Decision planning method and decision planning device for excavator
By integrating machine learning models in the excavator and using global three-dimensional environment diagram information for independent decision-making and trajectory planning, the problem that traditional excavators are difficult to operate efficiently in complex scenarios is solved, and more efficient and safer independent operations are achieved.
Patent Information
- Application Number
- CN202510112089.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-01-23
- Publication Date
- 2025-05-13
AI Technical Summary
Traditional manual operation excavators are difficult to meet the safety, efficiency, precision and low-cost operation needs, especially in complex and changeable operation scenarios.
By obtaining the excavator's position state, motion state, load state and global three-dimensional environment diagram information, the machine learning model is used to determine the target task guidance information, and based on this information, the future trajectory of the excavator is planned to achieve independent decision-making and trajectory planning.
It realizes independent decision-making in complex and changing scenarios, improves the adaptability of independent operation scenarios, improves the efficiency and quality of excavator operations, and reduces manual participation and labor costs.
Smart Images
Figure CN119987202A_ABST
Abstract
Description
Technical Field
[0001] The present disclosure relates to the technical field of intelligent engineering machinery, and in particular to an excavator decision-making planning method and a decision-making planning device. Background Art
[0002] With the rapid development of technologies such as artificial intelligence and the continuous upgrading of the industry, traditional manually operated excavators have gradually failed to meet the needs of safe, efficient, precise and low-cost operations. Intelligent excavators can perceive, understand and make decisions about the working environment by integrating advanced sensor technology, computer vision, machine learning algorithms, etc. They can automatically plan the working path, adjust the excavation action, avoid obstacles according to the requirements of the construction task, and continuously optimize the working strategy by collecting and analyzing the working data, improving the working safety, working efficiency and working precision, while reducing manual participation and reducing labor costs. Summary of the invention
[0003] The disclosed embodiment determines the target task guidance information based on the global three-dimensional environment map information; based on the guidance of the target task guidance information, the trajectory information of the excavator in the future is determined according to the excavator's posture state information, motion state information, load state information and global three-dimensional environment map information. Thus, autonomous task decision-making and trajectory planning are realized, and autonomous decision-making in complex and changeable scenes is realized based on the global three-dimensional environment map information, improving the adaptability of autonomous operation scenes without human participation, and based on the guidance of the target task guidance information, the motion trajectory is quickly and accurately planned autonomously, improving the excavator's operation efficiency and operation quality.
[0004] Some embodiments of the present disclosure propose a decision-making planning method, including: obtaining the excavator's posture state information, motion state information, load state information and global three-dimensional environment map information; determining target task guidance information based on the global three-dimensional environment map information; based on the guidance of the target task guidance information, determining the excavator's trajectory information in the future according to the excavator's posture state information, motion state information, load state information and global three-dimensional environment map information.
[0005] In some embodiments, determining the trajectory information of the excavator at a future time includes: determining the historical local three-dimensional environment map information, historical trajectory information and historical load information of the excavator based on the excavator's posture state information, motion state information, load state information and global three-dimensional environment map information; based on the guidance of the target task guidance information, determining the trajectory information of the excavator at a future time based on the excavator's historical local three-dimensional environment map information, historical trajectory information and historical load information.
[0006] In some embodiments, determining the historical local three-dimensional environment map information of the excavator includes: selecting a focus area based on the global three-dimensional environment map information and the range of the working space in which the excavator can move within a first preset time; cropping the global three-dimensional environment map with the excavator as the center to obtain a local three-dimensional environment map of the size of the focus area.
[0007] In some embodiments, the length and width of the area of interest are determined based on the maximum travel speed of the excavator chassis, the first preset time, and the farthest telescopic distance of the excavator tooth tip; the height of the area of interest is determined based on the highest telescopic distance of the excavator tooth tip.
[0008] In some embodiments, the region of interest is represented as W×L×H, where , , W is the width of the area of interest; L is the length of the area of interest; H is the height of the area of interest; , is the proportionality factor of interest; is the maximum travel speed of the excavator chassis; The maximum telescopic distance of the excavator tooth tip; is the maximum telescopic distance of the excavator tooth tip, and n is the first preset time.
[0009] In some embodiments, the historical trajectory information of the excavator is determined based on the posture state information and motion state information within the previous second preset time; the historical load information of the excavator is determined based on the load state information within the previous second preset time.
[0010] In some embodiments, determining target task guidance information based on global three-dimensional environmental map information includes: using a first machine learning model to process the global three-dimensional environmental map information to obtain target task guidance information, wherein the first machine learning model is learned from the driver's excavator operation data, wherein the excavator operation data is marked with work tasks and guidance points in the current environment.
[0011] In some embodiments, determining the trajectory information of the excavator at a future time includes: using a second machine learning model to process the target task guidance information, the historical local three-dimensional environmental map information of the excavator, the historical trajectory information and the historical load information to obtain the trajectory information of the excavator at a future time, wherein the second machine learning model is trained using a data set, the data set is obtained based on the excavator operation data of the driver, and the data set is a collection of the work task and its guidance points, environmental information, trajectory information, load information and actual motion trajectory information.
[0012] In some embodiments, the posture state information includes at least one of the center position of the excavator chassis and the position of the bucket tooth tip; the motion state information includes the rotation speed of the left and right travel motors of the excavator chassis, the rotation angle of the upper vehicle slewing device, the boom lifting angle, the angle between the boom and the boom, and the angle between the bucket and the boom; and / or the load state information includes at least one of the pressure difference between the oil inlet and outlet ports of the left and right travel motors of the excavator chassis, the pressure difference of the upper vehicle slewing motor, the force of the boom cylinder, the force of the boom cylinder, and the force of the bucket cylinder.
[0013] In some embodiments, the target task guidance information includes at least one of the target task category and the position coordinates; and / or the trajectory information at future time includes at least one of the posture state information and motion state information of the excavator at future time.
[0014] Some embodiments of the present disclosure provide a decision planning device, comprising: a memory; and a processor coupled to the memory, wherein the processor is configured to execute a decision planning method based on instructions stored in the memory.
[0015] Some embodiments of the present disclosure provide a decision planning device, including: a module for executing a decision planning method.
[0016] Some embodiments of the present disclosure provide an excavator, comprising: a decision-making planning device configured to execute a decision-making planning method.
[0017] Some embodiments of the present disclosure provide a computer-readable storage medium having computer instructions stored thereon, which implement a decision planning method when executed by a processor.
[0018] Some embodiments of the present disclosure provide a computer program product, comprising computer instructions, wherein the computer instructions implement a decision planning method when executed by a processor. BRIEF DESCRIPTION OF THE DRAWINGS
[0019] The following is a brief introduction to the drawings required for use in the embodiments or related technical descriptions. The present disclosure can be more clearly understood according to the following detailed description with reference to the drawings.
[0020] Obviously, the drawings described below are only some embodiments of the present disclosure, and a person skilled in the art can obtain other drawings based on these drawings without creative work.
[0021] Figure 1 A schematic diagram showing a decision planning device according to some embodiments of the present disclosure.
[0022] Figure 2 A schematic diagram showing a decision planning method according to some embodiments of the present disclosure.
[0023] Figure 3 A schematic diagram showing a decision planning device according to some embodiments of the present disclosure. DETAILED DESCRIPTION
[0024] It should be noted that the relative arrangement of components and steps, the numerical expressions and numerical values set forth in these embodiments do not limit the scope of the present disclosure unless specifically stated otherwise.
[0025] Those skilled in the art can understand that the terms "first" and "second" in the embodiments of the present disclosure are only used to distinguish different steps, devices or modules, etc., and neither represent any specific technical meaning nor indicate the necessary logical order between them.
[0026] It should also be understood that in the embodiments of the present disclosure, “plurality” may refer to two or more than two, and “at least one” may refer to one, two, or more than two.
[0027] It should also be understood that any component, data or structure mentioned in the embodiments of the present disclosure can generally be understood as one or more, unless explicitly limited or otherwise indicated in the context.
[0028] In addition, the term "and / or" in the present disclosure is only a description of the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / " in the present disclosure generally indicates that the associated objects before and after are in an "or" relationship.
[0029] It should also be understood that the description of the various embodiments in the present disclosure focuses on the differences between the various embodiments, and the same or similar aspects thereof can be referenced to each other, and for the sake of brevity, they will not be described one by one.
[0030] At the same time, it should be understood that for the convenience of description, the sizes of the various parts shown in the drawings are not drawn according to the actual proportional relationship.
[0031] The following description of at least one exemplary embodiment is merely illustrative in nature and is in no way intended to limit the present disclosure, its application, or uses.
[0032] Technologies, methods, and equipment known to ordinary technicians in the relevant art may not be discussed in detail, but where appropriate, the technologies, methods, and equipment should be considered part of the specification.
[0033] It should be noted that like reference numerals and letters refer to similar items in the following figures, and therefore, once an item is defined in one figure, it need not be further discussed in subsequent figures.
[0034] In addition, in order to avoid obscuring the present disclosure due to unnecessary details, only the processing steps and / or device structures closely related to at least the scheme according to the present disclosure are shown in the drawings, and other details that are not closely related to the present disclosure are omitted. It should also be noted that similar reference numerals and letters in the drawings indicate similar items, and therefore once an item is defined in one drawing, it does not need to be discussed again for subsequent drawings.
[0035] Figure 1 A schematic diagram showing a decision planning device according to some embodiments of the present disclosure.
[0036] like Figure 1 As shown, the decision-making and planning device of this embodiment is used for task decision-making and trajectory planning, and includes: a data processing module, a decision module and a planning module. The data processing module inputs the current posture state information, motion state information, load state information and global three-dimensional environment map information of the excavator, and outputs historical local three-dimensional environment map information, historical trajectory information and historical load sequence information. The decision module inputs the global three-dimensional environment map information and outputs the target task guidance information (target task guidance point); the planning module inputs the target task guidance information (target task guidance point), historical local three-dimensional environment map information, historical trajectory information and historical load information, and outputs the future trajectory information of the excavator (i.e., the trajectory information at future time).
[0037] The following describes each input information and output information.
[0038] Global 3D environment map information: refers to the surrounding environment information perceived by the excavator, such as the position and shape information of materials, trucks, the excavator itself and other objects, and its own information includes the material information in the bucket. Global 3D environment map information is perceived and obtained through cameras, laser radars, etc.
[0039] Position state information: for example, the center position of the excavator chassis, set to (0,0,0), and the position of the bucket tooth tip. The position state information is sensed and obtained through, for example, the global navigation satellite system (GNSS), inertial measurement unit (IMU), laser tracker, etc.
[0040] Motion state information: for example, the speed of the left and right travel motors of the excavator chassis, the rotation angle of the upper vehicle slewing device, the boom lifting angle, the angle between the boom and the boom, and the angle between the bucket and the boom. Motion state information is sensed and obtained through speed sensors, angle sensors, encoders, etc.
[0041] Load status information: For example, it includes the pressure difference between the inlet and outlet ports of the left and right travel motors of the excavator chassis, the pressure difference of the upper vehicle rotary motor, the force of the boom cylinder, the force of the arm cylinder, and the force of the bucket cylinder. The load status information is sensed and obtained through pressure sensors, torque sensors, etc.
[0042] Historical local 3D environment map information: 3D environment map of the local area of the excavator within the previous m seconds (including the current time, for example, within the previous 2 seconds, with an interval of 100ms). m seconds is the preset time, and other times can be configured.
[0043] Historical trajectory information: the excavator’s position status information and motion status information within the previous m seconds (including the current time, for example, within the previous 2 seconds, with an interval of 100ms).
[0044] Historical load information: load status information of the excavator within the previous m seconds (including the current time, for example, within the previous 2 seconds, with an interval of 100ms).
[0045] Target task guidance information: for example, including the target task category (such as excavation, loading, leveling, etc.), and the coordinates of a certain position in the global three-dimensional environment map (i.e., the guidance point).
[0046] Future trajectory information: the excavator's position and motion status information in the next n seconds (for example, in the next 3 seconds, with an interval of 100ms). n seconds is the preset time, and other times can be configured.
[0047] Data processing module: It is configured to cache m seconds of historical information, crop the global 3D environment map into a local 3D environment map of the area of interest, and output it to the planning module.
[0048] Decision module: It is configured to output target task guidance information based on the global three-dimensional environment map information as the input of the planning module, guiding the planned trajectory to gradually approach the correct task area.
[0049] Planning module: It is configured to plan the optimal motion trajectory for a period of time in the future based on various historical information and target task guidance information to complete various target tasks.
[0050] (1) Data processing module
[0051] 1) Historical local 3D environment map information processing
[0052] According to the input global three-dimensional environment map information, combined with the working space range that the excavator can move within n seconds, the focus area is selected. The focus area is represented by a rectangular block, for example. The length and width of the focus area are determined according to the maximum driving speed of the excavator chassis, the first preset time, and the farthest telescopic distance of the excavator tooth tip; the height of the focus area is determined according to the maximum telescopic distance of the excavator tooth tip.
[0053] The region of interest is represented as W×L×H, for example. , W is the width of the area of interest; L is the length of the area of interest; H is the height of the area of interest; , is the proportionality factor of interest; is the maximum travel speed of the excavator chassis; The maximum telescopic distance of the excavator tooth tip; is the maximum telescopic distance of the excavator tooth tip, and n is the first preset time.
[0054] According to the selected area of interest (W×L×H), the global 3D environment map is cropped with the excavator coordinates as the center to a local 3D environment map of size W×L×H.
[0055] Set up a storage cache to store historical local 3D environment map information within m seconds (100ms interval) according to the time dimension.
[0056] 2) Historical trajectory information processing
[0057] Set the storage cache to store historical trajectory information within m seconds (100ms interval) according to the time dimension.
[0058] 3) Historical load information processing
[0059] Set the storage cache to store historical load information within m seconds (100ms interval) according to the time dimension.
[0060] (2) Decision-making module
[0061] The decision module (set as D) is a mathematical model with learning ability, such as a machine learning model (the first machine learning model), which inputs the global three-dimensional environment map information (set as ), output the target task guidance information (set to ),refer to Figure 1 The processing process can be expressed by the formula as follows: .
[0062] The first machine learning model learning and training process can be based on a supervised approach, based on the excavator operation data of experienced drivers, marking the work tasks and guidance points in the current environment, and using them as the true value of model training to carry out model training until convergence, so that the model learns where the excavator should go and what to do in the current situation, so that the decision-making module has experienced target task selection capabilities and strong generalization capabilities, and uses guidance points to issue work tasks to the planning module.
[0063] (3) Planning module
[0064] The planning module (set as P) is a mathematical model with learning ability, such as a machine learning model (the second machine learning model), which inputs the historical local three-dimensional environment map information (set as ), historical trajectory information (set to ), historical load information (set to ) and target task guidance information (set to ), output future trajectory information (set to ),refer to Figure 1 The processing process can be expressed by the formula as follows: .
[0065] The second machine learning model learning and training process can be based on a supervised approach, based on the excavator operation data of experienced drivers, processed into a data set (a collection of work tasks and their guide points, environmental information, trajectory information, load information and actual motion trajectory information), and the trajectory sequence based on the time dimension will be used as the true value of the model training to carry out model training until convergence, so that the model learns what the excavator does under the current circumstances, so that the planning module has fast, accurate and smooth trajectory planning capabilities, and uses future trajectory sequences as a reference for control execution.
[0066] Here are some examples of what you can use as your first machine learning model.
[0067] Convolutional Neural Network (CNN)
[0068] Principle: CNN performs well in processing images and spatial data. Its core components include convolutional layers, pooling layers, and fully connected layers. The convolutional layer slides the convolution kernel on the data to perform convolution operations and extract local features; the pooling layer compresses the features to reduce the amount of data while retaining key information; the fully connected layer integrates all features and outputs the final result. For a three-dimensional environment map, it can be regarded as a series of two-dimensional slices, and CNN can automatically learn various feature patterns in the environment.
[0069] Example: Assume that the global 3D environment map of the excavator's working environment is represented in a voxel-like (3D pixel) format. These voxel data are input into the CNN model, and the convolution kernel of the convolution layer slides on different voxel slices to capture features such as obstacles (such as boulders, building remains), excavation areas (such as the mine area to be excavated), and terrain undulations. For example, the CNN model recognizes that there is an excavation area with a specific shape and depth requirement ahead. Based on the learning of these features, the CNN model outputs the target task guidance information, such as "go to the coordinates (x, y, z) to prepare for excavation operations, and the excavation depth is h".
[0070] Semantic segmentation network (such as U-Net)
[0071] Principle: The semantic segmentation network aims to classify each pixel in an image into a specific category. U-Net is a successful semantic segmentation model that has an encoder, a decoder, and a jump connection. The encoder part is used for downsampling and gradually extracting high-level features; the decoder restores the feature map to the original image size through upsampling and makes classification predictions; the jump connection connects the low-level features of the encoder to the corresponding layer of the decoder to retain the detail information. For 3D environment maps, it can be extended to process 3D data to achieve classification of each voxel.
[0072] Example: In an excavator operation scenario, the U-Net model processes the global 3D environment map and classifies each voxel into different categories, such as "digable area", "undigable obstacle", "digged area", etc. By identifying and analyzing these categories, the target task guidance information is obtained. For example, if a large area of "digable area" is detected and the current position is close to the area, the model may output the task guidance information of "move forward to the edge of the digable area and start digging".
[0073] Transformer-based models
[0074] Principle: The Transformer architecture is based on the self-attention mechanism, which can effectively capture long-distance dependencies in sequence data. When processing a three-dimensional environment map, each position in the three-dimensional space can be regarded as an element in the sequence. Through the self-attention mechanism, the model can dynamically pay attention to the relationship between different positions, thereby better understanding the overall structure and semantic information of the environment.
[0075] Example: Convert the global 3D environment map into a sequence form and input it into the Transformer-based model. The model uses the self-attention mechanism to analyze the spatial relationship between different areas, such as judging the distance and relative position between the area to be excavated and the surrounding obstacles. For example, when obstacles are detected around the area to be excavated, the target task guidance information generated by the model may be "bypass the obstacles first, approach the area to be excavated from the side, and then start the excavation operation."
[0076] Here are some examples of what could be used as a second machine learning model.
[0077] Long Short-Term Memory (LSTM)
[0078] Principle: LSTM is a special recurrent neural network (RNN) designed specifically to handle long-term dependency problems in time series data. It can effectively control the inflow, outflow and long-term storage of information by introducing input gates, forget gates and output gates, as well as memory units. When processing excavator-related data, LSTM can capture the changing patterns of historical data over time, and combine the target task guidance information to predict future trajectories.
[0079] Example: Assume that the historical local 3D environment map information is represented in the form of a time series, recording the changes in the local environment around the excavator at different times; the historical trajectory information records the position and movement direction of the excavator at each time point; the historical load information reflects the load of the excavator at different operation stages. Input these time series data and target task guidance information (such as excavation depth, coordinates of the excavation area, etc.) into the LSTM model. The LSTM model predicts the trajectory of the excavator in the future by learning patterns in historical data, such as how the excavator adjusts its trajectory according to task requirements under specific local environment and load conditions, such as predicting that the excavator will move in a certain direction at a certain speed to reach the target excavation area.
[0080] Gated Recurrent Unit (GRU)
[0081] Principle: GRU is a simplified variant of LSTM and is also used to process time series data. It merges the input gate and forget gate in LSTM into an update gate, and merges the memory unit and hidden state, simplifying the model structure while still being able to effectively capture long-term dependencies in time series. GRU models the input data by deciding which information to retain or update through a gating mechanism.
[0082] Example: For the time series of the excavator's historical local 3D environment map, historical trajectory, and historical load information, the GRU model can learn the dynamic relationship between them. For example, when the target task guidance information requires digging an area of a specific shape under specific terrain conditions, the GRU model can learn from historical data the trajectory pattern of the excavator under similar terrain and task requirements. Based on this, it is predicted that the excavator will first rotate to the left by a certain angle and then move forward a certain distance in the future to meet the requirements of the excavation task.
[0083] Multilayer Perceptron (MLP) combined with reinforcement learning (such as DQN - MLP)
[0084] Principle: Multilayer Perceptron (MLP) is a simple feedforward neural network consisting of an input layer, a hidden layer, and an output layer, which can learn complex nonlinear relationships between input and output. When combined with reinforcement learning, such as the Deep Q Network (DQN), MLP can be used as an approximator of the Q function. The excavator takes different actions (such as moving forward, turning, lifting the excavating arm, etc.) in different states (consisting of target task guidance information, historical local three-dimensional environmental map information, historical trajectory information, and historical load information) and obtains rewards (such as the progress and efficiency of completing the excavation task) by interacting with the environment. The DQN-MLP model continuously learns and optimizes strategies to select actions that can obtain the maximum cumulative reward in the current state, thereby predicting future trajectories.
[0085] Example: The current state information of the excavator is encoded and input into the MLP, which outputs the Q value of each possible action. For example, when facing a specific excavation task (target task guidance information) and in a specific local environment (historical local 3D environment map information), combined with its own historical trajectory and load information, the model evaluates that the action of "moving forward" has the highest Q value, that is, it predicts that the excavator should move forward next. As it continuously interacts with the environment and updates the model, the model can more accurately predict future trajectories that meet the task requirements, such as how to adjust the direction and speed of movement according to real-time conditions during the excavation process.
[0086] Graph Neural Networks (GNN)
[0087] Principle: Graph neural networks are suitable for processing data with graph structures. In the scenario of an excavator, the historical local 3D environment map information, historical trajectory information, and historical load information can be constructed into a graph structure. For example, different locations (points in the historical trajectory) are used as nodes, and the spatial relationship, time sequence relationship, and load change relationship between nodes are used as edges. GNN learns the feature representation of nodes through the message passing mechanism between nodes, thereby capturing the complex relationship between data. Combined with the target task guidance information, GNN can predict the future trajectory of the excavator.
[0088] Example: Assume that the target task is to dig a specific path in a complex terrain. The GNN model analyzes the graph structure constructed by historical data to understand the accessibility between different locations, the terrain difficulty (converted from the local 3D environment map information), and the impact of load changes on movement. Based on this, the future trajectory of the excavator is predicted, such as bypassing a node with complex terrain (high load area) first, choosing a relatively flat path, and moving towards the target excavation path.
[0089] Figure 2A schematic diagram showing a decision planning method according to some embodiments of the present disclosure.
[0090] like Figure 2 As shown, the decision planning method of this embodiment includes the following steps.
[0091] In step 210, the position state information, motion state information, load state information and global three-dimensional environment map information of the excavator are obtained.
[0092] In step 220, target task guidance information is determined based on the global three-dimensional environment map information.
[0093] The global three-dimensional environment map information is processed using a first machine learning model to obtain target task guidance information, wherein the first machine learning model is learned from the driver's excavator operation data, wherein the excavator operation data is annotated with the operation tasks and guidance points in the current environment.
[0094] In step 230, based on the guidance of the target task guidance information, the trajectory information of the excavator in the future is determined according to the posture state information, motion state information, load state information and global three-dimensional environment map information of the excavator.
[0095] The step of determining the trajectory information of the excavator at a future time includes, for example, step 231 and step 232 .
[0096] In step 231, based on the posture state information, motion state information, load state information and global three-dimensional environment map information of the excavator, the historical local three-dimensional environment map information, historical trajectory information and historical load information of the excavator are determined.
[0097] Determining the historical local three-dimensional environment map information of the excavator includes: selecting the focus area according to the global three-dimensional environment map information and the range of the working space that the excavator can move within the first preset time; determining the length and width of the focus area according to the maximum driving speed of the excavator chassis, the first preset time, and the farthest telescopic distance of the excavator tooth tip; determining the height of the focus area according to the highest telescopic distance of the excavator tooth tip. With the excavator as the center, the global three-dimensional environment map is cropped to obtain a local three-dimensional environment map of the size of the focus area.
[0098] The historical trajectory information of the excavator is determined based on the posture state information and motion state information within the previous second preset time.
[0099] The historical load information of the excavator is determined according to the load status information within the previous second preset time.
[0100] In step 232, based on the guidance of the target task guidance information, the trajectory information of the excavator in the future is determined according to the historical local three-dimensional environment map information, historical trajectory information and historical load information of the excavator.
[0101] The target task guidance information, the historical local three-dimensional environment map information, the historical trajectory information and the historical load information of the excavator are processed by the second machine learning model to obtain the trajectory information of the excavator in the future. The second machine learning model is trained by using a data set, the data set is obtained according to the excavator operation data of the driver, and the data set is a collection of the operation task and its guidance point, environment information, trajectory information, load information and actual motion trajectory information.
[0102] The disclosed embodiments implement autonomous decision-making in complex and ever-changing scenarios based on global three-dimensional environmental information, and improve the adaptability of autonomous operation scenarios without human intervention. The conversion relationship between historical states is acquired and utilized, and combined with target task guidance information, fast, accurate, and smooth trajectory planning is implemented, thereby improving the operating efficiency, operating quality, and operating smoothness of the excavator.
[0103] Figure 3 A schematic diagram showing a decision planning device according to some embodiments of the present disclosure is shown. Figure 3 As shown, the decision planning device 300 of this embodiment includes: a memory 310 and a processor 320 coupled to the memory 310 , and the processor 320 is configured to execute the decision planning method in each embodiment based on the instructions stored in the memory 310 .
[0104] The decision planning device 300 may further include an input / output interface 330 , a network interface 340 , a storage interface 350 , etc. These interfaces 330 , 340 , 350 , the memory 310 , and the processor 320 may be connected, for example, via a bus 360 .
[0105] The memory 310 may include, for example, a system memory, a fixed non-volatile storage medium, etc. The system memory may store, for example, an operating system, an application program, a boot loader, and other programs.
[0106] The processor 320 may be implemented by a general purpose processor, a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, discrete gates or transistors and other discrete hardware components.
[0107] Among them, the input and output interface 330 provides a connection interface for input and output devices such as a display, a mouse, a keyboard, and a touch screen. The network interface 340 provides a connection interface for various networked devices. The storage interface 350 provides a connection interface for external storage devices such as SD cards and USB flash drives. The bus 360 can use any bus structure among a variety of bus structures. For example, the bus structure includes but is not limited to the Industry Standard Architecture (ISA) bus, the Micro Channel Architecture (MCA) bus, and the Peripheral Component Interconnect (PCI) bus.
[0108] An excavator includes: a decision-making planning device configured to execute the decision-making planning method in each embodiment.
[0109] It should be understood by those skilled in the art that the embodiments of the present disclosure may be provided as methods, systems, or computer program products. Therefore, the present disclosure may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present disclosure may take the form of a computer program product implemented on one or more (non-transient) computer-readable storage media (including but not limited to disk storage, CD-ROM, optical storage, cloud storage, etc.) containing computer program code. A computer program product should be understood as a software product that implements its solution mainly through a computer program.
[0110] The present disclosure is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present disclosure. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of the processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.
[0111] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.
[0112] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.
[0113] The above description is only a preferred embodiment of the present disclosure and is not intended to limit the present disclosure. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present disclosure should be included in the protection scope of the present disclosure.
Claims
1. A decision planning method, comprising: Obtaining the excavator's posture state information, motion state information, load state information, and global three-dimensional environment map information; Determine the target task guidance information based on the global three-dimensional environment map information; Based on the guidance of the target task guidance information, the trajectory information of the excavator in the future is determined according to the excavator's posture state information, motion state information, load state information and global three-dimensional environment map information.
2. The decision planning method according to claim 1, wherein: The trajectory information of the excavator at a future time includes: Determine the historical local three-dimensional environment map information, historical trajectory information and historical load information of the excavator according to the posture state information, motion state information, load state information and global three-dimensional environment map information of the excavator; Based on the guidance of the target task guidance information, the trajectory information of the excavator in the future is determined according to the historical local three-dimensional environmental map information, historical trajectory information and historical load information of the excavator.
3. The decision planning method according to claim 2, wherein: Determine the historical local 3D environment map information of the excavator including: Selecting a focus area based on the global three-dimensional environment map information and the range of the working space that the excavator can move within a first preset time; With the excavator as the center, the global 3D environment map is cropped to obtain a local 3D environment map of the size of the focus area.
4. The decision planning method according to claim 3, wherein: The length and width of the area of interest are determined according to the maximum travel speed of the excavator chassis, the first preset time, and the maximum telescopic distance of the excavator tooth tip; The height of the area of concern is determined by the maximum telescopic distance of the excavator tooth tip.
5. The decision planning method according to claim 4, wherein: The region of interest is represented as W×L×H, where , , W is the width of the area of interest; L is the length of the area of interest; H is the height of the area of interest; , is the proportionality factor of interest; is the maximum travel speed of the excavator chassis; The maximum telescopic distance of the excavator tooth tip; is the maximum telescopic distance of the excavator tooth tip, and n is the first preset time.
6. The decision planning method according to claim 2, wherein: Determine historical trajectory information of the excavator according to the posture state information and motion state information within the second preset time; The historical load information of the excavator is determined according to the load state information within the previous second preset time.
7. The decision planning method according to claim 1 or 2, wherein: According to the global three-dimensional environment map information, the target task guidance information is determined to include: Using the first machine learning model, the global three-dimensional environment map information is processed to obtain the target task guidance information. The first machine learning model is learned from the driver's excavator operation data, wherein the excavator operation data is annotated with the work tasks and guidance points in the current environment.
8. The decision planning method according to claim 2, wherein: The trajectory information of the excavator at a future time includes: The second machine learning model is used to process the target task guidance information, the historical local 3D environment map information, historical trajectory information and historical load information of the excavator to obtain the trajectory information of the excavator in the future. Among them, the second machine learning model is trained using a data set, and the data set is obtained based on the driver's excavator operation data. The data set is a collection of work tasks and their guide points, environmental information, trajectory information, load information and actual motion trajectory information.
9. The decision planning method according to claim 1, wherein: The posture state information includes at least one of the center position of the excavator chassis and the position of the bucket tooth tip; The motion state information includes at least one of the rotation speed of the left and right travel motors of the excavator chassis, the rotation angle of the upper vehicle slewing device, the boom lifting angle, the angle between the boom and the excavator arm, and the angle between the bucket and the boom; and / or The load status information includes at least one of the pressure difference between the oil inlet and outlet ports of the left and right travel motors of the excavator chassis, the pressure difference of the upper vehicle rotary motor, the force of the boom cylinder, the force of the dipper cylinder, and the force of the bucket cylinder.
10. The decision planning method according to claim 1, wherein: The target task guidance information includes at least one of the target task category and the location coordinates; and / or The trajectory information at the future time includes at least one of the position state information and the motion state information of the excavator at the future time.
11. A decision-making planning device, comprising: Memory; And a processor coupled to the memory, wherein the processor is configured to execute the decision planning method according to any one of claims 1 to 10 based on instructions stored in the memory.
12. A decision-making planning device, comprising: A module for executing the decision planning method according to any one of claims 1 to 10.
13. An excavator, comprising: A decision planning device, configured to execute the decision planning method according to any one of claims 1-10.
14. A computer-readable storage medium having computer instructions stored thereon, which, when executed by a processor, implement the decision-making planning method according to any one of claims 1 to 10.
15. A computer program product comprising computer instructions, characterized in that When the computer instructions are executed by a processor, the decision-making planning method described in any one of claims 1 to 10 is implemented.