Systems and methods for generation of validated machine control models and construction equipment control models

Abstract

Systems, methods, and computer readable media for machine control model generation and use are discussed. An example method includes receiving a construction plan. The method includes maintaining an accuracy of an AI Platform, the AI platform including an algorithm and a model, by performing periodic updated training of the model of the AI platform. The at model of the AI platform is trained based at least in part on prior engineering data, prior construction data, or any combination of the prior engineering data and the prior construction data. The method includes processing the construction plan using the AI platform. The method includes generating a machine control model by inputting at least one model input to the AI platform. The machine control model is generated based on the processed construction plan, the machine learning model of the AI platform, and the at least one model input.

Description

Patent Application Attorney Docket # 073233-00006SYSTEMS, METHODS, AND COMPUTER READABLE MEDIA FOR GENERATION OF VALIDATED MACHINE CONTROL MODELS AND CONSTRUCTION EQUIPMENT CONTROL MODELSCROSS-REFERENCE TO RELATED APPLICATIONS

[0001] This application claims priority to U.S. Provisional App. No. 63 / 688,056, filed August 28, 2024, and titled “SYSTEMS AND METHODS FOR THE PRODUCTION OF MACHINE CONTROL MODELS, and U.S. Provisional App. No. 63 / 765,059, filed February 28, 2025, and titled “SYSTEMS AND METHODS FOR THE PRODUCTION OF MACHINE CONTROL AND OTHER CONSTRUCTION EQUIPMENT MODELS,” the contents of which are incorporated by reference herein in their entirety.TECHNICAL FIELD

[0002] The technology described herein generally relates to the automated production of models, where a “model” as used herein refers to a digital representation of construction, engineering, or architectural data, including but not limited to Machine Control Models, layout models, site plans, architectural plans, indicate-only models, digital construction models, site guidance models, earthwork models, surface models, operational field models, GNSS-compatible models, or other digital design models that can be validated, standardized, and transformed into machine-executable or computer- interpretable formats for use in construction equipment, layout devices, or other engineering workflows (together, “Machine Control Models”) and, more particularly, to the generation and validation of Machine Control Models, site plans, and architectural plans using custom configured neural network and machine learning implementations trained based on targeted data, for example engineering standards, construction standards, local codes, existing plans, models, and surveys.Patent Application Attorney Docket # 073233-00006BACKGROUND

[0003] The production of Machine Control Models in various contexts is generally a complex, expensive, and disorganized effort resulting from inputs from multiple parties. The production of Machine Control Models also generally lacks accepted standards, which makes accuracy, risk assessment, and liability determinations challenging. Machine Control Model creation generally remains a substantially manual process that, once completed, may still result in unvalidated and / or inaccurate models regardless of the process followed.

[0004] The Machine Control Models can undergo validation to ensure accuracy and compliance with design specifications. Quality control checks can be performed to identify and correct any errors or discrepancies between the machine-readable format(s) and the detailed and certified design, construction, or architectural plans.

[0005] Validated Machine Control Models can then be uploaded to equipment for use. The equipment may be equipped with any number of 2D or 3D machine control technology, which uses or interprets the Machine Control Models and other hardware and software for use in controlling a machine according to the model. Equipment or machines outfitted with machine-control technology, current state-of-the-art, often need a unique machine-control model for every job or project to utilize or benefit from the machine-control technology. Additionally, proper equipment calibration and maintenance are often required to ensure that the machine control systems align with actual site conditions.

[0006] Current systems and / or methods can be limited by a lack of resources, understanding, accountability, attention to detail, accuracy, and workflow in creating Machine Control Models. For example, manual conversion from a model designed for human use to a model intended for machine use is costly, resource, and time intensive. Further, the accurate use of many existing methods often results in inaccurate models, a lack of standards (making models difficult to use in real time), or modelsPatent Application Attorney Docket # 073233-00006 containing too many data points to be used and read effectively by the limited computing power onboard construction equipment or machines. The impact of current methods results in a lack of machine control adoption, which further results in increased construction errors and rework, one of the leading causes of global carbon emissions. Further impacts include an increase in the time to complete projects, a risk to human life and safety, increased costs of all types of construction, and delays in improving infrastructure. Further still, the lack of timely, cost-effective, and accurate Machine Control Models limits the adoption and use of machine control technology.

[0007] Consequently, there is a need for improved systems and methods for the timely creation, generation, and validation of Machine Control Models and standards that offer automation, accuracy, consistency, and compatibility across various machine and software systems.SUMMARY

[0008] This summary is provided to introduce a selection of concepts in a simplified form that are further described below in the detailed description. This summary is not intended to identify key features or essential features of the claimed subject matter, nor is it intended to be used in isolation as an aid in determining the scope of the claimed subject matter.

[0009] Embodiments of the technology described herein are generally directed towards computing systems and methods for the generation and validation of Machine Control Models, construction plans, surveys, layout plans, paving plans, engineering plans, and the like, which incorporate the use of machine learning and a generation model, such as a neural network, which trained on existing collected data, such as plans, surveys, local regulations, and models.

[0010] According to some embodiments, a method is provided for the generation of Machine Control Models, site plans, architectural plans, and / or regulation validation. In some aspects, a methodPatent Application Attorney Docket # 073233-00006 comprises receiving a construction plan and survey data from a client, analyzing, by a machine control algorithm, the construction plan and other data, and generating a Machine Control Model or validation of an existing Machine Control Model based on the analyzed construction plan and one or more model inputs.

[0011] According to some embodiments, a system for the creation or validation of construction plans, layout plans, site plans and / or Machine Control Models, among others, comprises a plan and design recognition component configured to integrate data from one or more relevant sources and store the data as a set of collected data and a model build component configured to generate or validate a Machine Control Model based on a received construction plan and site data, wherein the Machine Control Model is generated or validated by a machine control algorithm, the machine control algorithm being trained at least in part on the set of previously collected data. In addition, a method is provided to correct irregularities in models (including but not limited to surface models) generated using Triangular Irregular Network and / or other algorithms and models, typical in existing design software algorithms, without oversaturating data points.

[0012] In some embodiments, a method is provided to correct irregularities in models (including but not limited to surface models) generated using Triangular Irregular Network algorithms and models, typical in existing design software algorithms, without oversaturating data points. In some aspects, a method is provided that can implement Al (e.g. an Al model, algorithm) to fix errors found in surfaces created with existing methods, using Triangular Irregular Networks.

[0013] Additional objects, advantages, and novel features of the invention will be set forth in part in the description that follows, and in part will become apparent to those skilled in the art upon examination of the following or can be learned by practice of the invention.Patent Application Attorney Docket # 073233-00006BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Aspects of the technology presented herein are described in detail below with reference to the accompanying drawings and figures, with emphasis upon clearly illustrating the principles of the disclosure, wherein:

[0015] FIG. 1 illustrates an example system in accordance with at least some embodiments of the present disclosure;

[0016] FIG. 2 illustrates an example apparatus for machine control model generation and use in accordance with at least some embodiments of the present disclosure;

[0017] FIGs. 3 A and 3B illustrate an example workflow for generalized model generation and use;

[0018] FIGs. 4A and 4B illustrates an example flowchart for machine control model generation and use in accordance with at least some embodiments of the present disclosure;

[0019] FIG. 5 illustrates a flowchart with operations of an example process for elevation, contour, and / or topography extraction in accordance with at least some embodiments of the present disclosure;

[0020] FIG. 6 illustrates a flowchart with operations of an example process for spot elevation data processing in accordance with at least some embodiments of the present disclosure;

[0021] FIG. 7 illustrates a flowchart with operations of an example process for BERT model use in accordance with at least some embodiments of the present disclosure;

[0022] FIG. 8 illustrates a flowchart with operations of an example process for CNN model use in accordance with at least some embodiments of the present disclosure;

[0023] FIG. 9 illustrates a flowchart with operations of an example process for GNN model use in accordance with at least some embodiments of the present disclosure;Patent Application Attorney Docket # 073233-00006

[0024] FIGS. 10A-10E illustrate a flowchart with operations of an example process for machine control model generation and use in accordance with at least some embodiments of the present disclosure;

[0025] FIG. 11 illustrates a flowchart with operations of an example process performed by a particular system or apparatus for machine control model generation and use in accordance with at least some embodiments of the present disclosure;

[0026] FIG. 12 illustrates a flowchart with operations of an example process for data processing and extraction in accordance with at least some embodiments of the present disclosure;

[0027] FIG. 13 illustrates a flowchart with operations of an example process for generation of a construction plan used in generation of a Machine Control Model in accordance with at least some embodiments of the present disclosure;

[0028] FIG. 14 illustrates a flowchart with operations of an example process for updating an Al platform based on user interaction with construction plans in accordance with at least some embodiments of the present disclosure;

[0029] FIG. 15 illustrates a flowchart with operations of an example process for direct machine control model generation and use in accordance with at least some embodiments of the present disclosure; and

[0030] FIG. 16 illustrates a flowchart with operations of an example process for generation of a plan, for example architectural plans or site plans, in accordance with at least some embodiments of the present disclosure.Patent Application Attorney Docket # 073233-00006DETAILED DESCRIPTION

[0031] The subject matter of aspects of the present disclosure is described with specificity herein to meet statutory requirements. However, the description itself is not intended to limit the scope of this patent. Rather, the inventors have contemplated that the claimed subject matter might also be embodied in other ways, to include different steps or combinations of steps similar to the ones described in this document, in conjunction with other present or future technologies. Moreover, although the terms “step” and / or “block” can be used herein to connote different elements of methods employed, the terms should not be interpreted as implying any particular order among or between various steps disclosed herein unless and except when the order of individual steps is explicitly described.

[0032] Accordingly, embodiments described herein can be understood more readily by reference to the following detailed description, examples, and figures. Elements, apparatus, and methods described herein, however, are not limited to the specific embodiments presented in the detailed description, examples, and figures. It should be recognized that the exemplary embodiments herein are merely illustrative of the principles of the invention. Numerous modifications and adaptations will be readily apparent to those of skill in the art without departing from the spirit and scope of the invention.

[0033] Disjunctive language such as the phrase “at least one of X, Y, or Z,” unless specifically stated otherwise, is otherwise understood with the context as used in general to present that an item, term, etc., may be either X, Y, or Z, or any combination thereof (e.g., X, Y, and / or Z). Thus, such disjunctive language is not generally intended to, and should not, imply that certain embodiments require at least one of X, at least one of Y, or at least one of Z to each be present.

[0034] Various terms are utilized throughout the disclosure to refer to particular subject matter within one or more contexts. “Construction plan” refers to at least one plan for construction, installation, or of a site for construction and / or installation. “Implementation” with respect to a construction planPatent Application Attorney Docket # 073233-00006 refers to a construction and / or installation that is performed in with details and / or data in the construction plan. For example, if a construction plan includes plans for construction of a building at a site, an implementation of the construction plan refers to a construction of that building at the site — or in other words the details and / or data defining the actual building as built. “Field data collector” refers to a remote sensor that is connected to or otherwise integrated with an Al platform to automatically collect at least one data point for at least one type of data, and automatically transmits or otherwise communicates the collected data points to the Al platform, or outputs such collected data points for collection by a data collector (e.g., a user) that provides the collected data points to the Al platform, for use in generating a machine control model. Non-limiting examples of a “field data collector” include a smartphone, sensor, loT device, laser scanners, imaging system, camera, photogrammetry device, drone, LiDAR system, total stations, GNSS / GPS device, and a device used to collect geospatial and / or construction site data. The term “auditable format” with respect to a log refers to data in a log that is provided in a manner suitable for subsequent review and / or insurance underwriting.

[0035] The term “control” should be understood broadly to include both operation of a machine (or “equipment” generally) and indicating actions for operating a machine. For example, “control” includes both autonomous and / or semi-autonomous action that is performed to operate a particular machine, and displaying of an interface or multiple interfaces that indicate or instruct an action (or actions) to be performed in operation of the machine.

[0036] The term “breakline” or “break line” refers to a visual depiction or symbol within a construction plan or drawing that indicate a lengthened portion of an object that is removed from the construction plan or drawing. For example, a breakline typically is comprised of a zig-zag that indicates a missing elongated portion of a line, which could be of any length or distance or includes an indicated length or distance, for ease of depiction. Additionally, in some contexts, multiple breaklines facing onePatent Application Attorney Docket # 073233-00006 another are presented facing one another to indicate a gap in a part of an object, structure, or the like depicted by the construction plan or drawing.

[0037] Creating Machine Control Models generally starts with surveying and / or scanning a construction site, conducted using global positioning systems (GPS), drones, laser scanning, optical scanning, or traditional surveying equipment. Additionally, topographic, soil, environmental, architectural, and other relevant site information may be collected further to understand a site's terrain and existing conditions.

[0038] Civil engineers, architects, and / or designers use the data collected and other data to create detailed design plans or a Digital Terrain Model (DTM), which represents a two-dimensional (2D) or three-dimensional (3D) surface of the site. The developed model can provide a foundation for further design and planning with respect to the site.

[0039] Construction design and modeling for use in a field (e.g., at a particular site to build a particular constructed building, for example) can be a tedious, complex, and often times multi-party endeavor. In some contexts, construction Design and / or BIM (Building Information Modeling) software can be implemented (e.g., AutoCAD Civil 3D, Tremble Business Center, Revit, Archicad, Bentley MicroStation, Bentley AECOsim Building Designer) to create the design plans and DTMs and then develop the sub-plans, which may include roadways, sewer, erosion control, utilities, grading, structural, and other infrastructure elements. Detailed design, construction, and architectural plans are intended for licensed and professional human use and interpretation and typically include differing (e.g., more, or less — but not the same amount of) details and accuracy than is utilized for robots or machines to consume. As a result, plans may be reviewed and augmented by machine control or grading professionals before being converted into machine-readable format(s) (e.g., Machine Control Models) which are compatible with construction, layout, survey, or engineering equipment or software, thePatent Application Attorney Docket # 073233-00006 formats generally often include 2D models, 3D models, line-work, and surface models that machine control systems of the construction, layout, survey, or engineering equipment can interpret and utilize for various functionality. In this regard, conventional construction plans may not be accurately or readily interpretable by a construction machine or operator of a machine via user interface, and in some aspects of the disclosure, embodiments of the present disclosure are specially configured to utilize an Al platform that generates Machine Control Models that are readily interpretable by particular systems, such as construction machines or other engineering / construction equipment, for use in accurately controlling such systems autonomously, semi-autonomously, and / or via operator control.Construed on / engineering equipment (“equipment”) may include various machines utilized for any of a myriad of engineering, construction, design, and / or building purposes, for example falling within categories of construction, paving, layout, survey, boring, framing, or bricklaying equipment, and the like, (for example but without limitation, excavators, asphalt pavers, paving systems, welding robots, timber framing machines, robotic concrete printers, robotic total stations, automated drones, GNSS / GPS based total stations, total station controlled machines, bulldozers, graders, robotic loaders, construction layout printers, and the like). Hardware and software, including GPS, sensors, onboard computers, and / or the like, can be used to inform the operator (e.g., in an indicate mode) or guide the equipment or machinery (e.g., in an automatic mode). Testing can be conducted to verify that the equipment accurately follows the Machine Control Models and performs the required tasks corresponding to the models and the detailed design plans.EXAMPLE SYSTEMS

[0040] FIG. 1 depicts an example system 100. The system 100 includes a machine control modeling system 102, equipment (e.g., a construction machine 106), and an optional user device 104. In somePatent Application Attorney Docket # 073233-00006 embodiments, one or more of the machine control modeling system 102, the construction machine 106, and / or the optional user device 104 are connected via a network 108. The network 108 in some embodiments is configured to enable data transmission between one or more systems and / or devices connected to the network. The network 108 in some embodiments embodies a wired and / or wireless connection between devices. In some embodiments, the network 108 embodies the Internet, an intranet, and / or a hybrid network. In some embodiments, the network 108 is optional, and one or more of the systems and / or devices may be integrated into a single system.

[0041] In some embodiments, the construction machine 106 includes or embodies one or more pieces of equipment and corresponding systems associated with performing one or more construction tasks. In some embodiments, the construction machine 106 includes one or more sub-systems that enable operation in accordance with embodiments herein. As illustrated, for example, in some embodiments the construction machine 106 includes an equipment system 106A, a hydraulics / movement system 106B, and a GPS system 106C. In some embodiments, the equipment system 106A is embodied by or includes one or more processors, memories, transceivers or communication devices, input / output devices, peripherals, and / or the like utilized for operating and / or otherwise controlling a construction machine. Additionally, or alternatively, in some embodiments, the equipment system 106A is configured to communicate with another device, for example the machine control modeling system 102, to retrieve and / or receive a Machine Learning Model utilized in controlling the construction machine 106. The construction machine 106 in some embodiments is an autonomous and / or semi-autonomous construction machine, such that operation of the construction machine may be performed at least in part automatically based on instructions from a control system, for example based on the equipment system receiving and / or processing the Machine Control Model received from the machine control modeling system 102 and utilizing the Machine Control Model toPatent Application Attorney Docket # 073233-00006 generate and / or determine such instructions. Additionally, or alternatively, in some embodiments, the equipment system 106A receives and / or retrieves a machine control model generated by the machine control modeling system 102, and utilizes the machine control model to output a user interface that indicates instructions to an operator where such instructions inform control of the construction machine. In some embodiments, the equipment system 106A includes or embodies a user interface, for example that is configured to provide user output (e.g., display data via the user interface) and / or receive user input (e.g., in response to user interaction with the user interface). In some embodiments, the equipment system 106A is onboard and / or otherwise integrated with the construction machine 106 to be controlled, for example via autonomous or semi -autonomous action initiated via the equipment system 106A and / or by instructions output to and performed by an operator of the construction machine 106. In other embodiments, the equipment system 106A is separate from the construction machine to be controlled, at least in part. Additionally, or alternatively, in some embodiments, the equipment system 106 A includes one or more sub-systems, for example. For example, in some embodiments, the equipment system 106A includes a CAN bus that communicates with electronic control units of the construction machine.Additionally, in some embodiments, the equipment system 106A includes a central processing unit and / or other processor that determines operations to be performed by the construction machine.

[0042] Additionally, or alternatively, in some embodiments the construction machine 106 includes one or more additional sub-systems utilized in performing construction tasks. For example, in some embodiments, the construction machine 106 includes hydraulics / movement system 106B. The hydraulics / movement system 106B includes hydraulics, actuators, specialized tools, and / or other components that are activated and / or moved to perform a particular construction task. For example, in some embodiments, the hydraulics / movement system 106B includes hydraulics that drive movement of a backhoe, excavator, bulldozer, and / or the like. Additionally, or alternatively, in some embodiments,Patent Application Attorney Docket # 073233-00006 the hydraulics / movement system 106B includes mechanical components, for example and without limitation steering systems, acceleration systems, braking systems, suspension systems, and the like, that enable driving of the construction machine 106 in a field. In some embodiments, the hydraulics / movement system 106B is controlled by the equipment system 106A (e.g., via instructions received and / or generated by the equipment system 106 A, and autonomously and / or semi-autonomously initiated, based on a Machine Control Model received or retrieved via the equipment system 106A). In some embodiments, the hydraulics / movement system 106B is controlled by an operator, for example based on instructions and / or a displayed interface based on a Machine Control Model received or retrieved by the equipment system 106A.

[0043] In some embodiments, the construction machine 106 includes a GPS system 106C. In some embodiments, the GPS system 106C is configured to determine and / or provide GPS data (e.g., GPS coordinates) representing a current position of the construction machine 106. The GPS data in some embodiments is utilized to control the construction machine 106 to operate along a particular path (e.g., autonomously or semi-autonomously). Additionally, or alternatively, in some embodiments, the GPS system 106C provides GPS data associated with the construction machine 106 to the machine control modeling system 102 for processing in generating a Machine Control Model, as discussed herein. In some embodiments, the GPS system 106C includes or embodies a laser / location system that utilizes the laser-based system for determining data representing a location of the construction machine.

[0044] The machine control modeling system 102 includes hardware, software, firmware, and / or a combination thereof that enables Machine Control Model generation and / or use. In some embodiments, the machine control modeling system 102 is embodied by at least one server, at least one database, and / or at least one custom configured computing system that performs and / or provides access to functionality for Machine Control Model generation and / or use. In some embodiments, for example, thePatent Application Attorney Docket # 073233-00006 machine control modeling system 102 is configured for generating and / or maintaining an Al platform.The Al platform is specially configured to generate a Machine Control Model from input files, for example inputted PDF and / or CAD files representing construction plans, where the Machine Control Model is specially configured to generate a corresponding Machine Control Model that is specially configured for processing by a particular machine, for example a construction machine as discussed herein. For example, in some embodiments the Al platform is specially configured to utilize one or more of the machine learning models, algorithms, and the like discussed herein to remove, modify, and / or add elements to inputted construction plans (e.g., PDF and / or CAD plans) and generate a corresponding Machine Control Model.

[0045] Additionally, or alternatively, in some embodiments, for example, the machine control modeling system 102 is configured for processing one or more construction plans, such as using an Al platform. Additionally, or alternatively, in some embodiments, for example, the machine control modeling system 102 is configured for generating a Machine Control Model as discussed herein.Additionally, or alternatively, in some embodiments, for example, the machine control modeling system 102 is configured for generating a construction plan, an architectural plan, and / or a site plan.Additionally, or alternatively, in some embodiments, for example, the machine control modeling system 102 is configured for receiving and / or maintaining particular specialized data utilized in generating an Al platform, generating a construction plan, an architectural plan, and / or site plan, generating a Machine Control Model, and / or performing sub-processes associated therewith.

[0046] In some embodiments, the user device 104 includes hardware, software, firmware, and / or a combination thereof that enables interaction with one or more applications, systems, devices and / or the like. The user device 104 embodies a system or single device that enables interaction with the functionality of the machine control modeling system 102. The user device 104 may be one or morePatent Application Attorney Docket # 073233-00006 peripherals, end terminals, and / or the like communicatively coupled with the machine control modeling system 102. Additionally, or alternatively, in some embodiments, the user device 104 embodies a separate device that communicates with the machine control modeling system 102 over the network 108 to initiate and / or interact with particular functionality of the machine control modeling system 102 (e.g., via an application executed or otherwise accessed via the user device 104, including a native application, web application, and / or the like). In some embodiments, the user device 104 is embodied by a smartphone, tablet, laptop, desktop, smart device, and / or the like communicable with the machine control modeling system 102. Additionally, or alternatively, in some embodiments, the user device 104 is embodied as a sub-system of the machine control modeling system 102.

[0047] FIG. 2 depicts a system architecture diagram of an example apparatus for machine control model generation and use, specifically a machine control model generation and use apparatus 200 (“apparatus 200”). In some embodiments, the apparatus 200 embodies the machine control modeling system 102 as depicted and described with respect to FIG. 1 herein. In this regard, the apparatus 200 in some embodiments is specially configured to perform one or more methods for machine control model generation and use, as described herein. For example, in some embodiments, the apparatus 200 is specially configured to perform the operations associated with the processes depicted and described further herein.

[0048] As illustrated in FIG. 2, the apparatus 200 includes a processor 202, a memory 204, input / output (“I / O”) circuitry 206, communications circuitry 208, model analysis circuitry 210, model generation circuitry 212, and model deployment circuitry 214. The apparatus 200 may be configured to execute some or all of the operations described herein with respect to machine control model generation and use. Although the circuitry 202-214 are described with respect to functional limitations, it should be understood that the particular implementations necessarily include the use of particular hardware. ItPatent Application Attorney Docket # 073233-00006 should also be understood that certain of these components 202-214 may include similar or common hardware. For example, two modules may both leverage use of the same processor, network interface, storage medium, or the like, to perform their associated functions, such that the duplicate hardware is not required for each individually named circuitry.

[0049] The use of the term “circuitry” as used herein with respect to the components of the apparatus will be understood to include particular hardware configured to perform the functions associated with the particular module circuitry depicted and described. The term “circuitry” should be understood broadly to include hardware, software that configures the hardware, firmware that configures the hardware, and / or any combination thereof. For example, in some embodiments, “circuitry” may include processing circuitry, storage media, network interfaces, input and / or output devices, and the like. In some embodiments, other elements of the apparatus 200 may provide or supplement the functionality of particular circuitry. For example, in some embodiments, the processor 202 provides processing functionality, the memory 204 provides storage functionality, the communications circuitry 208 provides network interface functionality, and the like.

[0050] It should be appreciated that, in some embodiments, some or all of the circuitry may be associated with a separate device, server, and / or associated computing hardware, which may be in communication with one or more of the other circuitry components of the apparatus 200. For example, in some embodiments, the model analysis circuitry 210, model generation circuitry 212, and / or model deployment circuitry 214 is included in and / or embodied by a separate computing apparatus. The separate computing apparatus in some embodiments may include a separate processor, memory, I / O circuitry, and / or communications circuitry.

[0051] In some embodiments, the processor 202 (and / or co-processors in some embodiments) is in communication with the memory 204 via a bus for passing information among components of thePatent Application Attorney Docket # 073233-00006 apparatus 200. The memory 204 may be non-transitory and in some embodiments includes one or more volatile and / or non-volatile memories. In other words, for example, the memory 204 in some embodiments is an electronic storage device (e.g., a computer-readable storage medium). The memory 204 in some embodiments is configured to store and / or provide access to data maintained by the apparatus 200, for example to enable the apparatus 200 to carry out various functions utilizing such data as described herein.

[0052] The processor 202 may be embodied in any of a myriad of different ways. For example, in some embodiments, the processor 202 includes one or more processing devices and / or sub-processors configured to perform independently. Additionally, or alternatively, in some embodiments the processor 202 may include one or more processors configured to operate in tandem via a bus, for example to enable independent execution of instructions, pipelining, and / or multithreading. The use of the terms “processing device,” “processor,” and / or “processing circuitry” may be understood to include a single core processor, a multi-core processor, multiple processors internal to the apparatus 200, and / or one or more separate, remote, and / or “cloud” processors.

[0053] In some embodiments, the processor 202 is configured to execute computer-coded instructions stored in the memory 204, or otherwise accessible to the processor 202. Alternatively, or additionally, the processor 202 in some embodiments is configured to execute hard-coded functionality. As such, whether configured by hardware or software, or by a combination of hardware with software, the processor 202 in some embodiments represents an entity (e.g., physically embodied in the circuitry) capable of performing operations in accordance with one or more embodiments of the present disclosure when configured accordingly. Alternatively, as another example, when the processor is embodied as an executor of software instructions, the computer-coded instructions in some embodiments specificallyPatent Application Attorney Docket # 073233-00006 configure the processor to perform steps described here, for example embodying one or more algorithms and / or operations thereof.

[0054] In some embodiments, the apparatus 200 includes I / O circuitry 206 that may, in turn, be in communication with the processor 202 to provide output to the user associated with the apparatus 200. Additionally, or alternatively, in some embodiments, the I / O circuitry 206 is in communication with the processor 202 to receive input from a user. The I / O circuitry 206 in some embodiments comprises a user interface, for example a device display, web interface, mobile application, client device, and / or the like. Additionally, or alternatively, in some embodiments, the I / O circuitry 206 includes one or more input devices, for example a keyboard, a mouse, a joystick, a touch screen, a microphone, and / or input / output mechanisms. The I / O circuitry 206, alone or together with the processor 202, in some embodiments controls one or more functions of the user interface, for example through executing computer-coded instructions stored on the memory 204 or otherwise accessible to the processor 202 (e.g., embodied in software and / or firmwarej.The communications circuitry 208 in embodied in hardware, or a combination of both hardware and software, that is configured for data receiving and / or data transmission, for example over a network. The communications circuitry 208 may transmit data from the apparatus 200, and / or receive data at the apparatus 200 from another device, system, and / or the like. In some embodiments, the communications circuitry 208 includes at least one network card, at least one antenna, at least one switch, at least one router, at least one modem, at least one bus connecting components, and / or supporting hardware and / or software of any such components. In some embodiments, the communications circuitry 208 is a separate device that is configured to enable the apparatus 200 to perform such data receiving and data transmitting. For example, in some embodiments, the communications circuitry 208 is configured to interact with at least one antenna to facilitate signal transmission, and / or facilitate signal reception via the at least one antenna. In some embodiments, thePatent Application Attorney Docket # 073233-00006 apparatus 200 is configured to communicate via the communications circuitry 208 utilizing any communications protocol, or a combination of multiple communications protocol. Non-limiting examples of such communications protocols include Bluetooth Low Energy, infrared wireless communication, ultra-wideband communication, Wi-Fi, Near Field Communication, Worldwide Interoperability for Microwave Access, and / or the like.

[0055] In some embodiments, the model analysis circuitry 210 includes hardware, software, and / or any combination thereof, that provides data receiving, retrieval, and analysis capabilities. In some embodiments, the model analysis circuitry 210 is specially configured to receive a construction plan, digital representations of real-world regulations and / or laws (e.g., zoning laws, technical standards, and / or the like), field data collector data such as model data points and equipment telemetry data (e.g., smartphone data, sensor data, loT device data, and / or the like), basic site data (e.g., survey files, preliminary design inputs, site condition data, and / or the like), geospatial data, positioning data, topographic data, and / or the like. Additionally, or alternatively, in some embodiments, the model analysis circuitry 210 is specially configured to perform one or more data processing operations that generate and / or modify data utilized in subsequent Machine Control Model generation and / or use. Nonlimiting examples of such data processing operations include processes for data removal from one or more construction plans, correcting one or more irregularities in data files, extracting data from one or more data files, comparing and / or altering data files with one another, vector-based PDF data processing, raster-based PDF data processing, inferring missing data from one or more files, extracting annotation data from one or more files, and / or the like. Additionally, or alternatively still, in some embodiments the model analysis circuitry 210 is specially configured for validating one or more files and / or models, validating one or more models (e.g., at least one Machine Control Model), and / or validating one or more plans. Additionally, or alternatively still, in some embodiments, the modelPatent Application Attorney Docket # 073233-00006 analysis circuitry 210 is specially configured for process data utilizing an Al platform, for example to perform training and / or configuration of one or more algorithms, models, and / or the like.

[0056] In some embodiments, the model generation circuitry 212 includes hardware, software, firmware, and / or any combination thereof, that is specially configured for generating a Machine Control Model. In some embodiments, the model generation circuitry 212 is specially configured for generating a Machine Control Model using an Al platform. Additionally, or alternatively, in some embodiments, the model generation circuitry 212 is specially configured for storing one or more generated Machine Control Models. Additionally, or alternatively, in some embodiments, the model generation circuitry 212 is specially configured to update at least one Machine Control Model based on updated, newly received, and / or other data. Additionally, or alternatively still, in some embodiments, the model generation circuitry 212 is specially configured to validate a generated Machine Control Model.

[0057] In some embodiments, the model deployment circuitry 214 includes hardware, software, firmware, and / or any combination thereof, that is specially configured for deployment of at least one Machine Control Model to at least one construction machine. For example, in some embodiments, the model deployment circuitry 214 is configured for transmitting a Machine Control Model to another device, system, and / or the like (e.g., a construction machine). Additionally, or alternatively, in some embodiments, the model deployment circuitry 214 is configured for initiating control of a construction machine. Additionally, or alternatively, in some embodiments, the model deployment circuitry 214 is configured for transmitting an updated (or new) Machine Control Model upon re-training and / or other updating of a Machine Control Model.EXAMPLE MODEL DEPLOYMENT MECHANISMS

[0058] FIGs. 3A and 3B depict an example workflow for generalized model generation and use. In some embodiments, existing mechanisms may perform the workflow to generate and deploy a modelPatent Application Attorney Docket # 073233-00006 used for controlling equipment. For example, the workflow may be performed manually (e.g., by human actors) alone or in combination with existing computing systems. As depicted, the workflow is performed based on actions by a surveyor 399A, a CAD technician 399B (e.g., responsible or otherwise assigned for erosion control plan functions), another CAD tech 399C (e.g., responsible or otherwise assigned for finishing a grade plan), a customer service manager 399D, and a customer themselves 399E. Each of these entities may be embodied by one or more human actors, with the customer 399E seeking to deploy the model.

[0059] A customer 399E may request a quote for a model at step 302. For example, in doing so, the customer may request a particular model to be generated. The customer 399E may then provide a construction plan at step 304. The construction plan includes or is embodied by a plurality of construction plan files, such as CAD and / or PDF files that indicate the requirements of the construction project. The construction plan may include erosion control plan files, grading plan files, site plan files, construction details, alignments, utility plan files, and / or other data indicating improvements to be performed. The customer service manager 399D may provide an estimated work time and price quote, and work with the customer 399E to provide an updated price quote as needed and / or ensure that proper construction plan files are submitted for processing. If the customer does not approve at step 308, no project is initiated (at step 310 and the workflow ends), however if the customer does agree then the work begins to create a model corresponding to the construction plan.

[0060] Once work begins, the CAD tech 399C transfers civil linework from the CAD files into new a project file (or multiple project files) at step 312. These files may be processed and / or transferred utilizing any of a myriad of existing software. At step 314, the CAD tech 399C assigns the linework to correct layers. Typically, this may be done by color-coding the linework, but in other cases might bePatent Application Attorney Docket # 073233-00006 assigned differently. The CAD tech 399C at step 316 additionally may remove duplicate or unnecessary lines, symbols, and / or texts, which the CAD tech 399C defines as not needed.

[0061] The CAD tech 399C at step 318 elevates the linework. For example, the CAD tech 399C may elevate the linework by transferring it from 2D to 3D. The CAD tech 399C at step 320 then draws break lines, for example lines marking bottom and top of all slopes, for example within data of the construction plan as depicted in the construction plan files.

[0062] At step 322, the CAD tech 399C creates trim boundary lines. For example, the trim boundary lines in some contexts represent edges of a model. The CAD tech 399C may again put the line markings and / or boundary lines at their discretion.

[0063] At step 324, the CAD tech 399C generates the surface file using PDF spot grades. The surface file in some contexts represents a 3D representation of ground topography in a particular site or otherwise associated with the construction project in the construction plan files. At step 326, the CAD tech 399C reviews the file for break line corrections as needed. If the review fails and break line changes are needed, the flow may return to step 320 (via the “no” line). If the review succeeds and it is configured that break line corrections are not needed (via the “yes” line), flow proceeds to step 328.

[0064] At step 328, another CAD tech. 399B elevates the linework. For example, the other CAD tech. 399B may raise the linework from 2D to 3D. At step 330, the other CAD tech. 399B draws breaklines. Specifically, the other CAD tech. 399B may similarly draw breaklines that mark a bottom and a top of all slopes within the construction plan submitted. The other CAD tech. 399B at step 334 then generates the surface file based on the generated data. For example, in contexts, the CAD tech.399B generates a surface file including or embodying a 3D representation of ground topography as generated above.Patent Application Attorney Docket # 073233-00006

[0065] At step 336, the CAD tech. 399B reviews the generated file, for example the surface file, for any breakline corrections. If breakline corrections are needed, again flow returns to step 330 (e.g., via the “no” line). If no corrections are needed, flow proceeds to step 338 (e.g. via the “yes” line).

[0066] The flow returns to customer 399E. At step 346, the customer 399E sends the model to internal or external surveyor for localization. The customer 399E may forward, transmit, or deliver the model to whichever surveyor the customer 399E desires, and using any manner desired. The customer 399E typically employs another third-party software application.

[0067] At step 348, the surveyor 399A collects GPS coordinates at the site associated with the model sent from the customer 399E. For example, in some embodiments, the surveyor 399A determines the site from data via inspection of the model, and collects GPS coordinates accordingly. The surveyor 399A then adds localization details to the model embodying a machine control file at step 350. The localization details may be specific to the GPS coordinates for the site corresponding to the model. The surveyor 399A then returns the updated model to the customer 399E.

[0068] The customer then, at step 352, may be ready to deploy the model. The customer 399E uploads a machine control plan (e.g., embodied by the model or based on the model) to the cloud. The cloud may enable a particular equipment to download the corresponding machine control plan. At step 354, the customer 399E downloads the model (e.g., embodying the uploaded machine control plan) to an equipment. For example, the model may be downloaded to a dozer or excavator. In some contexts, the model is downloaded directly from the cloud. The model may be downloaded automatically when a new uploaded model is detected, or automatically when a user (e.g., an operator of the equipment) triggers manual download. The operator may then use the downloaded model to control the equipment in accordance with the model. It will be appreciated that this process is functional, however the use of multiple distinct users as individual entities and the manual nature of the process can be improved.Patent Application Attorney Docket # 073233-00006Additionally, or alternatively, manual decisions in some contexts lead to improper selections, which could jeopardize the accuracy of the model and the use of the model to construct a project accordingly.AI-BASED MULTI-MODEL RECOGNITION, STANDARDIZATION, AND VALIDATION WORKFLOW FOR GENERATING MACHINE CONTROL MODELS FROMHETEROGENEOUS CONSTRUCTION PLAN DATA

[0069] According to aspects of the present technology, and at a high level, systems and methods are provided for (1) the creation or development of a Neural Network used to create Machine Control Models from existing engineering documents and / or construction plans automatically and further, in some instances, (2) monitoring and validation of the data created or predictions made by the Neural Network, and (3) continuously training the Neural Network based on all new data, data created or predictions made by the Neural Network, and other data uploaded by engineers or other users (all together, in some instances, the “Al Platform”).

[0070] In some aspects, the Al Platform in some embodiments includes a plan, design, building, material, survey, layout, or construction object recognition library or repository (e.g., the “Plan Recognition Library”). The Plan Recognition Library in some embodiments systematically gathers information (e.g., insights) or pieces of information from existing or previously developed models or plans by retrieving and / or receiving and integrating data and information from one or more relevant data sources, for example, one or more databases or via an Application Programming Interface (API), or by importing plans or data from 3rdparty design software applications (e.g., AutoCAD Civil 3D, Tremble Business Center, Revit, Archicad, Bentley MicroStation, Bentley AECOsim Building Designer). Accordingly, one or more datasets can be built which are diverse, accurate, and comprehensive. The Al Platform, or one or more components, can perform one or more operations on a dataset. In some instances, to prepare, cleanse, remove, and / or process data stored in the Plan Recognition Library,Patent Application Attorney Docket # 073233-00006 missing values, outliers, and duplicates in a dataset can be handled (e.g. via imputation of data, thresholding, deduplication, and / or the like), and a dataset can further be normalized and standardized with respect to any information or data points such that consistency in information and / or datasets stored in the Plan Recognition Library can be maintained. Data retrieved and / or gathered and / or collected and stored in the Plan Recognition Library (e.g., as one or more datasets) can then be labeled for use in a supervised or semi-supervised machine learning (ML) process or for instance, for use in training one or more neural networks (e.g., convolutional neural networks, generative adversarial neural networks, or other machine learning framework). One or more datasets, or labeled datasets, in some embodiments are stored in the Plan Recognition Library and subsequently used in one or more systems and or processes to generate Machine Control Models. The Plan Recognition Library in some embodiments is trained or further updated (e.g., continuously or batch update) to further enrich and / or process the data by adding new plans and values, which can improve the scope and quality of image recognition over time.

[0071] In some aspects, the Al Platform can include a machine control algorithm that can design, create, or otherwise generate Machine Control Models based on one or more inputs. In some aspects, a newly generated Machine Control Model is based at least in part on a previously created Machine Control Model, and in some instances, the newly generated Machine Control Model is stored in a database or repository. In some aspects, a Machine Control Model is generated via the machine control algorithm based on defined engineering principles, specific parameters, zoning data, structural data, architectural plans, and / or geospatial considerations. A machine control algorithm in some embodiments analyzes or utilizes any number of inputs in the generation or creation of a Machine Control Model, for example, inputs or data relating to site-specific data, regulatory requirement data, and best practice data, among other data to generate Machine Control Models, or other plans and / or designs. In some aspects, the models, designs, and / or plans can be optimized with respect to any determined parameter, andPatent Application Attorney Docket # 073233-00006 compliant with respect to rules or regulations. As will be appreciated, by integrating one or more of various parameters into the generation of a Machine Control Model, for instance soil composition, topography, structural conditions, climate conditions, infrastructure needs, and / or the like, the machine control algorithm in some embodiments ensures that each newly generated Machine Control Model is specially configured for any number of unique characteristics of a project location (e.g., a site). As will be appreciated, the machine control algorithm in some embodiments continuously learns and adapts over time from the growing data pool, thereby refining its output and, therefore, producing more efficient, accurate, and improved design solutions over time.

[0072] In some aspects, the Al Platform includes a model build component, a monitoring component, and a maintenance component, and may include other components. As described above, a machine control algorithm can be implemented to develop an initial version of a model (e.g., an initial model, or a “first” model). The machine control algorithm (or in some embodiments, a set of algorithms, which can be implemented simultaneously or in series) in some embodiments incorporates various machine learning techniques and domain-specific knowledge (e.g., in the field of engineering, construction, building, architecture, and / or the like) in order to generate accurate designs that can be relied upon by equipment, for example which run such Machine Control Models. In order to drive consistency and development of standards, new models (e.g., Machine Control Models) can be developed using existing, imported, or retrieved data and datasets. During a model build or generation, for instance, via the model build component, the existing data used to build the model in some embodiments is divided or portioned into two or more subsets, for example, training data, validation data, and test data sets. As will be appreciated, this process in some embodiments is implemented to ensure that the model is trained, validated, and / or tested on distinct portions of data or a dataset to prevent bias and overfitting. In some aspects, the monitoring component monitors the performance ofPatent Application Attorney Docket # 073233-00006 the Machine Control Model, and in some instances provides monitoring in real-time, and further set ups and provides any alerts for any deviations or issues associated with the model. In some aspects, the maintenance component analyzes and updates the model (e.g., at regularly intervals, irregular intervals, and / or ad-hoc) with new data, and in some embodiments further retrains the model as needed, and in some instances implements a feedback loop to continuously improve the model, based on additional or new data, performance analysis, regression testing, and / or other factors.

[0073] According to some aspects, a validation component and / or engine is provided, for example a zoning and code validation component. In some instances, zoning and code validation components is implemented to test one or more aspects of compliance with construction, building, and site plans and / or generated construction, building, and site plans. In some instances, zoning and code validation components in some instances incorporates or calls an artificial intelligence (Al) software program, model, and / or module (for example, a generative Al program or model) to test one or more aspects of compliance of construction and site plans and / or generated construction and site plans against, for example, federal and / or local zoning and planning laws, building codes, restrictions, and / or rules. In some instances, zoning and code validation components search one or more third-party database(s) for zoning and planning laws, restrictions, and / or rules for use as a validation basis. In some aspects, zoning and code validation component can incorporate or use a crawler tool (e.g., a web crawler or other type of network crawling device) to find one or more data records in one or more data sources associated with any number of zoning and / or planning laws, or regulations, or rules.

[0074] FIGs. 4A and 4B depict an example flowchart for machine control model generation and use in accordance with at least some embodiments of the present disclosure. The flowchart embodies an example process that may be performed in accordance with at least some embodiments of the present disclosure. Additional details and / or embodiments in accordance with the details, and / or includingPatent Application Attorney Docket # 073233-00006 narrowed or specific implementation details depicted and described herein, may be performed as discussed further below. The process is performed between a system including one or more sub-systems, devices, and / or components embodying an equipment 499 A, and / or configured for model creation 499B, intelligent line work 499C, and plan recognition 499D, and enabling performance of operations by a user, for example a CAD engineer 499E. The system configured for model creation 499B, intelligent line work 499C, and / or plan recognition 499D in some embodiments includes or is embodied by an Al platform as depicted and described herein.

[0075] At step 402, a user (e.g., the CAD engineer 499E) scans and loads (or “uploads”) a construction plan into the interface, for example of the system including or embodying the Al platform. At step 404, if the system (e.g., via a component for plan recognition 499D) does not identify objects and layers, flow proceeds to step 406. The system may attempt to identify objects and layers at step 404 via use of a plan recognition library 408, for example configured based on previously submitted construction plans and / or associated data. At step 406, the system (e.g., via a component for plan recognition 499D) trains an Al model to recognize new objects and layers. In some embodiments, the Al platform is trained based on data stored to the plan recognition library 408, for example as depicted and described herein.

[0076] If the system (e.g., via a component for plan recognition 499D) identifies objects and layers at step 404, flow proceeds to step 410 (e g., via the “yes” line). At step 410, the system (e.g., via a component for intelligent line work 499C). At step 410, the system (e.g., via a component for intelligent line work 499C) separates objects and layers. In some embodiments, the objects and layers are separated via one or more machine learning and / or Al workflows, for example by the Al platform. At step 412, the system (e.g., via a component for intelligent line work 499C) validates the separated objects and layers.Patent Application Attorney Docket # 073233-00006The objects and layers may be validated with authenticated data, for example to confirm accuracy of the separated objects and layers.

[0077] If the objects and layers are not successfully validated, flow proceeds to step 414 (e.g., via the “no” line). A user, for example the CAD engineer 499E, interacts with the system including or embodying the Al platform to add the missing layers. Once added, the system (e.g., via a component for plan recognition 499D) continues to train the Al model to recognize objects and layers (e.g., the newly added layers), and the process proceeds as discussed above with respect to step 406. If the system (e.g., via a component for intelligent line work 499C) does validate the objects and layers, flow proceeds to step 416 (e.g., via the “yes” line).

[0078] At step 416, the system (e.g., via a component for intelligent line work 499C) removes duplicates or unnecessary lines, symbols, and text. In some embodiments, the Al platform, or a machine learning model or algorithm thereof, determines the lines, symbols, and / or text to remove, and performs the removal, as discussed herein. At step 418, the system (e.g., via a component for intelligent line work 499C) elevates the line work, for example by transferring from 2D to 3D. Additionally, at step 420, the system (e.g., via a component for intelligent line work 499C) draws breaklines, for example marking the bottom and the top of all slopes in the submitted construction plan as separated and updated. At step 422, the system (e.g., via a component for intelligent line work 499C) creates trim boundary lines, for example embodying edges of the model. In some embodiments, steps 418-422 are performed automatically by the system via at least one machine learning, Al, and / or other custom-trained algorithm embodying a workflow specific to the sub-process being performed.

[0079] At step 424, the system (e.g., via a component for intelligent line work 499C) generates the surface file using PDF spot grades. The PDF spot grades include 3D representations of ground topography. At step 426, the system (e.g., via a component for intelligent line work 499C) compares thePatent Application Attorney Docket # 073233-00006 layers in the created surface file with one or more standards, for example layer standards. The layer standards may be identified and / or retrieved from a machine control standards library 428. The machine control standards library 428 may include standards data for different jurisdictions, project types, and / or the like, such that appropriate standards for a particular project at a particular site and jurisdiction may be retrieved and utilized.

[0080] If the system fails to compare the layer standards successfully, flow proceeds to step 430(e.g., via the no line). At step 430, the system (e.g., via a component for intelligent line work 499C) amends the construction plans to recreate the machine control plan as currently generated (e.g., modified based on the steps 410-424). Flow may then return to the user (e.g., the CAD engineer 499E) for further processing. If the system (e.g., via a component for intelligent line work 499C) does compare the layers standard successfully, flow proceeds to step 432 (e.g., via the “yes” line).

[0081] At step 432, the system (e.g., via a component for model creation 499B) completes the CAD Machine Control Model. For example, the system (e.g., via a component for model creation 499B) in some embodiments generates one or more files embodying the Machine Control Model including the construction plan as modified and / or updated based on the prior steps. At step 434, the system (e.g., via a component for model creation 499B) inputs GPS coordinates associated with the Machine Control Model, which may be included in and / or extracted from the Machine Control Model (e.g., from the construction plan originally inputted, and / or other data associated therewith). At step 436, the system (e.g., via a component for model creation 499B) adds localization details for the Machine Control Model, for example the localization details may be based on the GPS coordinates associated with the Machine Control Model. In this regard, the Machine Control Model may be localized to ensure compliance with standards, regulations, laws, and the like in a particular location of the site associated with the Machine Control Model and / or site plan associated therewith.Patent Application Attorney Docket # 073233-00006

[0082] At step 438, the system (e.g., via a component for model creation 499B) generates theMachine Control Model plan in a desired manufacturer format. In some embodiments, the desired manufacturer format is specified, for example in the construction plans or other data inputted by a user. Additionally, or alternatively, in some embodiments, the desired manufacturer format is identified for the corresponding equipment 499A that is to receive the Machine Control Model.

[0083] In some embodiments, the generated Machine Control model is returned to the user, for example the Engineer 499E. The generated Machine Control Model may be manually deployed and / or used to instruct operation of one or more equipment, for example the equipment 499A. The equipment in some embodiments may be a construction machine, such as an excavator or a dozer or the like. In some embodiments, the system then transmits the Machine Control Model, for example embodying the Machine Control Model plans, to the equipment 499A. The Machine Control Model may be automatically transmitted to deploy the Machine Control Model, for example to cause the equipment 499A to operate in accordance with the Machine Control Model, and / or to instruct a user (e g., an operator) of the equipment 499A how to operate the equipment 499A in a manner consistent with the Machine Control Model. For example, the system may transmit to the equipment 499A via a public and / or private network, the Internet, a wired connection, and / or the like. As depicted, the system (e.g., via a component for model creation 499B) uploads the Machine Control Model to the cloud at step 442. The equipment 499A then automatically (or in response to a manual interaction in some embodiments) downloads the Machine Control Model to the equipment 499A at step 444.GENERATION OF MACHINE CONTROL MODELS

[0084] According to some aspects of the present technology, and at a high level, systems and methods are provided for the automatic generation of Machine Control Models. As will be appreciated, a Machine Control Model (or a set of Machine Control Models) can be uploaded or transmitted to aPatent Application Attorney Docket # 073233-00006 piece of equipment, which can interpret and run the Machine Control Models. The transmitted MachineControl Model, in some embodiments, generally control various aspects of the equipment or otherwise inform the equipment operator. In some instances, Machine Control Models can control and coordinate a plurality of equipment at a given site or location with respect to a given project or detailed design plan. Accordingly, in some instances, systems and / or methods described herein in some embodiments provide for the automation of Machine Control Model generation and in some embodiments further manage other aspects of an engineering and / or construction workflow, such as managing machine angle and depth but also can also provide oversight to an entire excavation or grading process and manage productivity and workflow.

[0085] In some aspects, the Al Platform enables and handles model deployment by way of a model deployment component, which in some embodiments leverages cloud and edge computing techniques that in some embodiments enable model development and deployment anywhere with any number of user devices, and / or which in some instances ensure scalability, reliability, and security of the system and the generated models. In some aspects, the Al Platform is configured to enable the development of open APIs or interfaces for interaction with one or more generated models and / or the machine control algorithm that in some embodiments connect other applications, platforms, devices, and solution providers.ADDITIONAL IMPLEMENTATION ASPECTS

[0086] According to some other aspects, systems and / or methods described herein in some embodiments utilize and / or integrate other software and tools, such as for three-dimensional (3D) modeling, Building Information Modeling (BIM), and Geographic Information Systems (GIS). In some other aspects, systems, and / or methods described herein can ensure or otherwise provide compatibility and integration between various disparate software systems and platforms and, in some instances, ensurePatent Application Attorney Docket # 073233-00006 or otherwise provide the precision of site surveys and measurements and / or further incorporate detailed data (e.g., topographical data, structural data) into existing and / or designed infrastructure.

[0087] According to some aspects, systems and / or methods according to the present technology can provide additional improvements in engineering and building (for example, in the civil engineering context) for instance, enabling the identification of potential risks and developing or forming mitigation strategies to ensure safety and compliance with various standards, such as occupational health standards; navigating complex zoning laws, building codes, environmental regulations, and securing various permits and approvals from issuing authorities; addressing issued related to soil stability, water drainage, and impact on local ecosystems and infrastructure to provide sustainable design capabilities and minimize environmental impact; and improve time to market for project timelines and workplan.

[0088] In some aspects, the Al Platform can include one or more security and / or governance components. For example, the Al Platform in some embodiments incorporates one or more build-in governance mechanisms to ensure compliance with relevant regulations and standards. In one example, each model undergoes review by a user process to ensure accuracy and accountability. In some instances, noted deviations trigger a comprehensive model variation analysis to identify and rectify any discrepancies found. In some instances, models undergo a variation assessment as some given time period (e.g., every quarter) or ad-hoc to maintain consistency and reliability. In some aspects, the Al Platform incorporates a reporting component that can report issues with a model either internally within the system, or to an external third-party system or application. In some instances, the Al Platform incorporates one or more security components, which can be implemented to safeguard any data and models, and / or which include encryption, access control, and other data security protocols. Additionally, in some instances, security audits and assessments are conducted to identify and mitigate any issues or vulnerabilities in the Al Platform, associated data, and associated functionality.Patent Application Attorney Docket # 073233-00006

[0089] According to some aspects, the Al Platform provided herein provides digitization and automation capabilities for machine control (e.g., Machine Control Model generation), as well as layout models, take-offs, site (e.g., a build site) and architecture, engineering, and / or design plans.

[0090] In some aspects, the Al Platform implements a multi-stage, Al-driven workflow for automatically generating standardized, validated Machine Control Models (MCMs) from heterogeneous construction plan inputs, including, but not limited to, PDF and CAD formats. This workflow enables rapid, accurate, and consistent transformation of engineering documents into manufacturer-compatible MCMs for deployment to construction equipment.

[0091] In some aspects, the process begins when a user, via a native, browser-based, or other interface to the AECAD platform, creates a project and uploads one or more source fdes. The files may include CAD files (e.g., DWG) and / or PDF files containing plan data in either vector or raster form. For CAD inputs, an ODA converter in some embodiments is used to extract objects and layer data, while PDF inputs in some embodiments are processed via distinct raster and vector pipelines to capture all relevant graphical and textual information. In some aspects, this includes the Elevation, Contour / Topography Extraction Process (for example, see FIG. 5) and the Spot Elevation Data extraction process (for example, see FIG. 6), which together enable accurate identification and mapping of contours, elevations, and spot grades from PDF sources.

[0092] Continuing to FIG. 5, a flowchart with operations of an example process for elevation, contour, and / or topography extraction (“E / C extraction”) is depicted. The flowchart represents an example process 500. The process 500 in some embodiments is implemented by the system (e.g., including or embodying the Al platform as discussed herein). In some embodiments, an Al platform embodied by or included as part of the machine control modeling system 102 is specially configured toPatent Application Attorney Docket # 073233-00006 perform the process 500 as part of a process for generating a Machine Control Model, for example as discussed herein with respect to FIGs. 10A-10E.

[0093] At step 502, one or more systems (e.g., a service manager for elevation and / or contour / topography extraction) is initiated to control and / or manage PDF processing. In some embodiments, the service manager is embodied by the Al platform.

[0094] In some embodiments, the file to be processed (e.g., the construction plan) is embodied as a DXF file 504. Alternatively, in some embodiments, the file is embodied as a PDF file, such that PDF to DXF conversion is performed at step 506. The DXF version of the file is then processed as depicted and described herein.

[0095] At step 508, the E / C extraction process 500 includes layer mapping of the DXF file. The various layers may be mapped for further processing of individual layers and / or data therein. At step 510, the E / C extraction process 500 includes identifying contours. In some embodiments, the contours are identified from the mapped layers, for example where the contours represent elevation changes in a site. Additionally, or alternatively, at step 512, the E / C extraction process 500 includes raster image elevation detection. For example, raster-based processing of the DXF file may be performed to detect features representing elevation changes depicted in image data of the rasterized DXF file. At step 514, the E / C extraction process 500 includes initiating an OCR model 514. The OCR model may be used to detect, identify, and / or convert text in the DXF file. At step 516, the E / C extraction process 500 includes mapping the contours with the identified elevations to coordinates of the file (e.g., PDF) coordinate system). The mapping of the contours to the coordinate system of the file enables the depicted contours to be paired with processable data representations accordingly.Patent Application Attorney Docket # 073233-00006

[0096] The E / C extraction process 500 in some embodiments further includes pipeline 516. In some embodiments, the pipeline 516 includes one or more Al and / or machine learning models, and / or algorithms, specifically trained for E / C extraction from the PDF and / or DXF version of a file.

[0097] At step 518, E / C extraction process 500 renders the DXF to an image (e.g., a PNG) and Captures data and display transforms. At step 520, the E / C extraction process 500 includes running an OCR model. At step 522, the pipeline includes processing for each detection by the OCR model. For example, for each detection the pipeline of the E / C extraction process 500 includes step 524 of normalizing a box to quad and centroid pixels. The pipeline further includes processing by converting pixels model space using the transforms at step 528. The pipeline further determines whether a confidence of such a conversion satisfies an applicable threshold, such as where a confidence score exceeds (or is greater than) the threshold. If not, the flow returns to step 522 for the next detected element. If the confidence value does exceed the applicable threshold, flow proceeds to step 532. At step 532, the pipeline of the E / C extraction process 500 includes collecting text, scores, the bounding box, and centroids for each detection. The collected data may form a particular set of collected data for a particular file and / or construction plan thereof.

[0098] Once processing each detection is completed, the E / C extraction process 500 saves the data collected to an outputted file. For example, in some embodiments, the E / C extraction process 500 includes step 526 of saving the collected data to a JSON file. It should be appreciated that in other embodiments, implementations for file saving other than JSON may be used.

[0099] Continuing to FIG. 6, a flowchart with operations of an example process for spot elevation data processing. The flowchart represents an example process 600. The process 600 in some embodiments is implemented by the system (e.g., including or embodying the Al platform as discussed herein). In some embodiments, an Al platform embodied by or included as part of the machine controlPatent Application Attorney Docket # 073233-00006 modeling system 102 is specially configured to perform the process 600 as part of a process for generating a Machine Control Model, for example as discussed herein with respect to FIGs. 10A-10E.

[0100] At step 602, one or more systems (e.g., a service manager for spot elevation data processing) is initiated to control and manage PDF processing. In some embodiments, the service manager is embodied by the Al platform.

[0101] In some embodiments, the file is embodied as a PDF file, such that PDF to DXF conversion is performed at step 604. The DXF version of the file is then processed as depicted and described herein.

[0102] At step 606, the spot elevation data process includes performing spot detection. A spot detection machine learning model, algorithm, and / or Al model in some embodiments is initiated for processing accordingly. At step 608, the spot elevation data process includes initiating a YOLO model. In some embodiments, the YOLO model performs object detection of one or more object types within the DXF-converted data.

[0103] Alternatively, or additionally, in some embodiments the spot elevation data process includes raster image spot grade detection at step 610. In this regard, rasterized data in the DXF file may be processed to perform spot gradation within such data. At step 610, the spot elevation data process includes initiating an OCR model 612. The OCR model may be utilized to detect characters and / or text within the DXF file.

[0104] In some aspects, the platform’s feature extraction process identifies and classifies existing text, layers, images, and objects. The system employs multiple, independently trained Al models, including: (i) a BERT -based model for text-based layer classification (see, e.g., FIG. 7); (ii) a convolutional neural network (CNN) for object and image-based classification (see, e.g., FIG. 8); and (iii) a proprietary graph neural network (GNN) for Object classification using nodes, edges, and graphs (see, e.g., FIG. 9). In some embodiments, the CNN model is implemented within the Al Layer CreationPatent Application Attorney Docket # 073233-00006Model to generate standardized layers for the Machine Control Model and is trained to process raster and vector data for object identification, classification, and grouping into the appropriate standard layers. The CNN Model Use process provides a summary of CNN operation across the workflow, including input preprocessing, feature extraction, and final classification stages.

[0105] FIG. 7 depicts a flowchart with operations of an example process for BERT model use. The flowchart represents an example process 700. The process 700 in some embodiments is implemented by the system (e.g., including or embodying the Al platform as discussed herein). In some embodiments, an Al platform embodied by or included as part of the machine control modeling system 102 is specially configured to perform the process 700 as part of a process for generating a Machine Control Model, for example as discussed herein with respect to FIGs. 10A-10E.

[0106] In some embodiments, the process 700 is performed by one or more sub-layers, for example a model training layer 750A and a data processing layer 750B. The different layers in some embodiments are embodied by different sub-systems and / or devices, or in some embodiments are embodied by different functionality of the same sub-system. In some embodiments, the two layers function in conjunction with one another to complete model training and use of the BERT model.

[0107] At step 702, model training data is obtained. In some embodiments, the model training data is pre-stored and / or collected. In some embodiments, the model training data is received from an external system. The model training data may include text data for processing.

[0108] At step 704, label encoding is performed. For example, the model training data may be labelled for training, and the labels may be encoded accordingly for processing. In some embodiments, the labels are manually applied (e.g., by a separate user or data scientist), and / or in some embodiments the labels are processed and automatically determined and applied to records.Patent Application Attorney Docket # 073233-00006

[0109] At step 706, a training dataset is created. In some embodiments, the model training data of step 702 and the labels encoded in step 704 are utilized and / or combined to create the dataset at step 706. In some embodiments, the dataset is created for a particular platform enabling training and / or use of a trained BERT model.

[0110] At step 708, tokenization occurs for the created data set. For example, in some embodiments, the data elements of the created dataset are tokenized for inputting to the model during training. At step 710, the dataset is split into a test set and a training set. For example, the tokenized version of the created dataset may be split into a training set and a test set.

[0111] The model and training layer 750A begins a subprocess at step 712 and / or 714, which represent data received for training. At step 712, a BERT encoder is received, retrieved, or otherwise identified. The BERT encoder may be stored by the system, or in some embodiments is received from an external system. Additionally, or alternatively, at step 714 training configuration and training arguments are received, retrieved, or otherwise identified. In some embodiments, the training configuration and / or training arguments are pre-stored or previously received. In some embodiments, the training configuration and / or training arguments are retrieved from a separate system.

[0112] At step 716, trainer orchestration is performed. The training orchestration performs training of the BERT encoder identified at step 712, specifically utilizing the training configuration and training arguments 714 and the test set and training set of data. For example, the BERT encoder may be specially configured based on the training configuration and training arguments at step 714, and trained to account for the training data represented in the training split of the dataset identified at step 710. The BERT model may be trained in one or more phases based on the training set identified at step 710, and validated based on the test set of the data identified at step 710.Patent Application Attorney Docket # 073233-00006

[0113] At step 718, an evaluation loop is performed. In some embodiments, the evaluation loop validates the BERT model trained at step 716. The evaluation may confirm that the accuracy and / or other metrics associated with the training of the BERT model satisfy applicable thresholds for use.

[0114] At step 720, the model and tokenization are saved. For example, in some embodiments, the model and tokenization are saved in response to the evaluation loop 718 indicating that the model satisfies applicable thresholds for one or more metrics.

[0115] In some embodiments, the model and tokenizer are exported. For example, at step 722, in some embodiments the model and tokenizer are exported to a particular framework. In some embodiments, the BERT model is exported to convert the neural network to an Open Neural Network Exchange format (ONNX).

[0116] Continuing to FIG. 8, FIG. 8 depicts a flowchart with operations of an example process for CNN model use. The flowchart represents an example process 800. The process 800 in some embodiments is implemented by the system (e.g., including or embodying the Al platform as discussed herein). In some embodiments, an Al platform embodied by or included as part of the machine control modeling system 102 is specially configured to perform the process 800 as part of a process for generating a Machine Control Model, for example as discussed herein with respect to FIGs. 10A-10E. In some embodiments, the process 800 includes one or more sub-processes, for example a data preparation process, a model structure process, a model training process, and a model evaluation process.

[0117] At step 802, the process includes step 802 of gathering images and labels. At step 804, the process includes extracting contextual metadata. At step 806, the process includes normalizing metadata features.Patent Application Attorney Docket # 073233-00006

[0118] At step 808, the process includes a setting up a data loading pipeline. In some embodiments, the data loading pipeline is set up including an augment option or an original option. The process may continue to step 810 and / or step 812 based on the setup.

[0119] In a circumstance where the data loading pipeline is setup with an original option selected, at step 812 the original images are used for further processing. Alternatively, or additionally, in a circumstance where the data loading pipeline is setup with an augment option selected, at step 810 the images are augmented and mixed up. For example, in some embodiments, the images may be augmented from the set of images gathered at step 802. In some embodiments, the images are augmented to remove, alter, and / or update one or more features or portions of data within the images. The augmented images may then be mixed up for further processing.

[0120] At step 814, the images are batched and prefetched. The batched and prefetched images may be made available for further processing. The sub-process for model structuring may then be initiated.

[0121] The sub-process begins at step 816 and / or step 818, which may be performed in parallel. At step 816. An image branch is initiated at step 816. At step 816, a pre-trained CNN is accessed. The pretrained CNN may be specially configured to process the images and identify and / or extract sub-features thereof. At step 820, the process continues with pooling visual features. The visual features may be extracted via the pre-trained CNN.

[0122] At step 818, a metadata branch begins via a small network. For example, the process includes processing metadata features. In some embodiments, the small network processes the metadata features to extract the metadata features from the images as processed in the first sub-process.

[0123] At step 824, the process includes combining visual and metadata features extracted from the images. For example, the visual features identified by the CNN branch and the metadata featuresPatent Application Attorney Docket # 073233-00006 identified from the metadata branch may be re-combined for each image processed via both branches. A set of combined images may then be generated.

[0124] At step 826, the process includes decision layers that refine features of the combined images. The decision layers may be specially configured to refine particular features, where the decision layers are specially configured, trained, or otherwise trained to refine such features. For example, features values may be refined upwards or downwards by the decision layer.

[0125] At step 828, the process includes predicting class probabilities. In some embodiments, the model is configured to utilize the feature values for the class prediction. A training sub-process may then be started.

[0126] The training sub-process begins at step 830. At step 830, feedback for further tuning is received. The feedback in some embodiments is associated with current features utilized by the model and / or accuracy of the model processing the training data. In some embodiments, the feedback is received in response to or based on user input.

[0127] At step 832, stage 1 of a training process is initiated. For example, the stage 1 may include training new classification layers. The new classification layers in some embodiments are trained based at least in part on the data and / or features of the data preparation process discussed above. At step 834, one or more intermediate training functions performed during training are monitored. For example, an early stop, ReduceLR, and a checkpoint may be monitored to continue training and, if one or more conditions are satisfied, terminate the training early.

[0128] At step 836, stage 2 of a training process is initiated. For example, the stage 2 may include fine-tuning part of the CNN. The fine-tuning in some embodiments includes adjusting one or more of the nodes within the CNN. At step 838, again one or more intermediate training functions performedPatent Application Attorney Docket # 073233-00006 during training are monitored. For example, an early stop, ReduceLR, and a checkpoint may be monitored to continue training and, if one or more conditions are satisfied, terminate the training early.

[0129] In some embodiments, iterations of training are repeated until convergence. For example, in some embodiments, until convergence the process may return to step 830 and repeat until convergence. Additionally, or alternatively, in some embodiments, the process may return to operation 828 and repeat until convergence, for example, where class probabilities are determined to be updated. Once training of the model is completed (e.g., at / after convergence), the best saved model may be saved for subsequent use.

[0130] The process continues to an evaluation sub-process. The evaluation sub-process begins at step 840. At step 840, the process includes loading a best saved model. The best saved model may be the model that performed the best with respect to one or more metrics. For example, a stored model representing the best saved model at a given time may be retrieved from a database.

[0131] At step 842, the process includes evaluating the model on a held-out test set. For example, the test set may be inputted to the retrieved model, and accuracy of performance of the model may be tracked for the samples in the test set. An overall accuracy of the model may thus be determined accordingly.

[0132] In some embodiments, the sample predictions (e.g., generated during testing) for the model are shown. For example, the model predictions may be outputted to a user via one or more interface. In some embodiments, the sample predictions are shown in an interface configured to receive user input representing feedback for the model. The feedback provided by the user may be utilized to further train the model, for example returning the process to step 830 as depicted and described herein. This process may continue until convergence, or until a user indicates that the model has reached a sufficient training accuracy.Patent Application Attorney Docket # 073233-00006

[0133] FIG. 9 illustrates a flowchart with operations of an example process for GNN model use. The flowchart represents an example process 900. The process 900 in some embodiments is implemented by the system (e.g., including or embodying the Al platform as discussed herein). In some embodiments, an Al platform embodied by or included as part of the machine control modeling system 102 is specially configured to perform the process 900 as part of a process for generating a Machine Control Model, for example as discussed herein with respect to FIGs. 10A-10E. In some embodiments, the process 900 includes one or more sub-processes, for example a data preparation process, a model structure process, a model training process, and a model evaluation process.

[0134] At step 902, the process includes DXF ingestion. The DXF ingestion includes extracting and / or otherwise inputting, from a DXF file, entities and / or labels for processing. In some embodiments a DXF file is inputted via upload from a user, or generated from inputted PDF and / or CAD files as discussed herein.

[0135] At step 904, the process includes entity parsing. The entity parsing in some embodiments parses various different types. Non-limiting examples of such entities include a line, a LW polyline, and a circle.

[0136] At step 906, the process includes extracting entity features. In some embodiments, the extracted entity features include data points associated with the features parsed at step 904. Non-limiting examples of the entity features include a length, an angle, a midpoint, a radius, and / or an area.

[0137] At step 908, the process includes normalizing coordinates. For example, the process may normalize coordinates that are extracted as entity features at step 906. For example, the coordinate normalizing may be translated and / or scaled for normalization. The normalized coordinates may be stored and / or otherwise made available for further processing.Patent Application Attorney Docket # 073233-00006

[0138] At step 910, the process includes creating graph data. In some embodiments, the graph data is created from a file, such as XLSX data representing the features (or normalized features) from steps 906 and / or 908. In some embodiments, the graph data includes graph nodes, edges, and / or other features.

[0139] At step 912, the process includes building edges of the graph. In some embodiments, an edge weighting algorithm, model, and / or the like is performed to build the edges. For example, in some embodiments, a K nearest neighbor implementation is utilized to build the edges.

[0140] At step 914, the graph data prepared from the sub-process is batched and prefetched. In some embodiments, a group per graph is generated and utilized for subsequent processing. A model structure sub-process may then be initiated.

[0141] The model structure sub-process begins at step 916. At step 916, the process includes receiving X inputs, where X is a set of features, and an edge index. The inputs may be processed by one or more subsequent algorithms.

[0142] At step 918, the process includes processing the inputs with a particular model and / or algorithm. For example, in some embodiments, the inputs are processed by a GraphSage algorithm. The inputs are processed via the model and / or algorithm multiple times.

[0143] At step 920, the process includes performing a softmax per entity. For example, in some embodiments, the process includes a softmax per entity of 36 layer classes.

[0144] At step 922, an output of the model structure process is provided. In some embodiments, for example, the outputs include per-entity predictions for further processing.

[0145] The process continues to a training sub-process, beginning at step 924. At step 924, the process includes performing a stratified K-fold algorithm. For example, in some embodiments, the K- folds are performed by drawing, utilizing 5 folds as an example.Patent Application Attorney Docket # 073233-00006

[0146] At step 926, the process includes tracking loss and updating weights. For example, in some embodiments, the updating includes tracking and updating based on a cross-entropy. Additionally, or alternatively, the class weights of the model are updated based on the stratified K-folds performed.

[0147] At step 928, the process includes updating the model based on an optimizer. For example, in some embodiments, an Adaptive Moment Estimation (ADAM) optimizer may be performed with a particular learning rate.

[0148] At step 930, the process includes scheduling a learning rate decay. For example, in some embodiments, a scheduler utilizes StepLR that decays the learning rate over epochs of training. In some embodiments, for example as depicted, every 15 epochs, a learning rate is decayed by 0.5.

[0149] At step 932, the process includes batching data. For example, in some embodiments, DataLoader is utilized to continue data fetching for training.

[0150] At step 934, a checkpoint is reached. For example, accuracy of the model at the checkpoint may be reached, evaluated, and / or the like. In some embodiments, the model with the best evaluated accuracy is maintained and / or further updated. An evaluation sub-process is then initiated.

[0151] The evaluation sub-process begins at step 936. At step 936, the process includes loading the best model. In some embodiments, the best model embodies the version of the model having the best (e.g., highest) accuracy.

[0152] At step 938, the process includes determining metrics accuracy per fold. The metrics per fold may be averaged for further comparison and / or processing.

[0153] At step 940, quantitative checks of visual samples are performed. One or more errors may be identified and / or the model or a user may correct such errors. In some embodiments, the correction includes inferring missing elements (e.g., missing elevation data). In some embodiments, the correction includes correcting surface irregularities.Patent Application Attorney Docket # 073233-00006

[0154] At step 942, one or more predictions are exported. For example, in some embodiments, the predictions are exported to a DXF. The predictions may be exported in a color-coded manner. In some embodiments, the predictions are exported in a CSV format.

[0155] Outputs from these models are aggregated using a specially-configured algorithm and / or Al aggregation logic that generates a unified set of standard layers and object associations.

[0156] In some aspects, the Al Platform maintains a standard layer database comprising, for example, thirty-two predefined layer classes (e.g., “3D-Break Line,” “3D-Contours,” “2D-Storm,” “2D- Sidewalk”). Objects are assigned to the appropriate layers and non-essential data (e.g., text, lines, or symbols not required for MCM generation) is removed. An Al Layer Scorecard can categorize layers into “Identified,” “Needs Review,” or “Deleted,” with confidence scores for each classification, enabling targeted human verification.

[0157] In some aspects, CAD engineers or other authorized users may review the “Needs Review” categories and apply updates. Surveyors can input GPS localization data to geo-reference the MCM accurately. The model can then be validated against approved plan sets (e.g., stamped and / or otherwise certified plans) to confirm consistency with design intent, regulatory requirements, and field conditions. Once the initial MCM with standard layers has been generated from the CAD data, the contours, elevations, and spot grades extracted from the authoritative PDF (as processed via the Elevation, Contour / Topography Extraction Process of FIG. 5 and the Spot Elevation Data process of FIG. 6) are mapped against the CAD-derived model to verify accuracy. Any discrepancies are resolved by updating the MCM to match the PDF, ensuring that the model reflects the approved, stamped design.

[0158] After validation and update, the linework within the model is elevated — transforming the 2D geometry into 3D linework — thereby enabling accurate volumetric and topographic representation. Following linework elevation, the GNN model is applied to generate breaklines, identifying the bottomPatent Application Attorney Docket # 073233-00006 and top of all slopes to avoid common Triangular Irregular Network (TIN) surface errors and / or irregularities, where TIN may traditionally be utilized to create a surface (see also FIG. 9). These breaklines are incorporated into the MCM to ensure geometric accuracy and machine-readability.

[0159] Once breaklines are established, in some embodiments the platform generates a complete 3D surface — a triangulated representation of ground topography — thereby finalizing the CAD portion of the MCM.

[0160] Once finalized, the validated, updated, and fully modeled Machine Control Model in some embodiments is automatically converted into the required manufacturer-specific format, such as Topcon formats TP3, GRD, or LN3 files, or Trimble TTM or DSZ / DSF files, and uploaded to a cloud-based storage system. From there, the model is downloadable directly to equipment for execution in the field, enabling real-time guidance, grading, excavation, or other machine control operations.

[0161] By combining heterogeneous input handling, multi-model Al recognition, standardized layer mapping, and automated validation, this workflow significantly reduces the time, cost, and labor associated with generating MCMs, improves accuracy, enhances interoperability across disparate systems, and promotes broader adoption of machine control technology in construction environments.

[0162] FIGS. 10A-10E depict a flowchart with operations of an example process for machine control model generation. The FIGS. 10A-10E depict operations of a process 1000. The process 1000 may implement one or more of the processes, models, and / or systems discussed herein. In this regard, the similarly named components, models, and operations may be performed in the manner described herein, including but without limitation with respect to performance of the discussed sub-models (e.g., BERT, CNN, and / or GNN models), PDF and / or DXF processing methodologies, and / or the like. For example, in some embodiments, the process 1000 is performed by an Al platform of the machine controlPatent Application Attorney Docket # 073233-00006 modeling system 102, such as embodied by the apparatus 200. In this regard, the Al platform may be specially configured to perform the various operations and sub-processes discussed herein.

[0163] As illustrated in FIG. 10 A, the process 1000 includes sub-process 1099A. The process 1000 includes a process initiated by a user 1002. At operation 1004, the process includes a user accessing a website associated with the platform. In some embodiments, the user device accesses the web platform via a browser on the user device. Via the website, for example utilizing one or more specially configured interfaces to create a project at step 1006.

[0164] The user 1002 may utilize the website to upload particular files. For example, at operation 1008 the user 1002 uploads a construction plans to the website. In some embodiments, the user 1002 uploads one or more PDF and / or CAD files embodying the construction plan. In some embodiments, a PDF file is processed with a sub-process, for example beginning at operation 1020. In some embodiments, a CAD drawing is processed with another sub-process, for example beginning at operation 1010.

[0165] In some embodiments, a Machine Control Model is generated from PDF-only inputs (e.g., without receiving any associated CAD (e.g., DWG) files. In some embodiments, the Al Platform is configured to generate a Machine Control Model directly from a PDF file without requiring any native CAD file (e.g., DWG) as an input. The system (e.g., including or embodying the Al platform) processes the PDF, which may be either vector-based or raster-based, using a multi-stage extraction pipeline described herein. In vector PDFs, the Al platform identifies geometric entities, linework, annotations, and symbols, and converts such data into structured CAD entities, which may be further processed as described herein. In raster PDFs, computer vision models, including convolutional neural networks (CNNs) and optical character recognition (OCR) models, detect and classify contours, spot grades, dimensions, and feature boundaries. The Al platform in some embodiments applies a specializedPatent Application Attorney Docket # 073233-00006 conversion process to transform the extracted geometric and attribute data into a DXF CAD drawing. From this DXF output, the Al platform executes the Al Layer Creation Model, standardizes the layers, elevates linework, generates breaklines, and produces the finalized 3D surface, thereby generating a complete, validated Machine Control Model as described herein with utilizing a native inputted CAD file.

[0166] In some other embodiments, the Al platform is configured to generate a Machine Control Model directly from a CAD file in DWG format (e.g., without needing a corresponding inputted PDF construction plan). In some embodiments, the system (e.g., including or embodying the Al platform) ingests the CAD file and applies an ODA converter or equivalent processing module to extract all available geometric entities, layers, and object data. The Al platform then processes the extracted data to unlock, unfreeze, and enable all layers; remove non-essential elements; and / or normalize the geometry for machine control compatibility. In some embodiments, an Al Layer Creation Model is applied to classify and standardize layers, while the Al platform’s object recognition modules associate objects with their appropriate standard layers. The system may then elevate the 2D linework to 3D geometry, generate breaklines, and build the final 3D surface to produce a completed Machine Control Model. By not requiring corresponding PDF file, such functionality enables rapid conversion of legacy CAD designs, design revisions, or third-party CAD files into fully functional, manufacturer-compatible MCMs without requiring access to a validated and / or stamped PDF plan.

[0167] At operation 1012, the process includes using an ODA converter to extract objects and layers data in the CAD file. At operation 1014, the process optionally includes using an ODA converter and visual special effects to create a visualization for the user. The visualization may be outputted via the web interface to the user.Patent Application Attorney Docket # 073233-00006

[0168] At operation 1016, the process includes unlocking, unfreezing, and / or enabling all layers.Additionally, or alternatively, at operation 1016, the process includes deleting unnecessary visual elements from the CAD file (e.g., hatches). Additionally, or alternatively still, in some embodiments, the process includes exploding objects (e.g., ungrouping) objects in the file. Such operations may be performed to save the data in a structured format for further processing.

[0169] At operation 1018, the process includes creating a DXF file. The DXF file in some embodiments is generated from the data saved in a structured format at operation 1016. The Figure continues in FIG. 10B.

[0170] As depicted in FIG. 10B, the process 1000 further includes a sub-process 1099B. At operation 1022, the process includes an Al layer model that controls and manages layer identification and processing using an Al model, for example an LLM. In some embodiments, the process further includes various sub-processes. At operation 1024A for example, includes preprocessing text data for a BERT model, for example a BERT model 1026A. The BERT model 1026A, after processing, is used in operation 1028 A to identify and classify one or more layers by text, and store such data in a structured format. The structured data may be processed at a subsequent step, for example operation 1034 discussed herein. In some embodiments, the BERT model processing discussed with respect to operations 1024A-1028A is performed or embodied by the process 700 described herein.

[0171] In parallel or in series, the process 1000 includes operation 1024B of preprocessing image data for a CNN model, for example the CNN model 1026B. The CNN model 1026B, after preprocessing, is used in operation 1028B to identify objects and images, and store such data in a structured format. The process includes operation 1030 of grouping object predictions into layer predictions, for example based on the stored data in the structured format from operation 1028B. The grouped objects may be processed in a subsequent step, for example operation 1034 discussed furtherPatent Application Attorney Docket # 073233-00006 herein. In some embodiments, the CNN model processing discussed with respect to operations 1024B-1028B and 1030 is performed or embodied by the process 800 described herein.

[0172] In parallel or in series, the process 1000 includes operation 1024C of preprocessing graph data for a GNN model, for example the GNN model 1026C. The GNN model 1026C, after processing, is used in operation 1028C to identify one or more objects, for example using nodes, edges, and / or graph data, and store such data in a structured format. The process includes operation 1032 of grouping object predictions into layer predictions, for example based on the stored data in the structured format from operation 1028C. The grouped objects may be processed in a subsequent step, for example operation 1034 discussed further herein. In some embodiments, the GNN model processing discussed with respect to the operations 1024C-1028 and 1032 is performed or embodied by the process 900 described herein.

[0173] In some embodiments, at operation 1034 the process 1000 includes using a specialized algorithm and / or Al platform (e.g., embodying Al aggregation logic) that analyzes the 3 model outputs to generate a layer output. At operation 1036, the process 1000 optionally includes generating an Al layer scorecard. The Al scorecard in some embodiments shows Al standard generated layers, categorized into the categories of (1) needs review, (2), identified, and (3), deleted. In some embodiments, the categories are presented with confidence levels corresponding to such categories. The Al layer scorecard in some embodiments is outputted to the user 1002, for example via the web interface. In some embodiments, at operation 1036 the process 1000 optionally includes receiving input where a user (e.g., a CAD engineer) updates the layers and objects indicated in the “needs review” category.

[0174] At operation 1040, the process 1000 includes training Al models with updated layers and objects data. The Al models may be of an Al platform, and trained to learn using Al observational learning. The Al models may be trained and / or updated, and stored for future use.Patent Application Attorney Docket # 073233-00006

[0175] In some embodiments the process continues to FIG. 10C from operation 1022, or from operation 1040. As depicted in FIG. 10C, the process 1000 includes maintaining a plan recognition library 1042. The plan recognition library may be maintained as a cloud repository, but in some embodiments may be embodied by a local repository and / or hybrid repository. As illustrated, in some embodiments the plan recognition library includes layer recognition library 1044A and object recognition library 1044B. The layer recognition library 1044A in some embodiments maintains data associated with Al models that detect, extract, identify, and / or otherwise process layers. For example, as illustrated the layer recognition library 1044A in some embodiments includes the 32 standard layer database 1050, including the data depicted therein. In some embodiments, the object recognition layer 1044B in some embodiments maintains data associated with Al models that detect, extract, identify, and / or otherwise process objects. For example, as illustrated, the object recognition library 1044A in some embodiments includes an object and line database 1046. The object and line database 1046 in some embodiments includes data associated with detected objects, for example a “ditch,” “road,” and “front of curb” as illustrated. The process 1000 includes creating and storing (for example, in the object recognition library 1044B and / or layer recognition library 1044A) associations of which objects belong in which layers, for example at operation 1048. In some embodiments, the process 1000 utilizes the data stored in the plan recognition library 1042 in one or more of the operations 1028 A, 1028B, 1028C, and / or 1040 as depicted and described.

[0176] Returning to FIG. 10B, in some embodiments, the process 1000 continues to FIG. 10D from the operation 1040 and / or operation 1022. As illustrated in FIG. 10D, a sub-process 1099D is performed. At operation 1052, the process 1000 includes performing elevation, contour, and / or topography extraction via a service manager programmed to control and manage PDF processing (e g., via the Al platform). In some embodiments, the elevation, contour, and / or topography extraction andPatent Application Attorney Docket # 073233-00006 processing is performed by the Al platform as discussed herein with respect to the process 500. At operation 1054, the process 1000 includes determining whether the file to be processed is a raster file or a vector file. In a circumstance where the file is a vector file, the process 1000 includes processing the vector file using a vector PDF pipeline at step 1056. The vector PDF pipeline in some embodiments includes one or more of steps 1058A and / or 1058B. For example, in some embodiments, the process 1000 includes text detection performed at step 1058 A. Additionally, or alternatively, in some embodiments, the process 1000 includes arrow head detection at operation 1058B. Such data may be processed for identifying particular data from the vector file, as discussed herein. At operation 1060, the process includes performing arrow-text mapping at operation 1060. In some embodiments, the arrowtext mapping is performed using one or more identified arrow shafts. The process may then continue to operation 1062, as discussed herein.

[0177] In a circumstance where, at operation 1054, the file to be processed is a raster file, flow continues to operation 1064. At operation 1064, the process 1000 includes processing the file using a raster PDF pipeline at step 1064. The raster PDF pipeline may include one or more subsequent steps, as discussed herein, for example steps 1066, 1068A, 1068B, 1070, 1072, and / or 1074. In some embodiments, the raster PDF pipeline includes one or more additional steps. For example, in some embodiments, the raster PDF pipeline uses the Al platform to infer missing data.

[0178] In some embodiments, the Al platform is configured to infer missing elevation data where survey or plan inputs contain gaps or inconsistencies. For example, when contour intervals in a grading plan are incomplete, the Al platform may analyze surrounding contour line geometry, slopes, and breaklines to predict intermediate elevations using interpolation methods (e.g., linear interpolation, spline fitting, kriging). Additionally, the Al platform may incorporate regression models trained on prior grading datasets to estimate missing elevation points consistent with typical civil engineering patterns.Patent Application Attorney Docket # 073233-00006By combining geometric interpolation with machine-learned predictions, the Al platform generates inferred elevation data that is within acceptable tolerance of surveyed values. In one example, when 20% of spot grades are absent from a site plan, the platform infers those values by applying slope-based regression, yielding elevation predictions that can be validated against subsequent survey data.

[0179] At step 1066, the process includes processing the raster fde using at least one Al raster image model at operation 1066. In some embodiments, the Al raster image model includes a YOLO model at operation 1068A and / or an OCR model at step 1068B. In some embodiments, the process 1000 includes using a pipeline that processes the file with a YOLO model 1068 A or object detection, and processing the file with an OCR model for text detection. At operation 1070, key points and text are identified. The key points and text may include the objects and / or text identified from the YOLO model 1068A and / or OCR model 1068B.

[0180] At operation 1072, the process 1000 includes arrowhead detection. In some embodiments, the arrowhead detection is performed based at least in part on the objects, text, and / or other key points identified at the step 1070. For example, at operation 1074, the process 1000 includes arrow-text mapping using dynamic cone search is performed. Additionally, or alternatively, in some embodiments, the arrow-text mapping maps both identified arrow heads and identified text to particular layers. The process continues to operation 1062 as discussed herein.

[0181] At operation 1062, the process 1000 includes generating information mapped from PDF output. In some embodiments, the information mapped from the PDF may represent the arrow-text mapping performed using the arrow shaft mappings at operation 1060 and / or dynamic cone search at operation 1074. The mapped PDF output in some embodiments is processed for further layer mapping. For example, at operation 1076 the process 1000 includes mapping layers with identified information (e.g., the information mapped PDF outputs at operation 1062) to DXF coordinates of the file. In thisPatent Application Attorney Docket # 073233-00006 regard, the layers and mappings may be performed to map the identified objects and layers to the coordinate system of the DXF file being processed.

[0182] Returning to operation 1052, the flow additionally in some embodiments proceeds to operation 1078. At operation 1078, the process includes identifying and extract spot grades and / or identifying and extracting topographical / contour grades and elevation data, as discussed herein. In some embodiments, the Al platform is configured to identify and / or extract spot elevation data (e g., grades) as discussed herein with respect to the process 600. The flow then continues to FIG. 10E, for example.

[0183] The FIG. 10E depicts a subprocess 1099E of the process 1000. At operation 1080, the process 1000 includes receiving or identifying an initial Machine Control Model. The initial Machine Control Model may include standard layers produced with unnecessary text, lines, objects, and / or layers removed. For example, in some embodiments, the initial Machine Control Model is generated based on the construction plan modified via the operations discussed above with respect to the process 1000.

[0184] At operation 1082, the process 1000 includes PDF and DXF mapping to match contours and spot grades. For example, in some embodiments, a PDF version of a construction plan (or a current Machine Control Model derived therefrom) and a DXF version of a construction plan (or a current Machine Control Model derived therefrom) is mapped accordingly. For example, the PDF may be a validated or otherwise approved PDF (e.g., a certified PDF plan), serving as a trusted version for comparison with the DXF version generated as discussed herein.

[0185] At operation 1084, the process 1000 includes validating and update elevation, spot grades, and / or topological and / or contour grades against an approved PDF. For example, the DXF mappings may be compared with validated, and / or approved PDF versions of the file for comparison. In a circumstance where the mappings are validated (or updated to be validated, such as by correctingPatent Application Attorney Docket # 073233-00006 identified errors), the flow continues to operation 1086. In some embodiments, the correction of identified errors includes correction of surface irregularities.

[0186] In some embodiments, the Al platform is configured to detect and correct surface irregularities created during surface model generation, such as artifacts in a triangulated irregular network (TIN). For example, the system may identify unrealistic grade breaks, spurious spikes, or depressions by analyzing slope continuity and curvature across a surface. The Al platform may then apply one or more corrective techniques, including and without limitation: (i) automatically inserting breaklines at slope boundaries, (ii) smoothing surface triangles using geometric algorithms, and / or (iii) training a machine learning model on historical surface data to distinguish between valid and invalid slope geometries. By applying these corrections, the system reduces common errors that otherwise lead to poor grading guidance in machine control systems, while preserving the fidelity of high-accuracy elevation points.

[0187] At operation 1086, the process 1000 includes elevating linework, for example transferring the linework from 2D to 3D. In some embodiments, the linework of the DXF version is elevated.

[0188] At operation 1088, the process 1000 includes using a GNN model to generate at least one break line. The GNN model may generate break lines embodying lines marking a bottom and / or top of all slopes. The break lines may be generated to avoid common TIN errors. For example, in some embodiments, the GNN is embodied by a break net model. In some embodiments the GNN model is embodied and / or functions as discussed herein.

[0189] In some embodiments, the Al platform is configured to automatically generate breaklines by combining geometric analysis of triangulated irregular network (TIN) surfaces with one or more models, including deep learning models. For example, in some embodiments, first, the system analyzes an initial TIN file to detect elevation and slope discontinuities across adjacent triangles. Edges or regions wherePatent Application Attorney Docket # 073233-00006 discontinuities exceed defined thresholds are flagged as candidate breakline locations. These candidate regions are then further processed by a three-dimensional convolutional neural network (3D CNN), which is trained on expert-annotated terrain datasets to distinguish true breaklines from noise and to extend partial breaklines for continuity.

[0190] The combined approach enables the system to automatically insert accurate breaklines marking the top and bottom of slopes, curbs, ridges, or other terrain discontinuities. By incorporating both rule-based TIN analysis and learned 3D CNN classification, the platform improves the quality of surface models, reduces common TIN errors, and decreases reliance on manual break line drafting. The validated breaklines are then incorporated into the Machine Control Model for final surface generation and deployment to equipment.

[0191] In some embodiments, hotspot areas of geometric roughness or inconsistency on a TIN surface can be extracted from a CAD file (e.g., in DXF format). An example method blends three surface cues, specifically edge blending, local inconsistency, and within-triangle height spread, into a single roughness score for each triangle. The method further denoises that score while preserving real edges, and then selects contiguous hotspot patches that are large enough to be actionable. In some embodiments, the resulting output includes (1) a full heatmap, (2) a clean hotspot overlay, (3) an interactive 3D viewer, and CAD-friendly exports (e.g., DXF points) that guide the system where to add breaklines to smooth the surface.

[0192] In some embodiments, the hotspot areas are extracted by loading and cleaning the triangles from a DXF surface (3DFACE / MESH / POLYFACE). Nearly-duplicate vertices may be “welded” to remove fake seams. The method may next score every triangle for roughness using three cues: (1) Edge bend of how sharply adjoining triangles meet, (2) an odd-one-out determination of how different a triangle’s tilt is from its neighbors, and (3) a relief determination of the Z-range across the triangle. ThePatent Application Attorney Docket # 073233-00006 method includes blend the cues into one score per triangle, then lightly smooth the score so small speckles fade but real edges remain. A hotspot may then be picked sing a fixed bend floor (e.g., > -10°), and / or based on the hotspot being in the top X% of the composite score (e.g., top 10%).In some embodiments, very skinny or tiny fragments (e.g., under a certain threshold) are ignored. Neighboring hot triangles in some embodiments are grouped into patches, and only patches with enough area are kept. The results may then be exported.

[0193] In some embodiments, TIN Pre-Analysis is performed. The TIN inputs may include a digital terrain model represented as a TIN comprising vertices, edges, and triangular facets. Local shape analysis may be performed for each shared edge between adjacent triangles, the system computes slope and elevation differences. Thresholding may additionally be performed, where edges with slope discontinuities or elevation differences exceeding a defined threshold are marked as candidate breakline edges. In some embodiments, clustering is performed so that candidate edges are grouped into contiguous regions representing potential linear features (e.g., ridges, valleys, road edges, curbs). A resulting output including a set of preliminary polylines, representing potential or “candidate” breaklines, may then be performed.

[0194] Additionally, or alternatively, in some embodiments, a 3D CNN analysis is performed. The 3D CNN analysis includes voxelization of local terrain around each candidate region, a 3D grid representation is constructed from TIN vertices and elevations. A trained CNN model (e.g., a trained 3D CNN) receives local patches of the voxelized terrain as input. The trained CNN model in some embodiments is trained based on annotated breaklines across various terrain types, construction sites and types, geographic locations and conditions, and the like. Other methodologies, such as unsupervised learning, may also be used. In some embodiments, the CNN model classifies candidate edges as true breaklines or noise and / or predicts extensions to ensure consistency and connectivity, and outputs suchPatent Application Attorney Docket # 073233-00006 data as output. The resulting breaklines may be smoothed, and snapped into vertices where applicable to integrate into the existing TIN or plan. The set of breaklines may then be outputted for inclusion to the Machine Control Model, for example in a particular format. Such breaklines may be utilized for any of a myriad of purposes, for example classifying particular edges of features in the site, detecting certain features in the site, and / or utilizing detected and / or classified features for machine control (e.g., to reduce field grading errors by ensuring that the Machine Control Model matches the intended design in a machine-interpretable format).

[0195] At operation 1090, the process 1000 includes generating a surface, for example of a Machine Control Model. The surface in some embodiments is a 3D representation of ground topography. The surface may be generated based on the linework and / or break lines, and / or other data generated as discussed at previous steps.

[0196] At operation 1092, the process 1000 includes completing a Machine Control Model. For example, in some embodiments, the Machine Control Model is completed by including the generated surface data, and / or other data, into the CAD file.

[0197] At operation 1093, the process 1000 includes providing access of the completed Machine Control Model to a surveyor. In some embodiments, the web platform is configured to grant access credentials to an identified surveyor to enable the Surveyor to input the GPS coordinates via the web platform. The Surveyor may receive a different interface than the user 1002, as discussed herein.

[0198] At operation 1094, the process 1000 includes incorporating localization details. In some embodiments, the process 1000 automatically incorporates localization details identified based at least in part on the inputted GPS coordinates. For example, the Al platform in some embodiments automatically identifies local regulations, zoning laws, building standards, and / or the like applicable to a Machine Control Model based on the inputted GPS coordinates, and incorporates such data accordingly.Patent Application Attorney Docket # 073233-00006

[0199] In some embodiments, the system embodying or including the Al platform maintains a dedicated zoning and planning regulation database (“Zoning Database”), for example utilized for zoning database maintenance and compliance validation. In some embodiments, the zoning database stores applicable codes, ordinances, setback requirements, easements, allowable use categories, height restrictions, grading limitations, and other regulatory parameters for multiple jurisdictions. The Zoning Database may be populated through a combination of automated ingestion from third-party data providers, web crawling of public agency websites, and manual curation by authorized users, and / or the like.

[0200] In some embodiments, the Zoning Database is version-controlled, allowing historical regulations to be preserved for reference in disputes or for projects initiated under prior rules. The database may be indexed by geographic coordinates, jurisdictional boundaries, and regulatory category, enabling rapid retrieval of applicable rules for a given project site.

[0201] Additionally, or alternatively, in some embodiments, during MCM generation or validation, the Al Platform automatically cross-references the model geometry, dimensions, and features against the Zoning Database for the relevant jurisdiction. Zoning validation may include checking model boundaries against required setbacks, verifying allowable grading or excavation volumes, ensuring compliance of structural features with engineering standards, and confirming site usage against permitted land use categories.

[0202] In some embodiments, the Al platform incorporates a zoning and code compliance component configured to cross-reference generated models with regulatory requirements. For example, based on GPS localization of a project site, the platform retrieves setback rules, height restrictions, or grading limitations from a zoning database. The system then evaluates the model geometry against these requirements, flagging violations such as a building footprint encroaching on a setback line orPatent Application Attorney Docket # 073233-00006 excavation volumes exceeding permitted limits. In some cases, the Al platform automatically suggests corrective design alternatives, such as adjusting grading slopes to comply with erosion control standards. Compliance results can be logged and output as a zoning validation report, suitable for submission to permitting authorities or insurers.

[0203] In some embodiments, in a circumstance where violations are detected, the Al platform generates a zoning compliance report identifying each issue, the specific regulation violated, and suggested corrective actions. In some embodiments, the Al platform can automatically adjust non- compliant features in the MCM to bring such features into compliance, optionally subject to user review. In some embodiments, zoning validation results are recorded in the audit logs described herein.

[0204] In other embodiments, the system embodying or including the Al platform is configured to generate Machine Control Models from a standard design (e.g., a construction plan) and GPS coordinates alone. In this regard, in some such embodiments, the Al Platform enables the creation of Machine Control Models based solely on known feature types or structures and corresponding GPS coordinates, without the need for a complete PDF or DWG plan. In some such embodiments, a user (e.g., the user 1002) selects a feature or structure from a library of standard designs stored within the platform, for example and without limitation swimming pools, parking lots, or retaining walls. Each standard design may include predefined geometric templates, dimensional parameters, and construction specifications. The user can specify the desired location by providing GPS coordinates and, optionally, orientation or alignment parameters. The Al platform in some such embodiments automatically scales, positions, and aligns the selected standard design within the geospatial coordinate system of the project site. From this positioned template, the Al platform generates the corresponding CAD geometry, elevates the linework, creates breaklines, and produces the 3D surface required for the Machine Control Model. By not requiring any PDF or CAD designs, such GPS and standard design-only implementationsPatent Application Attorney Docket # 073233-00006 enable rapid deployment of MCMs for repeatable structures, supports modular construction workflows, and reduces the need for full design plan sets for well-understood and pre-engineered features (which may not be available, for example).

[0205] In some embodiments, at operation 1095, the process 1000 includes generating a Machine Control plan (e.g., a Machine Control Model) in a desired manufacturer format. In some embodiments the desired manufacturer format is based on a particular equipment (e g., a construction vehicle) to which the Machine Control Model is to be transmitted.

[0206] At operation 1096, the process 1000 optionally includes reviewing and finalizing a Machine Control Model. For example, in some embodiments, the Machine Control Model is outputted via one or more interfaces provided to the user 1002, and / or users assigned to review and / or finalize the Machine Control Model. In some embodiments, such user(s) may adjust or otherwise update the Machine Control Model to finalize a version of the Machine Control Model for use.

[0207] In some embodiments, the Machine Control Model is configured for self-updating in response to newly received construction plan data, field data, or as-built measurements. For example, if a revised grading plan is uploaded, the Al platform identifies changes to contours and spot grades relative to the existing model, regenerates only the affected portions, and creates a new model version while preserving validated geometry elsewhere. Similarly, as-built survey points collected by field data collectors (e.g., GNSS rovers, drones, LiDAR) may be incorporated into the model, triggering an automatic retraining of surface prediction algorithms performed by the Al platform. This continuous learning process enables the platform to adapt and improve over time while maintaining audit logs of each model revision.

[0208] In some embodiments, the system embodying the Al platform is configured to estimate the carbon emissions associated with executing a given Machine Control Model. Emission calculations canPatent Application Attorney Docket # 073233-00006 be based on simulated machine usage derived from the MCM’s scope, material volumes, and estimated equipment cycles, and / or from actual telemetry data received from connected equipment, machinery, sensors, and / or the like in the field. In this regard, the Al platform may be configured to generate simulated carbon emissions data for a Machine Control Model, and / or may be dynamically updated based on detected changes, for example in a workflow, a material volume, site logistics data, and / or the like, or any combination thereof. The Al platform in some embodiments may integrate data streams such as engine run-time, fuel consumption rate, load factor, travel distance, and material movement metrics, and apply emissions conversion factors according to recognized standards (e.g., EP A, ISO).

[0209] In some aspects, the emissions model is dynamically updated to reflect changes in construction workflow, phasing, material volumes, or site logistics, ensuring that environmental impact assessments remain accurate throughout the project. The Al platform in some embodiments generates ESG (Environmental, Social, and Governance) compliance reports summarizing a carbon footprint, energy consumption, safety incidents, and regulatory compliance, with outputs formatted for submission to stakeholders, regulators, or sustainability certification bodies.

[0210] In some embodiments, the Al platform estimates carbon emissions and environmental impact associated with executing a given Machine Control Model. The system may simulate equipment operation based on grading volumes, cut / fill quantities, and expected machine cycles, or may integrate real-time telemetry data such as fuel consumption, engine runtime, and load factors from connected equipment. Emissions are calculated using standardized conversion factors (e.g., EPA or ISO protocols). The platform may generate ESG (Environmental, Social, Governance) compliance reports that summarize a project’s estimated carbon footprint, energy consumption, and safety metrics. In one example, when modeling an excavation requiring 10,000 cubic meters of earthmoving, the systemPatent Application Attorney Docket # 073233-00006 estimates carbon output for the associated dozer fleet and provides a comparative report showing potential savings from optimized grading sequences.

[0211] In some embodiments, the system embodying or including the Al Platform is configured to document and log each stage of model generation, validation, deployment, and updating. Such actions may be logged in an insurance-grade audit trail logging process. For example, in some embodiments such actions are logged in an immutable, auditable format suitable for use in insurance underwriting, dispute resolution, or risk management. Each record may include, without limitation, a model version identifier, creation and modification timestamps, user approvals, compliance validation results, and a cryptographic hash of the model data.

[0212] The audit log may be stored in a secure, access-controlled repository and exportable in formats compatible with commercial insurance platforms, risk assessment systems, or legal discovery processes. In some aspects, the audit trail includes cross-references between model changes and as-built validation results, enabling underwriters, claims adjusters, and project owners to verify adherence to design intent and identify potential causes of deviations.

[0213] At operation 1097, the process 1000 includes uploading the Machine Control Model to a cloud. The Machine Control Model is uploaded to a cloud accessible by a particular equipment, for example a particular construction machine or multiple construction machines.

[0214] At operation 1098, the process 1000 includes a user initiating downloading a Machine Control Model to equipment (e.g., a construction machine). In some embodiments, the Machine Control Model downloaded to the equipment is utilized by the construction machine to automatically control operation of the equipment (e.g., autonomous or semi-autonomous control of operation of a construction machine) in accordance with the Machine Control Model. Additionally, or alternatively, in some embodiments, the Machine Control Model is utilized to inform a user of a manner to control thePatent Application Attorney Docket # 073233-00006 equipment (e.g., inform control of operation of a construction machine). For example, in some embodiments, the Machine Control Model is configured for asphalt pavers to control paving thickness, material flow, and edge alignment based on a construction plan thereof. Alternatively, or additionally, in some embodiments, the Machine Control Model is configured for a line-painting machine that is autonomous or semi-autonomous and that performs precise line striping based on the Machine Control Model and in accordance with at least one site layout specification. Alternatively, or additionally, in some embodiments, the Machine Control Model is configured for a layout printer configured to mark at least one reference line or at least one control point directly onto a jobsite surface using the Machine Control Model based on one or more design specifications thereof. Alternatively, or additionally, in some embodiments, the Machine Control Model is configured for a heavy civil excavator, where the Machine Control Model guides the heavy civil excavator equipped with a 3D machine control system to execute a precise grading and trenching activity based on the Machine Control Model. Alternatively, or additionally, in some embodiments, the Machine Control Model is configured for an automated or semiautomated bulldozer, where the Machine Control Model guides a blade elevation, a tilt, and a slope during at least one grading operation based on 3D surface data. Alternatively, or additionally, in some embodiments, the Machine Control Model is configured for a light transport and logistics (LTL) machine, where the Machine Control model is used for material movement, routing, and delivery sequencing based at least in part on a site layout and construction phasing.

[0215] In some embodiments, the system embodying or including the Al Platform is configured to validate actual construction or installation against the corresponding Machine Control Model by ingesting “as-built” (or “as-implemented”) data collected from at least one field data collection devices (“Field Data Collectors”). Such data may be used for as-built validation, for example based on real-time or near-real-time collected data. Non-limiting examples of field data collectors may include, but are notPatent Application Attorney Docket # 073233-00006 limited to, smartphones, GNSS / GPS rovers, robotic total stations, terrestrial laser scanners, drone-based photogrammetry systems, LiDAR devices, imaging systems, loT-connected sensors, and automated site inspection robots. In some aspects, each Field Data Collector may be associated with a specific project, model version, or geographic boundary, enabling the system to automatically map collected measurements to the correct Machine Control Model.

[0216] In some embodiments, the Al Platform performs a geometric and semantic comparison of the as-built data points to the MCM. The comparison may include or result in identifying deviations, omissions, or unplanned modifications. In some aspects, the validation process may be performed in real time during construction activities, triggering alerts or notifications when deviations are detected, and / or exceed defined threshold tolerances. In other aspects, the process may be performed as a batch comparison at defined milestones. The system can record validation results in a structured format, associating each deviation with a location, timestamp, measurement source, and responsible party, thereby enabling continuous quality control and post-construction compliance verification.

[0217] In some embodiments, the system including or embodying the Al platform is configured to update an existing Machine Control Model when a revised construction plan is uploaded, for example by the user 1002. In some embodiments, the Al platform updates the existing Machine Control Model regardless of whether the updated construction plan is uploaded in PDF or CAD (e.g., DWG) format. Upon receipt of the updated file, for example, the system automatically detects the file type and initiates the appropriate processing pipeline (e.g., a raster PDF pipeline, vector PDF pipeline, or CAD ingestion pipeline, as discussed above).

[0218] In some embodiments, the updated plan is aligned to the coordinate system of the current MCM, and an Al-based comparison engine is configured to identify all geometric, annotation, or layer differences between the updated plan and the existing model. Unchanged portions of the MachinePatent Application Attorney Docket # 073233-00006Control Model are preserved, while modified portions are reprocessed. Such implementations reduce or minimize computation time and prevent unnecessary changes in updating the Machine Control Model.

[0219] For updated PDFs, in some embodiments the Al platform extracts contours, spot grades, and elevation data and validates such data against the MCM, applying the “Validate and Update” process described herein. For updated CAD files (e.g., updated DWGs), the Al platform in some embodiments extracts revised layers and object data, re-runs the Al Layer Creation Model, and regenerates affected breaklines, 3D linework, or surfaces

[0220] In some embodiments, each update produces a new version of the MCM with a unique identifier, while the previous version is archived in the audit logging discussed herein, along with compliance checks against the Zoning Database discussed herein and, if applicable, as-built validation results discussed herein. The Al platform may also generate an automated change report and notify applicable users (e g., stakeholders or other entities associated with the construction model and / or Machine Control Model) when an updated MCM is ready for deployment.

[0221] Further details regarding specific aspects of embodiments of the disclosure are discussed below. It should be appreciated that embodiments of the disclosure, such as those discussed above with respect to FIGS. 1-9 and 10A-10E, may incorporate any such details in one or more embodiments.GENERATION OF NEURAL NETWORK FOR GENERATION OF SITE PLANS

[0222] According to some aspects, a site plan component and / or engine is provided. A site plan component in some instances is provided and incorporates or calls an artificial intelligence (Al) software program, model, and / or module (for example a generative Al program or model) to automatically create or generate construction and / or site plans. In some instances, construction and / or site plans can be generated by one or more inputs, for instance from other sources (e.g., databases, other programs / modules / applications / retrieval-augmented generation (RAG)). In one aspect, constructionPatent Application Attorney Docket # 073233-00006 and / or site plans can be generated (or automatically generated) from at least one of an existing plan and a new plan or site survey. In some example instances, existing plans are stored in one or more databases connected to the system and provided to the site plan component. In some example instances, new site surveys are stored at one or more databases connected to the system and / or provided to the site plan component, or for instance, a new site survey can be provided by another application or held in a memory component of a user device implemented to carry out a new site survey.CREATION OF NEURAL NETWORK FOR GENERATION OF ARCHITECTURAL PLANS

[0223] According to some aspects, an architectural plan component and / or engine is provided. An architectural plan component in some instances incorporates or calls an artificial intelligence (Al) software program, model, and / or module (for example, a generative Al program or model) to automatically generate and / or create one or more architectural designs and / or plans. In some instances, architectural designs and / or plans can be generated one or more inputs, for instance from other sources (e.g., databases, other programs / modules / applications / retrieval-augmented generation (RAG)). In one aspect, architectural designs and / or plans are generated (or automatically generated) from at least one of an existing design or plan and a new plan and / or site survey and / or design. In some example instances, existing designs or plans are stored at one or more databases connected to the system and provided to the site plan component. In some example instances, new site surveys, plans, or designs are stored at one or more databases connected to the system and provided to the architectural plan component, or in some instances, a new site survey are provided to the architectural plan component by another application or held in a memory component of a user device implemented to carry out a new site survey and subsequently provided to the architectural plan component.Patent Application Attorney Docket # 073233-00006CREATION OF NEURAL NETWORK FOR GENERATION OF BIM (BUILDINGINFORMATION MODELING) LAYOUT

[0224] According to some aspects, architectural and / or conceptual design component and / or element is provided. An architectural and / or conceptual design component in some instances is provided and incorporate or call an artificial intelligence (Al) software program, model, and / or module (for example a generative Al program or model) to automatically create or generate layout. In some instances, a layout can be generated by one or more inputs, for instance from other sources (e.g., databases, other programs / modules / applications / retrieval-augmented generation (RAG)). In one aspect, layout is generated (or automatically generated) from at least one of an existing and a new architectural and / or conceptual design. In some example instances, existing architectural and / or conceptual design is stored in one or more databases connected to the system and provided to the architectural and / or conceptual design. In some example instances, a new architectural and / or conceptual design is stored at one or more databases connected to the system or and provide to the architectural and / or conceptual design component, or for instance, a new architectural and / or conceptual design is provided by another application or held in a memory component of a user device implemented to carry out a new architectural and / or conceptual design.

[0225] Various aspects of the technology described herein can be implemented with any construction / building technology that can utilize automation models, for instance machine control equipment. In some instances, these include automated machines and equipment that read CAD plans or models, and / or GNSS, and / or implement Al, for example for enhanced precision and efficiency. In some examples, the machines and equipment include construction machines, such as autonomous and semi- autonomous construction machines, which may include earthmoving and excavation equipment, such as GPS-enabled excavators, automated bulldozers, autonomous graders, and / or robotic loaders, amongstPatent Application Attorney Docket # 073233-00006 others. In some examples, the construction machines include machine control systems, such as 2D and3D machine control systems, automated paving systems, and / or total station-controlled machines, amongst others. In some examples, the construction machines include robotic surveying and layout tools, such as robotic total stations, autonomous drones for site surveying, and GNSS / GPS-based layout robots, amongst others. In some examples, the construction machines include concrete and rebar fabrication tools, such as automated rebar bending machines, robotic concrete printers (e.g. industrial 3D printers), and / or automated shotcrete robots, amongst others. In some examples, the construction machines include prefabrication and modular construction tools and equipment, such as CNC machines, automated timber framing machines, and / or modular home assembly robots, amongst others. In some examples, the construction machines include road and bridge constructions tools and equipment, such as automated asphalt pavers, bridge deck welding robots, and / or tunnel boring machines (TBMs), among others. In some examples, the construction machines include robotics and Al-assisted machines, such as bricklaying robots, Al-driven robotic arms, and / or autonomous site inspection robots, among others. In some examples, the construction machines include digital twin and Al-based systems, such as AI- powered site monitoring cameras, digital twin-enabled construction machinery, and / or autonomous warehouse and logistics robots, among others. As will be appreciated, aspects of the present technology can be implemented in automated equipment to enhance precision, efficiency, and automation of the equipment.EXAMPLE PROCESSES OF THE DISCLOSURE

[0226] Having described example systems and related implementation details, example processes in accordance with the present disclosure will now be discussed. In some embodiments, a process embodies a computer-implemented method that is implementable by one or more computers, devices, systems, and / or the like discussed herein. For example, in some embodiments, the process isPatent Application Attorney Docket # 073233-00006 implemented via a machine control modeling system 102 as depicted and described herein. The machine control modeling system 102 may be embodied by the apparatus 200, which may be specially configured to perform the processes described herein. In some embodiments, a process is embodied by computer-coded instructions stored on at least one non-transitory computer-readable storage medium, for example that may be executed by one or more processor to trigger and / or perform the process.

[0227] In some embodiments, a process may include one or more optional operation. As illustrated, an optional operation may be depicted in dashed (or “broken”) lines, indicating that the operation is optional. In some embodiments, all optional operations are performed as part of the process. In some embodiments, some optional operations are performed as part of the process. In some embodiments, none of the optional operations are performed as part of the process.

[0228] FIG. 11 depicts operations of a process for machine control model generation and use. In some embodiments, a Machine Control Model is generated using the process for deployment to one or more construction machine. The process may be performed to generate a Machine Control Model in accordance with specific data and / or information processed to generate the corresponding, desired Machine Control Model.

[0229] At operation 1102, the process includes receiving a construction plan. In some embodiments, the construction plan is received in response to upload by a user. In some embodiments, the construction plan is received upon retrieval from a database (e.g., based on prior upload to the system). In some embodiments, the received construction plan includes at least one PDF file. Additionally, or alternatively, in some embodiments, the received construction plan includes at least one CAD file. In some embodiments, the process includes validating the at least one PDF file of the construction plan with at least one CAD file of the construction plan, for example to confirm matching data within both files prior to further processing. In some embodiments an Al model is specially configured to comparePatent Application Attorney Docket # 073233-00006 the at least one PDF file with the at least one CAD file for validation, and output a determination and / or value that may be compared to an acceptable threshold. The construction plan may continue to be processed if validation is successful, and to output an alert or error indicating a failed validation otherwise.

[0230] In some embodiments, the Al platform validates Machine Control Models (e.g., including CAD-based plans) against at least one certified PDF (e.g., authoritative, certified PDF plan sets). The system rasterizes the certified PDF, extracts annotated text, symbols, and arrow-based dimensions using computer vision and OCR, and overlays these extracted features against the CAD-derived geometry. An Al comparison engine then identifies mismatches between the PDF annotations and the CAD geometry, for example detecting omitted spot grades, alignment errors, or dimension conflicts. Detected discrepancies are flagged for user review or automatically corrected, for example by the Al platform, ensuring that the machine control model reflects the official stamped plans. In some embodiments, this validation workflow provides an audit trail that reduces liability and ensures compliance with regulatory and contractual design intent.

[0231] At operation 1104, the process includes maintaining an accuracy of an Al platform, the Al platform including at least one algorithm and at least one model, by performing periodic updated training of the at least one model of the Al platform. In some embodiments, the at least one algorithm and at least one model includes a neural network or other specially trained network that processes input data (e.g., a construction plan) and generates the corresponding Machine Control Model.

[0232] At operation 1106, the process includes processing the construction plan using the Al platform. In some embodiments, the Al platform is configured to use at least one machine learning model and / or algorithm specially configured to process data inputted to the Al platform. For example, in some embodiments, the Al platform utilizes the at least one algorithm and / or at least one model toPatent Application Attorney Docket # 073233-00006 process prior engineering data, prior construction data, and / or other prior data, or any combination thereof, from which the at least one algorithm and / or at least one model is trained. The Al platform may be trained, in some embodiments, to generate a new construction plan, which may be modified from an inputted and / or otherwise processed construction plan.

[0233] At operation 1108, the process includes generating a machine control model by inputting at least one model input to the Al platform. The machine control model is generated based on the processed construction plan, the at least one machine learning model of the Al platform, and the at least one model input.

[0234] At optional operation 1110, the process includes transmitting the machine control model to a construction machine. In some embodiments, the machine control model is transmitted to trigger the construction machine to operate (or in other words, to be controlled) in accordance with the machine control model. In some embodiments, the Machine Control Model is validated before being transmitted to the construction machine. For example, in some embodiments the Machine Control Model is compared with a trusted or validated construction plan (e.g., not configured as a Machine Control Model), such as a construction plan inputted to the system by a user, to determine if the details of the Machine Control Model and trusted / vali dated construction plan represent the same construction. In some embodiments, the Machine Control Model is transmitted to the construction machine for controlling the construction model, such as by displaying and / or instructing information for use by an operator of the construction machine (e.g., to operate the construction machine in accordance with the Machine Control Model) or by autonomous or semi-autonomous control via the construction machine.

[0235] FIG. 12 depicts operations of a process for data processing and extraction. For example, the data processing and extraction may be performed as a subprocess of a process for machine control model generation and use. In some embodiments, the process 1200 is a sub-process of the process 1100.Patent Application Attorney Docket # 073233-00006

[0236] In some embodiments, the process 1200 branches off into one or more individual processes, which may be performed together and / or alternatively in parallel or in series. For example, in some embodiments, operations 1202 and 1204 are performed. Additionally, or alternatively, in some embodiments, operation 1206 is performed. Additionally, or alternatively, in some embodiments, operation 1208 is performed. Additionally, or alternatively, in some embodiments, operations 1210 and 1212 are performed. It should be appreciated that, in some embodiments, any combination of the different branches of the process may be performed for that particular embodiment.

[0237] At operation 1202, the process 1200 includes extracting textual data and annotations from the raster-based PDF construction plan by a raster data extraction pipeline that uses artificial intelligence. In some embodiments, the raster-based PDF construction plan is a particular file, data, and / or the like received from a user as described herein, for example with respect to FIG. 11. In some embodiments, the raster-based PDF construction plan comprises one or more cells that define areas and / or portions of a site or project. The raster-based PDF construction plan may include pixel -based geospatial information that represents any of a desired set of properties and / or characteristics, for example properties and / or characteristics regarding a particular site, project, and / or the like. In some embodiments, the raster data extraction pipeline is embodied by or includes a sub-step wherein the raster-based PDF construction plan is processed utilizing the Al. In some embodiments, the Al is specially configured to detect and / or extract particular data types, and / or data values corresponding thereto, from the raster-based PDF construction plan. In some embodiments, the Al includes or is embodied by a specially-trained neural network that identifies and / or extracts the particular relevant data.

[0238] At operation 1204, the process 1200 includes converting the textual data and annotations extracted from the raster-based PDF construction plan into geospatial and structured data utilized in generating the Machine Control Model. For example, in some embodiments, the data is associated withPatent Application Attorney Docket # 073233-00006 different areas of a site and / or project. The data may be converted by combining data associated with a same area and / or portion of a site and / or project.

[0239] Additionally, or alternatively to the operations 1202-1204, in some embodiments the process 1200 includes operation 1206. At operation 1206, the process 1200 includes extracting vector data from the raster-based PDF construction plan and automatically converting the vector data into DXF format using at least one scripted workflow. In some embodiments the at least one scripted workflow each extract independent portions of the vector data from the raster-based PDF construction plan.Additionally, or alternatively, in some embodiments, the raster-based PDF data in the raster-based PDF is identified and separated into vector-based PDF data, and is done using a plurality of learning workflows that are optimized to interpret each data type of a plurality of data types prior to generating the Machine Control Model.

[0240] Additionally, or alternatively to the operations 1202-1204 and / or 1206, in some embodiments the process 1200 includes operation 1208. At operation 1208, the process 1200 includes extracting annotation text by retrieving MTEXT entities from the DXF format of the vector data. In some embodiments, the MTEXT includes multiline text formats included within the raster-based PDF construction plan.

[0241] Additionally, or alternatively to the operations 1202-1204, 1206, and / or 1208, in some embodiments the process 1200 includes operation 1210. At operation 1210, the process 1200 includes determining at least one annotation layer having a highest density of detected arrows. For example, in some embodiments, each layer of the raster-based PDF construction plan is processed, and a number and / or density of detected arrows is determined. The number of annotations for each level may be utilized (e.g., via comparison) to determine the layer with the highest density.Patent Application Attorney Docket # 073233-00006

[0242] At operation 1212, the process 1200 includes selecting, based on determining the at least one annotation layer having the highest density of detected arrows, the at least one annotation layer as at least one primary annotation layer. The selected annotation layer may be subsequently processed, for example as part of processing a raster-based PDF construction plan.

[0243] FIG. 13 depicts operations of a process for generation of a construction plan used in generating a Machine Control Model. For example, the construction plan generation may be performed as a subprocess of a process for machine control model generation and use. In some embodiments, the process 1300 is a sub-process of the process 1100.

[0244] At operation 1302, the process 1300 includes integrating, using a site plan recognition component, basic site data from one or more sources. For example, in some embodiments, particular data is identified from one or more real-time data sources or previously-stored data is retrieved. In some embodiments, data types of a particular type are determined based on the site, construction plan to be generated, and / or the like, such that only particular relevant data (e.g., determined marked or otherwise indicated relevant to a site, project, and / or the like) is integrated into stored data associated with the construction model and / or Machine Control Model to be generated.

[0245] At operation 1304, the process 1300 includes organizing, using the site plan recognition component, the basic data as a set of collected data. In this regard, the integrated and organized basic data may be marked or otherwise determined, or indicated, for further processing (e.g., for generation of a construction plan as described herein).

[0246] At operation 1306, the process 1300 includes generating, using a model build component, a construction plan based on the set of collected data. In some embodiments, the construction plan is generated based on the collected data (e.g., to satisfy certain restrictions, with selections and / or determinations based on the set of collected data, and / or the like).Patent Application Attorney Docket # 073233-00006

[0247] At operation 1308, the process 1300 includes automatically generating a machine control model including at least one new construction plan based at least in part on one or more inputs. The Machine Control Model is generated by an Al platform using a construction plan algorithm trained on a library of prior construction plans and associated project data.

[0248] At operation 1310, the process 1300 includes maintaining an Al platform accuracy of the Al platform by performing periodic updated training of the at least one construction plan algorithm of the Al platform. In some embodiments, the Al platform accuracy is updated in training as new data is received and / or collected, for example new basic site data received and / or collected from one or more other device. Additionally, or alternatively, in some embodiments, the Al platform is updated based at least in part on feedback as discussed herein.

[0249] At optional operation 1312, the process 1300 includes transmitting the Machine Control Model to a construction machine. In some embodiments, the Machine Control Model is transmitted to deploy the Machine Control Model for controlling one or more construction machine. In some embodiments, the Machine Control Model is transmitted to a construction machine as discussed above with respect to optional operation 1110.

[0250] FIG. 14 depicts operations of a process for updating an Al platform based on user interaction with construction plans. For example, for updating of the Al platform based on user interaction with construction plans may be performed as a subprocess of a process for machine control model generation and use. In some embodiments, the process 1400 is a sub-process of the process 1300.

[0251] At operation 1402, the process 1400 includes generating a plurality of construction plans. Each construction plan includes at least one distinct option. In this regard, each construction plan of the plurality of construction plans may by the at least one distinct option to embody various distinctPatent Application Attorney Docket # 073233-00006 construction plans. Each plan option of the plurality of construction plans is based on an individual user, a company, and local demographic preferences.

[0252] At operation 1404, the process 1400 includes outputting the plurality of construction plans to a user for selection. In some embodiments, the plurality of construction plans is displayed to a user via one or more interfaces. The one or more interface may be configured to receive a user input indicating the selection of the user.

[0253] At operation 1406, the process 1400 includes receiving a selection by the user based on an interaction with the plurality of construction plans outputted to the user. In this regard, the selection may embody user input that indicates a plan option (e g., one of the plurality of construction plans) that the user selected, for example for use or further processing. The Al platform is configured to learn based on the selection by the user. For example, the selection by the user may be provided as input to the Al platform for subsequent training.

[0254] FIG. 15 depicts operations of a process for direct machine control model generation and use. The process may be performed to generate a Machine Control Model in accordance with specific data and / or information processed to generate the corresponding, desired Machine Control Model.

[0255] At operation 1502, the process 1500 includes receiving geospatial data, positioning data, and / or topographical data. In some embodiments, the geospatial data, the positioning data, and / or the topographical data is received from one or more other devices. Additionally, or alternatively, in some embodiments, the geospatial data, the positioning data, and / or the typographical data is received from real-time sensors, loT devices, and / or the like in a field. Additionally, or alternatively, in some embodiments, the geospatial data, the positioning data, and / or the topographical data is received from a database or other storage including such prior-collected data.Patent Application Attorney Docket # 073233-00006

[0256] At operation 1504, the process 1500 includes processing the received geospatial data, positioning data, and / or topographic data through an Al platform including one or more algorithms and machine learning models trained on at least one prior engineering pattern, at least one prior grading pattern, and at least one prior construction pattern.

[0257] At operation 1506, the process 1500 includes generating a machine control model directly from the processed geospatial inputs. For example, in some embodiments, the Al platform is specially configured to generate a Machine Control Model directly by processing the geospatial data, the positioning data, and / or the topographical data. The one or more algorithms and / or machine learning models of the Al platform in some embodiments are trained to generate a Machine Control Model based on the inputted geospatial data, positioning data, and / or topographical data.

[0258] At optional operation 1508, the process 1500 includes training the Al platform with additional data and feedback. In some embodiments, the feedback includes user selections as described herein. Additionally, or alternatively, in some embodiments the additional data includes subsequently collected sensor, loT, and / or other data associated with a particular Machine Control Model. For example, in some embodiments, updated and / or new geospatial data, positioning data, and / or topographical data is received, and such updated data is utilized to update training of the Al platform.

[0259] FIG. 16 depicts operations of a process for generation of a plan. For example, generating the plan may be performed as a subprocess of a process for machine control model generation and use. In some embodiments, the process 1600 is a sub-process of the process 1100, 1300, and / or 1500.

[0260] At operation 1602, the process 1600 includes integrating, by a plan recognition component, data from one or more relevant sources. In some embodiments, the relevant data is determined based on a type of plan to be generated. For example, a construction plan to be generated corresponds to first types of data from one or more particular relevant data sources, an architectural plan to be generatedPatent Application Attorney Docket # 073233-00006 corresponds to second types of data from one or more particular relevant data sources, and a site plan to be generated corresponds to third types of data from one or more particular relevant data sources. The relevant data sources may differ depending on the type of plan to be generated.

[0261] At operation 1604, the process 1600 includes storing, by the plan recognition component, the data as a set of collected data. In some embodiments, the plan recognition component stores the set of collected in a database and / or the like. The set of collected data may be associated with a particular site, construction plan, and / or other identifier (or set of identifiers) that links the portions of data together.

[0262] At operation 1606, the process 1600 includes generating, by a model build component, a plan based on received survey data, design data, and other site data of the collected data. The plan is generated by a plan algorithm trained at least in part on the set of collected data. In some embodiments, the plan that is generated embodies a construction plan, an architectural plan, and / or a site plan. The particular type of plan may be generated by a particular corresponding plan algorithm, for example the construction plan is generated by a construction plan algorithm, a site plan is generated by a site plan algorithm, and an architectural plan is generated by an architectural plan algorithm.

[0263] At optional operation 1608, the process 1600 includes indicating, by a monitoring component, a deviation in the plan. In some embodiments, data indicating a deviation from the plan is detected by one or more sensors, devices, and / or the like associated with a particular site, construction plan, and / or the like.IMPLEMENTATIONS

[0264] Certain implementations of systems and methods consistent with the present disclosure are provided as follows:

[0265] Example Clause 1. A method including receiving a construction plan; maintaining an accuracy of an Al Platform, the Al platform comprising at least one algorithm and at least one model, by performingPatent Application Attorney Docket # 073233-00006 periodic updated training of the at least one model of the Al platform, wherein the at least one model of the Al platform is trained based at least in part on prior engineering data, prior construction data, or any combination of the prior engineering data and the prior construction data; processing the construction plan using the Al platform; and generating a Machine Control Model by inputting at least one model input to the Al platform, wherein the Machine Control Model is generated based on the processed construction plan, the at least one machine learning model of the Al platform, and the at least one model input.

[0266] Example Clause 2. The method of Clause 1, wherein processing the construction plan by the Al platform comprises: indicating, by the Al platform, a subset of data in the construction plan as determined to be unnecessary data for visual presentation or machine performance; and removing, by the Al platform, the subset of data from the construction plan.

[0267] Example Clause 3. The method of Clause 1, wherein generating the Machine Control Model comprises: correcting, by the Al platform, at least one surface irregularity generated using at least one triangular irregular network surface algorithm.

[0268] Example Clause 4. The method of Clause 3, wherein the correcting the at least one surface irregularity comprises: using at least one computer-generated break line to correct the at least one surface irregularity; or using at least one geometric algorithm to correct the at least one surface irregularity; or using the at least one machine learning model to correct the at least one surface irregularity.

[0269] Example Clause 5. The method of Clause 1, wherein correcting the at least one surface irregularity comprises: reducing a file size of the Machine Control Model by removing at least one oversaturated point after creating a high-fidelity elevation model.

[0270] Example Clause 6. The method of Clause 1, wherein the Al platform is configured for: determining a representation of at least one zoning law or at least one technical standard applicable to the construction plan or the Machine Control Model; and validating the construction plan or the Machine Control ModelPatent Application Attorney Docket # 073233-00006 satisfies the at least one zoning law or the at least one technical standard.

[0271] Example Clause 7. The method of Clause 1, wherein the Al platform is configured for extracting geometric and geospatial data from the construction plan; and transforming the geometric and geospatial data from the construction plan into structured data further processed in generating the Machine Control Model.

[0272] Example Clause 8. The method of claim 1, wherein the Machine Control Model is configured to self-update by regenerating a new version of the Machine Control Model after receiving an updated version of data used in during generating of the Machine Control Model.

[0273] Example Clause 9. The method of Clause 8, wherein the Al platform is configured for: receiving, automatically from at least one field data collector associated with at least one site, at least one model data point collected by the at least one field data collector from the site.

[0274] Example Clause 10. The method of Clause 9, wherein the at least one field data collector is associated with a specific machine control model and collects as-implemented data points associated with the specific Machine Control Model, in real time or near real time, and wherein the Al platform is configured for using the as-implemented data points to validate an implementation of the construction plan.

[0275] Example Clause 11. The method of Clause 10, the method further comprising: detecting at least one change in field measurements represented by the as-implemented data points; and in response to detecting the at least one change in field measurements, trigger an alert to update the Machine Control Model.

[0276] Example Clause 12. The method of Clause 1, further comprising: updating the Machine Control Model to generate a new version of the Machine Control Model based at least in part on additional data.

[0277] Example Clause 13. The method of Clause 1, wherein the Al Platform is configured for:Patent Application Attorney Docket # 073233-00006 automatically validating an implementation of the construction plan by cross-checking as-implemented data points associated with the implementation as-implemented and that are received by the Al platform with the Machine Control Model.

[0278] Example Clause 14. The method of Clause 1, further comprising: receiving equipment telemetry data associated with at least one equipment; and generating a simulated carbon emissions data based on simulated machine usage of the at least one equipment and the equipment telemetry data.

[0279] Example Clause 15. The method of Clause 14, wherein the simulated carbon emissions data is dynamically updated based on a detected change in a workflow, a material volume, or site logistics data.

[0280] Example Clause 16. The method of Clause 1, further comprising: generating, based on the Machine Control Model, at least one ESG compliance report summarizing a project impact, a project safety, a project carbon footprint, or a project audit trail.

[0281] Example Clause 17. The method of Clause 1, further comprising: generating an auditing log that indicates at least one detail associated with each of the generation of the Machine Control Model, a validation of the Machine Control Model, and an update of the Machine Control Model, wherein the auditing log is automatically generated in an auditable format.

[0282] Example Clause 18. The method of Clause 17, wherein the auditing log comprises at least one timestamp, at least one user approval, and at least one compliance validation mapped to a specific version of the Machine Control Model.

[0283] Example Clause 19. A system comprising at least one non-transitory memory and at least one processor, the at least one non-transitory memory storing computer-coded instructions that, when executed by the at least one processor, cause the system to: receive a construction plan; maintain an accuracy of an Al Platform, the Al platform comprising at least one algorithm and at least one model, by performing periodic updated training of the at least one model of the Al platform, wherein the at least one model ofPatent Application Attorney Docket # 073233-00006 the Al platform is trained based at least in part on prior engineering data, prior construction data, or any combination of the prior engineering data and the prior construction data; process the construction plan using the Al platform; and generate a Machine Control Model by inputting at least one model input to the Al platform, wherein the Machine Control Model is generated based on the processed construction plan, the at least one machine learning model of the Al platform, and the at least one model input.

[0284] Example Clause 20. A non-transitory computer-readable storage medium storing computer-coded instructions stored thereon that, when executed by at least one processor, are configured for: receiving a construction plan; maintaining an accuracy of an Al Platform, the Al platform comprising at least one algorithm and at least one model, by performing periodic updated training of the at least one model of the Al platform, wherein the at least one model of the Al platform is trained based at least in part on prior engineering data, prior construction data, or any combination of the prior engineering data and the prior construction data; processing the construction plan using the Al platform; and generating a Machine Control Model by inputting at least one model input to the Al platform, wherein the Machine Control Model is generated based on the processed construction plan, the at least one machine learning model of the Al platform, and the at least one model input.

[0285] Example Clause 21. The method of Clause 1, further comprising: validating at least one CAD fde with at least one certified PDF plan a same project, wherein validating comprises detecting and flagging at least one inconsistency, omission, or mismatch between the at least one CAD file and the at least one certified PDF plan.

[0286] Example Clause 22. The method of Clause 21, wherein validating the at least one CAD file with the at least one certified PDF plan comprises: overlaying CAD geometries of the at least one CAD file onto a rasterized version of the at least one certified PDF plan; and initiating an Al-based geometric and semantic comparison between the CAD geometries of the at least one CAD file and the rasterized versionPatent Application Attorney Docket # 073233-00006 of the at least one certified PDF plan.

[0287] Example Clause 23. The method of Clause 21, wherein validating the at least one CAD file with the at least one certified PDF plan comprises: comparing a plurality of CAD points with PDF plan data of the at least one certified PDF plan using a neural network classifier.

[0288] Example Clause 24. The method of Clause 21, further comprising: identifying and separating vector-based PDF data from raster-based PDF data in the at least one certified PDF plan, wherein the identifying and separating is performed utilizing a plurality of machine learning workflows optimized for interpreting each data type of a plurality of distinct data types prior to generating the Machine Control Model.

[0289] Example Clause 25. The method of claim 1, further comprising: generating a design improvement associated with the construction plan based at least in part on a determination from the Al platform learned from historical data and model creation; and outputting the design improvement.

[0290] Example Clause 26. The method of claim 1, wherein the Al platform is configured for: automatically standardizing model visuals across a plurality of projects by applying a predefined set of visual keys, colors, line types, and symbols to model layers of the Machine Control Model.

[0291] Example Clause 27. The method of Clause 26, wherein the visual keys are automatically adapted for region-specific workflows.

[0292] Example Clause 28. The method of Clause 1, further comprising: detecting an update to the construction plan or the Machine Control Model; and automatically generating, in real-time, an alert notifying of the update to each entity of a plurality of entities associated with the construction plan or the Machine Control Model.

[0293] Example Clause 29. The method of Clause 1, further comprising: generating inferred missing elevation data using visual pattern recognition in at least one area of insufficient TIN points.Patent Application Attorney Docket # 073233-00006

[0294] Example Clause 30. The method of Clause 1, further comprising: transmitting the Machine ControlModel to a construction machine, wherein the Machine Control Model controls the construction machine or informs control of the construction machine.

[0295] Example Clause 31. The method of Clause 1, wherein the construction plan comprises a rasterbased PDF construction plan, the method further comprising: extracting textual data and annotations from the raster-based PDF construction plan by a raster data extraction pipeline that uses artificial intelligence; and converting the textual data and annotations extracted from the raster-based PDF construction plan into geospatial and structured data utilized in generating the Machine Control Model.

[0296] Example Clause 32. The method of Clause 31, wherein the raster data extraction pipeline is configured to identify at least one arrowhead, at least one mapping point, or at least one annotation from the raster-based PDF construction plan using at least one trained neural network model.

[0297] Example Clause 33. The method of Clause 32, wherein the raster data extraction pipeline is configured to identify the at least one arrowhead using probabilistic Hough line transforms that are configured to transform a plurality of arrow styles and determine a directional orientation.

[0298] Example Clause 34. The method of claim 32, wherein the raster data extraction pipeline is configured to identify and match text identify and match at least one text annotation to at least one corresponding arrowhead using cone modeling and spatial proximity analysis.

[0299] Example Clause 35. The method of Clause 32, wherein the raster data extraction pipeline is configured to transform image-based coordinates into DWG model space coordinates using a direct linear transformation (DLT) technique.

[0300] Example Clause 36. The method of Clause 31, further comprising: extracting vector data from the raster-based PDF construction plan and automatically converting the vector data into DXF format using at least one scripted workflow.Patent Application Attorney Docket # 073233-00006

[0301] Example Clause 37. The method of Clause 31, wherein the raster data extraction pipeline is configured to project at least one arrowhead to detect associated DWG layers and assign at least one corresponding annotation.

[0302] Example Clause 38. The method of Clause 32, wherein the raster data extraction pipeline is configured to export a final output as structured JSON data representing relationships utilized for generating the Machine Control Model.

[0303] Example Clause 39. The method of claim 38, wherein the at least one arrowhead is detected based on geometric properties comprising at least one hatch pattern, at least one closed triangle, or at least one predefined size threshold.

[0304] Example Clause 40. The method of Clause 39, further comprising: determining at least one annotation layer having a highest density of detected arrows; and selecting, based on determining the at least one annotation layer having the highest density of detected arrows, the at least one annotation layer as at least one primary annotation layer.

[0305] Example Clause 41. The method of Clause 33, further comprising: extracting annotation text by retrieving MTEXT entities from the DXF format of the vector data.

[0306] Example Clause 42. The method of Clause 41, wherein at least one arrowhead is paired with corresponding MTEXT annotations of the MTEXT entities by evaluating shaft geometry and line endpoints within defined tolerance thresholds.

[0307] Example Clause 43. The method of Clause 1, wherein the Machine Control Model is deployed to a GNSS rover that performs a field stakeout, a coordinate validation, or a surface verification.

[0308] Example Clause 44. The method of Clause 1, wherein the Machine Control Model is configured for asphalt pavers to control paving thickness, material flow, and edge alignment based on the construction plan.Patent Application Attorney Docket # 073233-00006

[0309] Example Clause 45. The method of Clause 1, further comprising: transmitting the Machine ControlModel to an autonomous machine or a semi-autonomous line-painting machine that performs precise lane striping based on the Machine Control Model and in accordance with at least one site layout specification.

[0310] Example Clause 46. The method of Clause 1, wherein the Machine Control Model is rendered and transmitted to a layout printer configured to mark at least one reference line or at least one control point directly onto the jobsite surface using the Machine Control Model and based on a design specification.

[0311] Example Clause 47. The method of claim 1, further comprising: transmitting the Machine Control Model to a heavy civil excavator, wherein the Machine Control Model guides the heavy civil excavator equipped with a 3D machine control system to execute a precise grading and trenching activity.

[0312] Example Clause 48. The method of Clause 1, further comprising: transmitting the Machine Control Model to an automated or semi-automated bulldozer, wherein the Machine Control Model guides a blade elevation, a tilt, and a slope during at least one grading operation based on 3D surface data.

[0313] Example Clause 49. The method of Clause 1, further comprising: transmitting the Machine Control Model to a light transport and logistics (LTL) machine, wherein the Machine Control Model is used for material movement, routing, and delivery sequencing based at least in part on a site layout and construction phasing.

[0314] Example Clause 50. The method of Clause 1, further comprising: transmitting the Machine Control Model to an automated bulldozer or a semi-automated bulldozer, wherein the Machine Control Model guides a blade elevation, a tilt, and a slope during at least one grading operation based on 3D surface data.

[0315] Example Clause 51. A system for the creation of construction plans, comprising: (a) a site plan recognition component configured to: integrate basic site data from one or more sources; and organize the basic site data as a set of collected data; and (b) a model build component configured to: generate a construction plan based on the collected data, (c) an Al platform configured to: automatically generate aPatent Application Attorney Docket # 073233-00006Machine Control Model comprising at least one new construction plan based at least in part on one or more inputs using a construction plan algorithm trained on a library of prior construction plans and associated project data; and (d) a monitoring component configured to: maintain an Al platform accuracy by performing periodic updated training of the construction plan algorithm of the Al platform.

[0316] Example Clause 52. The system of Clause 51, wherein the Al platform is further trained on prior construction plan inputs, the prior construction plan inputs comprising at least one CAD drawing, at least one annotated PDF, or at least one surface model.

[0317] Example Clause 53. The system of Clause 51, wherein the Al platform is further trained on Machine Control Model outputs that were previously generated by the Al platform and verified during a previous construction.

[0318] Example Clause 54. The system of Clause 51, wherein the Al platform is further configured to use data from completed projects to reverse at least one engineer standard site layout pattern and optimize a future plan creation.

[0319] Example Clause 55. The system of Clause 51, wherein the Al platform is further configured to: receive and learn from a representation of local zoning codes and planning regulations across multiple jurisdictions; and automatically generate the construction plan that is compliant with applicable data of the local zoning codes and the planning regulations based on a project location.

[0320] Example Clause 56. The system of Clause 51, wherein the Al platform is further configured to: receive a construction plan designed for one jurisdiction or site; and automatically regenerate a new compliant construction plan for a different site based on learnings of the Al platform for a different at least one zoning rule or at least one locally applicable planning rule.

[0321] Example Clause 57. The system of Clause 51, further comprising a user preference management engine configured to: track, store, and apply individual user preferences, location-specific preferences,Patent Application Attorney Docket # 073233-00006 and company-level configuration settings across a plurality of design inputs, a plurality of models, and a plurality of workflows.

[0322] Example Clause 58. The system of Clause 51, wherein the Al platform is configured to: generate a plurality of construction plans, each comprising at least one distinct option, in response to a given input, wherein each plan option of the plurality of construction plans is based on an individual user, a company, and local demographic preferences; and output the plurality of construction plans to a user for selection; and receive a selection by the user based on an interaction with the plurality of construction plans outputted to the user, wherein the Al platform is configured to learn based on the selection by the user.

[0323] Example Clause 59. The system of Claim 51 , wherein the Al platform is configured to generate the Machine Control Model based on one or more inputs comprising at least GPS coordinates and a localization input.

[0324] Example Clause 60. The system of Clause 51, further comprising: a validation component configured to validate the construction plan created, or otherwise input for validation, with one or more design standards or preferences, the one or more design standards or preferences including user design standards or preferences, company design standards or preferences, or demographic standards or preferences.

[0325] Example Clause 61. A method comprising: receiving geospatial data, positioning data, or topographic data; processing the received geospatial data, positioning data, or topographic data through an Al platform comprising one or more algorithms and machine learning models trained on at least one prior engineering pattern, at least one prior grading pattern, and at least one prior construction pattern; generating a machine control model directly from the processed geospatial inputs; and training the Al platform with additional data and feedback.

[0326] Example Clause 62. The method of Clause 61, wherein the geospatial data, the positioning data,Patent Application Attorney Docket # 073233-00006 or the topographic data comprises GPS points, survey data, drone capture data, or LiDAR scan data, and the geospatial data, the positioning data, or the topographic data lacks a construction plan or a CAD plan.

[0327] Example Clause 63. A system for the creation of architectural plans, the system comprising: a plan recognition component configured to: integrate data from one or more relevant sources; and store the data as a set of collected data; and a model build component configured to: generate an architectural plan based on a received survey, design data, and other site data, wherein the architectural plan is generated by an architectural plan algorithm, the architectural plan algorithm trained based at least in part on the set of collected data.

[0328] Example Clause 64. The system of claim 63, further comprising: a monitoring component configured to indicate a deviation in the architectural plan algorithm.

[0329] Example Clause 65. The system of claim 63, further comprising: a maintenance component configured to update the architectural plan algorithm based on additional data in the set of collected data.

[0330] Example Clause 66. The system of claim 63, wherein the architectural plan is generated based on a description of a desired outcome and one or more inputs comprising at least data associated with walls, data associated with doors, data associated with property boundaries, data associated with size of a building envelope or other architectural feature, and data associated with shape of the building envelope.

[0331] Example Clause 67. The system of claim 63, further comprising: a validation component configured to validate the architectural plan against one or more standards.

[0332] Example Clause 68. A system for the creation of site plans, the system comprising: a plan recognition component configured to: integrate data from one or more relevant sources; and store the data as a set of collected data; and a model build component configured to: generate a site plan based on a received survey, design data, and other site data, wherein the site plan is generated by a site plan algorithm, the site plan algorithm trained based at least in part on the set of collected data.Patent Application Attorney Docket # 073233-00006

[0333] Example Clause 69. The system of Clause 68, further comprising: a monitoring component configured to indicate a deviation detected in the site plan algorithm.

[0334] Example Clause 70. The system of Clause 68, further comprising: a maintenance component configured to update the site plan algorithm based on additional data in the set of collected data.

[0335] Example Clause 71. The system of Clause 68, wherein the model build component generates the site plan based on a description of a desired outcome and one or more inputs, the one or more inputs comprising open spaces data, contour lines, soil condition data, drainage conditions data, or surrounding road input data.

[0336] Example Clause 72. The system of Clause 68, further comprising: a validation component configured to validate the site plan with one or more standards.CONCLUSION

[0337] The present technology may be embodied as, among other things, a system, method, or computer product. Accordingly, embodiments may take the form of a hardware embodiment, a software embodiment, or an embodiment combining software and hardware. In one embodiment, the present invention takes the form of a computer program product that includes the computer-usable instructions embodied on one or more computer-readable media and executed by one or more processors. Some embodiments can be embodied as one or more, or combination of, computer software, a computer program, an application, one or more engines, one or more modules, and / or one or more hardware and / or software components. In some aspects, components of systems described herein can be employed as a distributed system or centralized system.

[0338] Computer-readable media includes volatile and nonvolatile media, removable and nonremovable media, and media readable by a database, a switch, and various other network devices. Network switches, routers, access points, and related components, in some instances, act as a means ofPatent Application Attorney Docket # 073233-00006 communication within the scope of the technology. By way of example, computer-readable media comprise computer storage media and communications media.

[0339] Computer storage media or machine-readable media can include media implemented in any method or technology for storing and / or transmitting information or data. Examples of such information include computer-usable instructions, data elements, data structures, programs and program modules, and other data representations.

[0340] Communications media generally store computer-usable or readable instructions, including data structures and program modules in a modulated data signal. A modulated data signal, in some instances, can be understood to be a propagated signal that has one or more of its characteristics set or changed to encode information in the signal. Communications media include any information-delivery media. By way of example and not limitation, communications media include wired media, such as a wired network or direct-wired connection, and wireless media, such as radio, cellular, spread-spectrum, and other wireless media technologies. Combinations of the above are included with the scope of computer-readable media and communications media.

[0341] Embodiments described herein can be understood more readily by reference to the examples described above. Elements, apparatus, and methods described herein, however, are not limited to any specific embodiment presented in the Examples. It should be recognized that these are merely illustrative of some principles of this disclosure and are non-limiting. Numerous modifications and adaptations will be readily apparent without departing from the spirit and scope of the disclosure.

[0342] Many different arrangements of the various components and / or steps depicted and described, as well as those not shown, are possible without departing from the scope of the claims below.Embodiments of the present technology have been described with the intent to be illustrative rather than restrictive. Alternative embodiments will become apparent from reference to this disclosure. AlternativePatent Application Attorney Docket # 073233-00006 means of implementing the aforementioned can be completed without departing from the scope of the claims below. Certain features and sub-combinations are of utility and can be employed without reference to other features and sub-combinations and are contemplated within the scope of the claims.

Claims

Patent Application Attorney Docket # 073233-00006CLAIMS1. A method comprising: receiving a construction plan; maintaining an accuracy of an Al Platform, the Al platform comprising at least one algorithm and at least one model, by performing periodic updated training of the at least one model of the Al platform, wherein the at least one model of the Al platform is trained based at least in part on prior engineering data, prior construction data, or any combination of the prior engineering data and the prior construction data; processing the construction plan using the Al platform; and generating a Machine Control Model by inputting at least one model input to the Al platform, wherein the Machine Control Model is generated based on the processed construction plan, the at least one machine learning model of the Al platform, and the at least one model input.

2. The method of claim 1, wherein processing the construction plan by the Al platform comprises: indicating, by the Al platform, a subset of data in the construction plan as determined to be unnecessary data for visual presentation or machine performance; and removing, by the Al platform, the subset of data from the construction plan.

3. The method of claim 1, wherein generating the Machine Control Model comprises: correcting, by the Al platform, at least one surface irregularity generated using at least one triangular irregular network surface algorithm.

4. The method of claim 3, wherein the correcting the at least one surface irregularity comprises: using at least one computer-generated break line to correct the at least one surface irregularity; or using at least one geometric algorithm to correct the at least one surface irregularity; or using the at least one machine learning model to correct the at least one surface irregularity.Patent Application Attorney Docket # 073233-000065. The method of claim 1, wherein correcting the at least one surface irregularity comprises: reducing a file size of the Machine Control Model by removing at least one oversaturated point after creating a high-fidelity elevation model.6 The method of claim 1, wherein the Al platform is configured for: determining a representation of at least one zoning law or at least one technical standard applicable to the construction plan or the Machine Control Model; and validating the construction plan or the Machine Control Model satisfies the at least one zoning law or the at least one technical standard.7 The method of claim 1, wherein the Al platform is configured for: extracting geometric and geospatial data from the construction plan; and transforming the geometric and geospatial data from the construction plan into structured data further processed in generating the Machine Control Model.8 The method of claim 1, wherein the Machine Control Model is configured to self-update by regenerating a new version of the Machine Control Model after receiving an updated version of data sed in during generating of the Machine Control Model.9 The method of claim 8, wherein the Al platform is configured for: receiving, automatically from at least one field data collector associated with at least one site, at least one model data point collected by the at least one field data collector from the site.10 The method of claim 9, wherein the at least one field data collector is associated with a specific machine control model and collects as-implemented data points associated with the specific Machine Control Model, in real time or near real time, and wherein the Al platform is configured for using the as- implemented data points to validate an implementation of the construction plan.11 The method of claim 10, the method further comprising: detecting at least one change in field measurements represented by the as-implemented dataPatent Application Attorney Docket # 073233-00006 points; and in response to detecting the at least one change in field measurements, trigger an alert to update the Machine Control Model.

12. The method of claim 1, further comprising: updating the Machine Control Model to generate a new version of the Machine Control Model based at least in part on additional data.

13. The method of claim 1, wherein the Al Platform is configured for: automatically validating an implementation of the construction plan by cross-checking as- implemented data points associated with the implementation as-implemented and that are received by the Al platform with the Machine Control Model.

14. The method of claim 1, further comprising: receiving equipment telemetry data associated with at least one equipment; and generating a simulated carbon emissions data based on simulated machine usage of the at least one equipment and the equipment telemetry data.

15. The method of claim 14, wherein the simulated carbon emissions data is dynamically updated based on a detected change in a workflow, a material volume, or site logistics data.

16. The method of claim 1, further comprising: generating, based on the Machine Control Model, at least one ESG compliance report summarizing a project impact, a project safety, a project carbon footprint, or a project audit trail.

17. The method of claim 1, further comprising: generating an auditing log that indicates at least one detail associated with each of the generation of the Machine Control Model, a validation of the Machine Control Model, and an update of the Machine Control Model, wherein the auditing log is automatically generated in an auditable format.Patent Application Attorney Docket # 073233-0000618. The method of claim 17, wherein the auditing log comprises at least one timestamp, at least one user approval, and at least one compliance validation mapped to a specific version of the Machine Control Model.

19. A system comprising at least one non-transitory memory and at least one processor, the at least one non-transitory memory storing computer-coded instructions that, when executed by the at least one processor, cause the system to: receive a construction plan; maintain an accuracy of an Al Platform, the Al platform comprising at least one algorithm and at least one model, by performing periodic updated training of the at least one model of the Al platform, wherein the at least one model of the Al platform is trained based at least in part on prior engineering data, prior construction data, or any combination of the prior engineering data and the prior construction data; process the construction plan using the Al platform; and generate a Machine Control Model by inputting at least one model input to the Al platform, wherein the Machine Control Model is generated based on the processed construction plan, the at least one machine learning model of the Al platform, and the at least one model input.

20. A non-transitory computer-readable storage medium storing computer-coded instructions stored thereon that, when executed by at least one processor, are configured for: receiving a construction plan; maintaining an accuracy of an Al Platform, the Al platform comprising at least one algorithm and at least one model, by performing periodic updated training of the at least one model of the Al platform, wherein the at least one model of the Al platform is trained based at least in part on prior engineering data, prior construction data, or any combination of the prior engineering data and the prior construction data; processing the construction plan using the Al platform; and generating a Machine Control Model by inputting at least one model input to the Al platform, wherein the Machine Control Model is generated based on the processed construction plan, the at least one machine learning model of the Al platform, and the at least one model input.

Images