System and method for providing train rail maintenance and inspection

Abstract

A railroad maintenance and inspection system (RMIS) streamlines railway upkeep through a decentralized network of portable, autonomous maintenance and / or inspection vehicles (AMIVs) configured to accept removable modular maintenance and / or inspection modules (MIMs) that accommodate a variety of accessory attachments, including rail brushes, scrubbing wheels, sprayers, and specialized sensors. The decentralized system preferably utilizes artificial intelligence (AI) to partition the rail lines into rail segments and allocate AMIVs from nearby storage sites to perform specific maintenance and / or inspection (MI) tasks on assigned rail segments based on track and weather conditions, and any scheduled maintenance and / or inspection requirements. AMIVs are equipped with optical sensor(s) and wireless communication systems, which provide real-time updates on MI activities to both train crew and a maintenance and inspection computer server (MICS), which also has access to multiple database servers, including rail and weather conditions. The RMIS minimizes downtime and total cost for performing optimal MI tasks, including a schedule of assigned MI tasks, and generating analytics associated with operation of AMIVs.

Description

Attorney Docket No. 36409-100110SYSTEM AND METHOD FOR PROVIDING TRAIN RAIL MAINTENANCE AND INSPECTIONCROSS-REFERENCE TO RELATED APPLICATIONS

[0001] This application claims the benefit of and priority to U. S. Provisional Application No. 63 / 730,454, filed December 11, 2024, entitled " Autonomous Electric Vehicle for Railroad Maintenance and Inspection, " the disclosure of which is hereby incorporated by reference in its entirety.TECHNICAL FIELD

[0002] The present disclosure relates to a system and method for providing train rail maintenance and inspection, and, more particularly, to a system and method for addressing precipitation and contaminants on train rails, such as pectin, black precipitate, snow, and ice, as well as for inspecting track conditions for flaws.BACKGROUND

[0003] During the fall season, trains traveling over leaves and other organic debris on the rails generate heat and friction, crushing these materials into a black precipitate that contains pectin, a thin, greasy film. This slippery residue significantly reduces friction between the train wheels and the rails, causing wheel slippage and sliding. In severe cases, onboard computers may trigger automatic emergency stops. Excessive sliding and braking can also cause flat spots on train wheels, requiring costly repairs and removing trains from service. This problem is particularly prevalent on open commuter rail networks connecting urban centers to outlying towns, often resulting in substantial service delays and / or cancellations.

[0004] In some instances, slippery tracks have even been identified as contributing factors in train collisions and derailments. In other cases, the buildup of leaves can reduce conductivity and electrically insulate the wheels from the rails, preventing shunting and causing the signaling equipment to fail to detect the train's presence.

[0005] Slippery rail conditions are not limited to fall foliage — they can occur year-round due to varying weather conditions. In winter, snow and ice reduce track adhesion, making itAttorney Docket No. 36409-100110difficult for trains to start and stop safely. Even in summer, light rain can mix with oxide debris on the tracks, forming a slick paste that similarly compromises traction.

[0006] Over the years, several solutions have been developed to address these issues. For example, large rail trains with mechanical wire brush heads scrape compressed leaves, snow, ice, and debris from the tracks. Others use high-pressure water jets to blast away buildup, often followed by applying high-friction coatings, such as sand paste or specialized gels, to improve wheel -to-rail adhesion. Still, others have explored advanced technologies, such as plasma jets and lasers, with plasma jets breaking down the slippery residue and lasers vaporizing it entirely.

[0007] While these methods have shown effectiveness in limited and localized applications, they are generally ill-suited for year-round, large-scale implementation across extensive rail networks to ensure safe and reliable commuter train rails. Their high operational costs, labor-intensive maintenance, and potential to disrupt commuter train schedules pose significant barriers. Additionally, these systems lack the robustness needed to ensure full maintenance of all rail lines.

[0008] The safety and reliability of train rail networks also require ongoing rail infrastructure inspection. A critical aspect of rail maintenance is the continuous inspection and monitoring of track conditions for flaws in the rails and deviations in track geometry, such as cracks, wear, misalignment, and improper gauge width, among other things, that can also lead to accidents, derailments, and costly service disruptions.

[0009] Advanced rail inspection systems play a crucial role in detecting and addressing these issues before they become hazardous. Technologies such as ultrasonic testing, groundpenetrating radar, and laser-based geometry measurements help identify internal rail defects, corrosion, and track misalignments that are not visible to the naked eye. However, with today's technologies, operating rail inspection systems regularly for predictive maintenance is timeconsuming and costly.

[0010] Accordingly, a cost-effective and robust system and method are needed that ensure rail safety and minimize downtime and costs.BRIEF SUMMARY

[0011] The present disclosure relates, among other things, to a railroad maintenance and inspection system (RMIS) that streamlines railway upkeep through a decentralized network ofAttorney Docket No. 36409-100110portable, autonomous maintenance and / or inspection vehicles (AMIVs) to minimize downtime and cost. This decentralized approach, preferably utilizing artificial intelligence (Al), affords train lines to be partitioned into rail segments and AMIVs allocated from nearby storage sites to perform specific maintenance and / or inspection (MI) tasks on assigned rail segments based on track and weather conditions, and any previously scheduled maintenance and / or inspection requirements. Rail segments may also be prioritized based on track conditions, past, current, and forecast weather conditions, rail contaminants, past maintenance and cleaning records, environmental data, and track usage frequency.

[0012] In an exemplary embodiment, the RMIS provides wireless communication among AMIVs, a maintenance and inspection computer server (MICS), and / or train crew mobile devices, including smartphones or smart tablets. Additionally, the MICS may communicate with database servers containing a variety of track condition data and topography, as well as maintenance and / or inspection data, such as rail lines and stations, trail schedules, AMIV status, sensor data relating to the surface conditions of the tracks and the identification of contaminants, debris, corrosion and unevenness, historical data on past maintenance and cleaning records, current and forecast weather conditions, environmental data on local vegetation growth, and track usage frequency, and / or other information generated and stored by the RMIS.

[0013] The MICS is preferably implemented with artificial intelligence and machine learning (AI / ML) functionality provided by an Al module or alternatively provided from cloud computing platforms such as AWS, Google Cloud, or Microsoft Azure. The MICS allocates and schedules MI tasks efficiently, while also generating analytics and reports to enhance overall system performance. Embedded Al / ML optimization algorithms may process incoming data and make real-time adjustments, improving task execution while continuously refining future maintenance and inspection strategies to minimize labor costs and downtime. Cost and downtime minimization follows a structured workflow, gathering first real-time information on rail conditions, train schedules, and environmental factors that are then processed by the Al module to compute optimal MI tasks, rail segments, and AMIV assignments. The scheduling may be adjusted based on real-time updates, particularly for immediate repairs and maintenance tasks.

[0014] In another aspect, the portable AMIVs are strategically stationed at nearby storage sites located judiciously throughout the train rail network. This setup enables train crew to quickly deploy AMIVs to specific rail segments, including those in non-contiguous areas, and preferablyAttorney Docket No. 36409-100110deployed during gaps in train operation for minimal disruption to active rail traffic. AMIVs are relatively mobile and lightweight, making for fast and safe on-and-off-track deployment.

[0015] In a further aspect, the AMIV includes a control processor module, a wireless communication module, an antenna, an audio module, a drive module coupled to a power module, and a maintenance or inspection module (MIM). Each AMIV is a self-powered vehicle, preferably configured to receive instructions for autonomous operation for completing designated MI tasks. Furthermore, the MIMs are preferably removably mounted and modular, allowing them to engage and disengage readily with the frame of the AMIVs. MIMs may include specialized accessory attachments, such as brushes, scrapers, or wires, configured to clean a particular type of contamination or precipitation off the tracks, such as ice, snow, or pectin. Alternatively, the MIMs may include specialized sensors to inspect and assess rail conditions, such sensors including cameras, lasers, and other diagnostic sensors. Still further, MIMs may address vegetation growth on the track, and grinding tasks, or may even be tailored to apply sand or a high-friction coating, such as a specialized gel, to enhance wheel-to-rail adhesion. AMIVs respond to queries from maintenance or the train crew through mobile web-based applications running on a mobile device, such queries including real-time maintenance statuses and / or reports, among other things. Authorized train crew can also manually override, if necessary, AMIV tasks and adjust operations, ensuring efficient and responsive maintenance and / or inspection management.

[0016] In yet another aspect, communication is established between the AMIVs and the MICS through wireless communication links, which preferably access the Internet through a wireless network, either through cellular, Wi-Fi, or a satellite connection for regions without service. Wireless communication enables real-time control, telemetry, and diagnostic functionality. Authorized train crew may attend the MICS through a computer terminal or workstation. Preferably, the train crew utilizes mobile devices that include web-based applications to access some or all of the MICS functionality via the wireless network.

[0017] In another aspect, the AMIVs may also include front and back diagnostic sensors, respectively, such as HD digital cameras, ultrasound sensors, or lasers that inspect and monitor the surrounding track environment. The number and positioning of the diagnostic sensors may be judiciously tailored to the specific needs of a given application, depending on the track environment and maintenance objectives. The front and back diagnostic sensors may use Al image analysis or LiDAR to aid navigation and inspection.Attorney Docket No. 36409-100110

[0018] In a further aspect, each AMIV is preferably configured to accept a removable modular MIM affixed preferably on either side of its frame and into some or all of a framed enclosure region beneath the outer body shell. The MIM is preferably constructed of a durable, modular framework enclosure designed to accommodate a variety of accessory attachments, such as rail brushes, scrubbing wheels, sprayers, and specialized sensors. A modular framework enclosure provides a protective housing while allowing flexibility in configuration. Some MIMs are designed to handle different types of precipitation and track contaminants, such as pectin, black precipitate, snow, or ice. Others are designed with specialized sensors for track inspection. Still, others have specialized accessories for controlling vegetation growth on the tracks, performing rail grinding operations, or improving wheel-to-rail adhesion by applying sand or a high-friction coating, such as a specialized gel.

[0019] In yet another aspect, the RMIS includes a plurality of AMIVs with their identification and location having been registered therewith, which AMIVs receive optimal MI tasks from the MICS. The AMIVs operate substantially autonomously on the rails once provided with optimal rail segments and their MI tasks to minimize train downtime and cost. AMIVs are preferably configured to capture and / or generate sensor data relating to environmental conditions, and the performance of the AMIVs. Environmental conditions may include weather, track containment conditions, track usage patterns, seasonal effects, and observed obstacles, including other vehicles, persons, or objects. Performance data may include battery levels, maintenance logs, vehicle usage, mileage, and the intensity of cleaning. The MICS may transmit any data to the AMIVs, including optimal assigned rail segments for maintenance and / or inspection, task instructions, and any modified rail segments based on environmental data received from other AMIVs.

[0020] In a further aspect, the database servers may include a weather database server associated with and / or maintained by third-party weather services, such as the National Weather Service (NWS), that provide current weather conditions and forecasts affecting the geographical region where the rail network is located. The database servers may also include a rail database server maintained by the train rail network. This database server provides rail line topography overlaid with a geographical map of the region, providing the MICS with information and data regarding the paths of the different train lines, including their location, stops, train schedules, and rider traffic. The MICS may also have access to database servers of the Federal RailroadAttorney Docket No. 36409-100110Administration (FRA) containing topography maps with geospatial resources for identifying stations and tracks for rail networks in the United States.

[0021] In yet another aspect, the Al module creates an efficient strategy for assigning AMIVs and their assigned rail segments for MI tasks. AMIV identification information and their associated storage sites, as well as sensor data from the database servers, are used to partition train lines, allocate AMIVs, and assign their MI tasks. These tasks and rail segment assignments are developed with routing strategies to optimize the use of the AMIVs and lower operating costs and downtime. In some embodiments, the Al module may develop strategies for maximizing maintenance system performance (e.g., cleaning) without prioritizing cost, or for minimizing cost while minimizing downtime. The Al module may also create optimal MI tasks based on a cost analysis for each task, taking into account labor costs and commuter train schedules. In partitioning the train lines and assigning MI tasks, the Al module may identify and prioritize train lines for maintenance and inspection while deprioritizing others with lower ridership. As such, the Al module may allocate more AMIVs to specific train lines than others and transmit corresponding instructions to the AMIVs. The Al module may identify rail segments for contaminant inspection rather than maintenance in partitioning the train lines. Alternatively, the Al module may instruct AMIVs to travel along designated tracks to map out obstructions, anomalies, potential hazards, rail flaws, or track geometry.

[0022] In a still further aspect, the rail segments may also be prioritized based on track conditions, past, current and forecast weather conditions, rail contaminants, past maintenance and cleaning records, environmental data on local vegetation growth, and track usage frequency. The Al module may be further configured to identify maintenance patterns or trends as historical data is collected with continued usage and stored in the database servers to augment prioritization. In one embodiment, the Al module identifies maintenance, inspection, and / or repair patterns, along with their corresponding rail locations. This allows the Al module to develop maintenance and / or inspection strategies to preemptively instruct AMIVs to perform MI tasks for assigned rail segments, and preferably schedule them during off-peak train hours, such as late at night or early in the morning, before heavy ridership occurs.

[0023] In another aspect, the MICS may be linked and / or communicate with automatic train protection (ATP) and automatic train control (ATC) systems of many of today's modem train systems, where normal signaling operations, such as route setting and train regulation, are carriedAttorney Docket No. 36409-100110out by the ATP and / or the ATC system. These later systems work together to maintain a train's safe distance apart by marginally adjusting moving and station dwell time to ensure safe spacing between trains. These systems may be a source of external data, such as train position, current speed, track conditions (including speed limits, gradients, and switch positions), signal information, upcoming station stops, timetable data, and any relevant trackside information transmitted via trackside beacons (balises), which the train receives and uses to automatically adjust its speed and movement according to safety regulations and operational schedules.

[0024] In a further aspect, a method for partitioning the train lines into rail segments and allocating AMIVs to partitioned rail segments for specific inspection and maintenance tasks may be provided to reduce downtime and costs. The method is executed by the MICS, utilizing AI / ML to improve the efficiency of MI tasks while optimizing their execution. The MICS is responsible for executing the Al and ML functionality using the Al module for various tasks, including routing, analytics, reporting, and system management. These operations may involve specialized instructions that require a central processor to work in conjunction with additional coprocessors. The method may include retrieving from the database servers rail network information, including the number of lines and branches, the number of stations, the mileage of rail for each line, train schedules, the geolocations or geographic coordinates of the track lines, and any other available data relevant to the commuter service of the train rail network. The MICS also accesses database server to retrieve RMIS information, including the identification and storage locations of the AMIVs, the operational status and capability of each AMIV, the personnel costs associated with operating each AMIV per mile of track, and historical data related to previous inspection and maintenance activities, available Ml tasks, the track conditions, current and forecasted weather, task priorities, track assignment records, optimization algorithms, and training datasets. Based on the rail network and RMIS information, the train lines are partitioned into rail segments. Specified MI tasks are also assigned for each rail segment based on track and weather conditions, including the start and estimated completion times of the tasks. The database servers contain specific details regarding the type of MI tasks required, which may depend on factors such as the type of contaminants, debris, or precipitation, as well as the priority levels of the specific tasks. The method further includes allocating the AMIVs to rail segments, preferably based on the proximity of their corresponding storage sites to the rail segments, ensuring that each AMIV is assigned to the nearest geographical rail segment. The MICS considers the train schedules of the trail railAttorney Docket No. 36409-100110network and limits the time AMIVs may operate, scheduling MI tasks preferably within train operation gaps to prevent service disruptions. The MICS applies Al and ML techniques to analyze rail network information, RMIS information, and task definitions to generate an assignment of the AMIVs, the rail segments, and MI tasks and schedules. The method considers the number of AMIVs required and their designated MI tasks while minimizing downtime and cost. Rail segments with the worst track conditions and / or high passenger traffic are assigned a higher priority, ensuring maintenance in those areas is completed before lower-priority segments. Rail segments may also be prioritized based on track conditions, current and forecast weather conditions, rail contaminants, past maintenance and cleaning records, environmental data, and track usage frequency. Al models utilize historical task performance data to refine assignments, leveraging reinforcement learning and deep neural networks to enhance future scheduling.

[0025] In another aspect, once the rail segments and AMIV assignments are determined, the MICS transmits the assigned rail segments and tasks to the allocated AMIVs and train crew. Each AMIV receives its assigned rail segment, the specific MI task to be performed, the scheduled time frame for execution, and any operational constraints based on train schedules. The train crew quickly deploys the AMIVs from nearby storage sites to their assigned rail segments. The AMIVs are fitted with an appropriate MIM for the assigned MI tasks.

[0026] In an aspect, as the AMIVs traverse along their assigned rail segments, real-time sensor data from the AMIVs and trackside sensors are monitored to update track conditions, weather hazards, and task completion status. If unexpected track obstructions, severe weather conditions, or maintenance failures are identified, an updated maintenance and inspection schedule is generated, and AMIVs may be added as needed, along with any updated Ml tasks, rail assignments, and scheduling times, which are transmitted to the AMIVs. If it is determined that the AMIVs have not yet completed their assigned MI tasks, tasks may be further updated until all tasks are completed. However, once the assigned MI tasks have been completed, the MICS alerts the train crew to load AMIVs back onto pick-up trucks, return them to their designated storage sites, and recharge them. The MICS then receives updated status on the MI tasks completed, as well as any updated RMIS and rail network information, including rail conditions, maintenance and / or inspection status, and AMIV status.

[0027] In yet another aspect, the method establishes an Al adaptive maintenance strategy that optimizes AMIV task assignments using real-time and historical data. By leveraging machineAttorney Docket No. 36409-100110learning for predictive maintenance scheduling, the method effectively reduces downtime, lowers costs, and enhances the reliability of rail infrastructure while dynamically adapting to changing operational conditions. The exemplary method above includes various processing steps, and these executable instructions may be embodied in computer-executable instructions and stored on any computer-readable medium, such as solid-state memories, optical media, and magnetic media.

[0028] In still another aspect, the optimization may be mathematically modeled as the sum of individual AMIV costs and task execution times, weighted by a balancing factor. The objective is to minimize the combined cost of task execution and total task completion time, ensuring that priority tasks are performed. The optimization of the mathematical model is subject to, but not limited to, the following constraints: task assignment - each task must be assigned to at most one AMIV in a given rail segment; AMIV availability - each AMIV can only be assigned one task at a time; total available AMIV - the total number of AMIV assigned at any time cannot exceed the fleet size; time constraints -tasks must be scheduled within available operational hours; priority consideration - higher priority tasks must be scheduled before lower priority ones; vehicle travel time - an AMIV must have sufficient time to travel to a task location before starting the task; emergency handling - emergency repairs must take precedence over all other tasks; task completion - each task has a predefined duration; off-peak scheduling - tasks must be scheduled when trains are not running; and high-priority rail segments - tasks in high-priority rail segments must be assigned first.

[0029] In an aspect, the Al-driven optimization system for the present RMIS requires realtime data from multiple sources, including weather forecasts, to prioritize the removal of snow, ice, or leaves and ensure that hazardous conditions are addressed. AMIV sensors can detect track flaws, and vegetation overgrowth, which are essential for scheduling inspections and maintenance tasks. Train schedules determine available time slots for maintenance, allowing tasks to be executed during off-peak hours. Ridership volume also provides data for priority scheduling, and GPS helps generate estimated task durations. Historical data provides valuable insights into past performance, enabling Al models to make more accurate predictions and enhance decisionmaking.

[0030] In yet another aspect, to ensure that the objective function does not simply avoid assigning MI tasks to minimize total cost, additional constraints or penalties may be imposed. InAttorney Docket No. 36409-100110some embodiments, an added constraint may be to require that every task of high priority is assigned to an AMIV, ensuring that all high priority tasks are performed.

[0031] In still another aspect, the objective function may include a penalty for any high priority tasks not assigned. Alternatively, the objective function may include a term that rewards assigning a task, incorporating priority levels. Alternatively, the model may set a usage rate for the AMIVs so that each AMIV performs at least one task within the available hours. Still further, the mathematical model may ensure that higher-priority MI tasks are maximized and completed. The objective function and constraints may be modified accordingly. Instead of only minimizing cost and task completion time, the model introduces a priority-weighted term to ensure that higher-priority tasks are assigned. The prioritization process may begin by defining the specific factors relevant to a particular Ml task. Such factors may include current rail conditions such as contamination from ice, snow, pectin, rail wear, corrosion, or geometric irregularities. For example, if the MI task involves snow or ice removal, the relevant factors might include the current thickness of ice accumulation, forecasted snowfall, historical frequency of ice-related issues, ridership levels, and any emergency indicators that signal urgent maintenance needs. For other tasks, such as contamination removal, including pectin, relevant factors might instead include the current thickness of pectin. For vegetation control or rail grinding, different factors, like vegetation growth rates, rail wear severity, historical inspection records, or the time since the last maintenance would be considered. Also, real-time and forecasted weather conditions may be gathered, focusing on precipitation levels, temperature extremes, seasonal conditions like leaf fall, and other potential adverse weather. Such information may be readily retrieved from the database servers, which have access to other real-time third-party databases, including the National Weather Service (NWS).

[0032] After identifying these relevant task-specific factors, each factor may be assigned a numerical score reflecting its severity or urgency, along with a corresponding weighting factor representing its relative importance to the overall prioritization of the task. These weighted factors are combined into a single numerical value known as the task-specific priority score. Moreover, for each rail segment, the priority score for the given task is calculated by multiplying each factor's measured severity by its assigned weight, then summing these products. Rail segments with conditions indicating immediate attention, such as a combination of severe weather impacts, high contamination, high historical maintenance frequency, or emergency signals - will naturally produce higher scores. Once calculated, these scores allow rail segments to be ranked, highlightingAttorney Docket No. 36409-100110which rail segments require immediate action for a specified MI task. Rail segments with the highest scores are assigned the highest priority, ensuring that critical tasks like urgent ice, snow or pectin removal, vegetation management, or track repairs are addressed promptly. This prioritization can be integrated into maintenance and / or inspection scheduling algorithms, allowing resources to be allocated efficiently and dynamically. And, by continuously recalculating priority scores as new data becomes available, this method ensures the MI scheduling remains responsive to changing conditions.

[0033] In another aspect, prioritization may use ML models to predict the urgency of MI tasks, or the likelihood of rail defects based on historical track conditions, weather patterns, traffic intensity, and previous maintenance outcomes. Such a model would continuously learn and adjust the priority scoring accordingly. This prioritization may be integrated directly into assigning rail segments with specific MI tasks. MI tasks on rail segments with scores surpassing a critical threshold are marked as mandatory, guaranteeing their completion. Regular updates ensure that the priority scores accurately reflect current conditions, enabling timely responses to emerging maintenance and inspection needs.

[0034] In still another aspect, each AMIV is preferably assigned to a rail segment based on the proximity of its storage site to the rail segment. Geographical data is compiled on the locations of the rail segments and the storage sites for the AMIVs. For each rail segment requiring an MI task, the distances from every storage site to the midpoint of that rail segment are calculated. Storage sites are then ranked according to their proximity to the rail segment. Assignments are first attempted by selecting AMIV from the storage site closest to the rail segment. If an AMIV is unavailable from that storage site due to other assignments or operational constraints, the next closest storage site is chosen, and so on. This iterative approach ensures that each rail segment is serviced by an AMIV from the nearest feasible storage site.

[0035] In another aspect, to determine the number of rail segments when partitioning train lines, a simplistic approach may be used, dividing the total track length by the desired segment length, with the constraint that the number of rail segments does not exceed the number of available AMIVs. Additionally, natural and operational constraints are considered in partitioning the train lines, such as breaking segments at access points where AMIVs may be readily loaded and unloaded, or breaking segments to areas historically requiring frequent maintenance. Typically, MI tasks are performed during gaps in train operation, so partitioning would be based on theAttorney Docket No. 36409-100110available maintenance time window so that rail segments could be serviced within that time. With the AMIVs having a defined operational range, rail segment lengths should be small enough that an AMIV could complete its MI task within its operational time and range.

[0036] In another aspect, the user interface of the RMIS may display key functionalities of the system, including the overall status, the current locations of all active AMIVs overlaid with an exemplary train rail network, maintenance and inspection schedules, and various analytics and reports.BRIEF DESCRIPTION OF THE DRAWINGS

[0037] The features and advantages of the present disclosure will become more readily apparent from the following detailed description of the embodiments in which like elements in the figures are labeled similarly. The figures herein of the autonomous maintenance and / or inspection vehicle (AMIV) may, in some instances, only depict one side. In such case, the other side is a mirror image and is not shown or described herein for clarity.

[0038] Fig. 1 is a high-level pictorial representation of a railway maintenance and inspection system (RMIS) in accordance with the present disclosure, including a maintenance and inspection computer server (MICS), and a fleet of portable autonomous maintenance and / or inspection vehicles (AMIVs) decentralized along storage sites of an exemplary train rail network;

[0039] Fig. 2 is a view of an exemplary AMIV;

[0040] Fig. 3 is a functional block diagram of an exemplary AMIV in the RMIS of Fig.1;

[0041] Fig. 4 illustrates a view of the outer body of an exemplary AMIV;

[0042] Fig. 5 illustrates a perspective view of the frame of an exemplary AMIV;

[0043] Fig. 6 illustrates a perspective view of an exemplary AMIV adapting the frame of Fig. 5;

[0044] Fig. 7 illustrates a view of an exemplary maintenance and inspection module (MIM), including scrub wheels for removing ice, snow, or pectin;

[0045] Fig. 8 illustrates a view of an exemplary MIM, including a rail brush having a set of spring steel tines for removing ice, snow, or pectin;

[0046] Fig. 9 illustrates a view of an exemplary MIM, including a rail grinder with rotating stones for correcting the rail profile;Attorney Docket No. 36409-100110

[0047] Fig. 10 illustrates a view of an exemplary MIM, including a sprayer and tank for spraying herbicides to help control invasive vegetation growth, or spraying traction gel on the head of the rails;

[0048] Fig. 11 is a view of an exemplary AMIV having a MIM with specialized sensors to measure track geometry and / or to inspect for flaws on train rails;

[0049] Fig. 12 illustrates a flow chart of an exemplary method for (i) partitioning the rail lines into rail segments (ii) assigning AMIVs from nearby storage sites, and (iii) generating maintenance and inspection tasks for each rail segment;

[0050] Fig. 13 depicts an exemplary screenshot of the " Homepage" displayed on a user interface (UI) of the RMIS; and

[0051] Fig. 14 depicts an exemplary screenshot of another page of a user interface (UI) displayed when the AMIV link on the " Homepage" of Fig. 13 is selected.DETAILED DESCRIPTION

[0052] The innovative railroad maintenance and inspection system (RMIS) of the present disclosure streamlines railway upkeep through a decentralized network of portable, autonomous maintenance and / or inspection vehicles (AMIVs) to minimize downtime and cost. This decentralized approach, preferably utilizing artificial intelligence (Al), affords the train lines to be partitioned into rail segments and AMIVs allocated from nearby storage sites to perform specific maintenance and / or inspection (MI) tasks on assigned rail segments based on track and weather conditions, and any previously scheduled maintenance and / or inspection requirements. Rail segments may also be prioritized based on track conditions, past, current and forecast weather conditions, rail contaminants, past maintenance and cleaning records, environmental data, and track usage frequency.

[0053] Extreme cold, snow, and ice can impact train lines that run at street level or above ground more than others, requiring snow and ice maintenance service on those corresponding rail segments. Other train lines may only need routine inspection for debris or flaws. And still others may need rail maintenance to attend to the pectin or leaves due to fall foliage. AMIVs are equipped with diagnostic sensors, for example, optical sensor(s) and wireless communication systems, which can provide real-time updates on MI tasks to both the train crew and a maintenance andAttorney Docket No. 36409-100110inspection computer server (MICS), which also has access to multiple database servers storing track and weather conditions.

[0054] As used herein, “autonomous maintenance and / or inspection vehicles” (AMIVs) refer to vehicles designed to perform assigned maintenance and inspection tasks with minimal or no human intervention. Although AMIVs are primarily intended to operate independently, they may also be manually operated by a human, either with or without assistance from automated systems such as obstacle detection, computer vision, or navigation technologies. Manual control may be necessary in specific situations, including but not limited to loading or unloading the AMIVs onto tracks, responding to emergency conditions, or complying with applicable local, state, or federal regulations.

[0055] Decentralized AMIVs operating concurrently across the train rail network reduce downtime by servicing small rail segments simultaneously, preferably during gaps in train operations. The number of rail segments 110 is determined by the size of train rail network 105, the number of storage sites 115, the number of available AMIVs 120, and the anticipated or needed MI tasks, primarily based on track and weather conditions.

[0056] Referring to Fig. 1, RMIS 100 provides wireless communication among AMIVs 120, MICS 125 and / or train crew mobile devices 130, including smartphones or smart tablets. Also, MICS 125 may communicate with database servers 135, including the database servers from train rail network 105, and other third-party database servers. MICS 125 is preferably implemented with artificial intelligence and machine learning (AI / ML) functionality provided by an Al module 140. MICS 125 may also include computing devices, such as a workstation 175, with internet access that alternatively provides the AI / ML functionality designed to operate on cloud computing platforms such as AWS, Google Cloud, or Microsoft Azure. Such an Al cloud platform provides scalable infrastructure for processing large-scale optimization problems, while also allowing seamless integration of real-time data and computationally intensive optimization algorithms. In other words, the AI / ML functionality may be cloud-based, but would involve integrating RMIS 100 with cloud Al services through APIs, SDKs, or specialized frameworks. A cloud-based RMIS also allows access to other Al-powered tools, natural language processing (e.g., OpenAI), and computer vision capabilities.

[0057] MICS 125 includes a main memory 145 and a central processor 150 that may communicate with one or more database servers 135, which contain a variety of track conditionAttorney Docket No. 36409-100110data and topography, as well as maintenance and / or inspection data. For example, database servers 135 may include information, such as rail lines and stations, trail schedules, AMIV status, sensor data relating to the surface conditions of the tracks and the identification of contaminants, debris, corrosion and unevenness, historical data on past maintenance and cleaning records, current and forecast weather conditions, environmental data on local vegetation growth, and track usage frequency, and / or other information generated and stored by RMIS 100.

[0058] In certain embodiments, the MICS 125 may employ an AI / ML module 140 trained on a range of information associated with the operation of the rail network 105. This training data may include earlier maintenance and inspection records describing the tasks previously performed on each rail segment 110, together with the conditions that led to those tasks; sensor data gathered during prior AMIV 120 deployments, such as images, contamination assessments, surface irregularity measurements, and track-geometry readings; and operational information, including train schedules, passenger-traffic levels, station and track layouts, and the availability of maintenance windows. Weather information — both historical and forecasted — may also form part of the training data, as may patterns associated with snowfall, temperature variation, leaf deposition, and other environmental effects. In addition to this historical material, the MICS 125 may generate and use simulated track-condition scenarios derived from physics-based or generative models incorporating the geometry of the rail network 105 and expected train movements. These simulations allow the system to develop representative patterns of rail wear or contamination and to explore potential maintenance routes even when comparable historical examples are sparse.

[0059] Using these materials, the Al / ML module 140 may learn how rail conditions evolve under different combinations of weather, usage, and maintenance history, and how conflicts in scheduling typically arise when maintenance must be coordinated with train operations. During normal operation, the module may receive as input the real-time sensor information transmitted from active AMIVs 120, updated weather data and forecasts, current train-operation information, and the status of the AMIV 120 fleet, including each vehicle’s location, battery level, and installed maintenance or inspection module. Information describing the rail network 105 itself, including station positions, track topology, and known access points, may also be provided as input. With these inputs, together with the relationships learned from the training data, the AI / ML module 140 may determine the likely maintenance and inspection needs of each rail segment 110, identify theAttorney Docket No. 36409-100110type of work required, and estimate the urgency of those tasks. The AI / ML module 140 may also identify which AMIVs 120 are best suited to perform the required tasks based on their proximity, readiness, and travel time to the relevant segment 110, and may further generate a recommended schedule for performing the work within available gaps in train traffic. In addition, the AI / ML module 140 may generate revised outputs whenever real-time conditions change — for example, if severe weather develops, if a new obstruction is detected on the track, or if train operations deviate from their expected schedule. In such cases, the system may adjust its earlier predictions, reassign tasks, or alter the proposed sequence of operations. Through repeated use of this combined predictive and scheduling approach, the MICS 125 may improve its ability to anticipate future maintenance needs, reduce conflicts with train movements, and produce maintenance and inspection schedules that reduce downtime while maintaining safe and reliable rail operations.

[0060] The AI / ML module 140 may utilize one or more models to facilitate efficient and optimal management of the rail network 105. For example, a condition model may be employed to evaluate the condition of the tracks, receiving inputs that may include data from various sensors or manual entry by a user. The condition model would output track metrics (for example, track geometry, track profile, identified surface cracks, internal defects, rotted ties, damaged hardware, water content in ballast, etc.) and would detect and / or identify obstacles on the tracks. The output of the condition model may be used by a human user to evaluate the condition of the tracks, or by a robot in a closed-loop feedback situation. Additionally, the AI / ML module 140 may utilize an assignment model to partition the tracks into rail segments and to assign MI tasks. This assignment model receives inputs such as track metrics from the condition model, maintenance history, track usage data, weather information, and additional sensor data. The assignment model may output track partitions (segments 110) and specific MI tasks. This output can be used, for example, by a human user to deploy robots (AMIVs 120) for specific tasks, or it may enable robots (AMIVs 120) to deploy autonomously without human intervention. The conditional model and the assignment model may be implemented using various Al architectures, including but not limited to transformer networks, neural networks (NN), and convolutional neural networks (CNN). Moreover, both models may be employed together during a training phase to achieve supervised learning, and then transition to unsupervised learning after a sufficient volume of data has been collected.

[0061] MICS 125 executes programming instructions 160 using main memory 145 and includes Al module 140 having AI / ML functionality to allocate and schedule MI tasks efficiently,Attorney Docket No. 36409-100110as well as generate analytics and reports to enhance overall system performance. Embedded AI / ML optimization algorithms may process incoming data and make real-time adjustments, improving task execution while continuously refining future maintenance and inspection strategies to minimize labor costs and downtime. However, it is important to note that ML models will be judiciously tailored to meet the specific maintenance needs of the train rail network 105. For example, ML may rely on convolutional neural networks (CNN) for visual analysis, recurrent neural networks (RNN), or long short-term memory (LSTM) models for temporal data and / or regression models.

[0062] Cost and downtime minimization - a type of optimization - follows a structured workflow, gathering first real-time information on rail conditions, train schedules, and environmental factors that are then processed by Al module 140. This minimization may utilize Mixed-Integer Linear Programming (MILP), Reinforcement Learning (RL), or heuristic-based methods to compute optimal MI tasks, rail segments, and AMIV assignments. The scheduling may be adjusted based on real-time updates, particularly for immediate repairs and maintenance tasks

[0063] It should be clearly understood that Al module 140 provides the AI / ML functionality for MICS 125 and should be able to handle large-scale optimization (e.g., minimization) and real-time processing, preferably running under a Linux or Windows Server operating system. In one embodiment, Al module 140 may include a multi-core central processing unit (CPU), memory, a graphical processing unit (GPU), a communication controller, and a storage controller. GPUs such as the NVIDIA Tesla A100, H800, H100, or RTX 4090 are particularly well-suited for Al applications. An Al accelerator application-specific integrated circuit, such as a Tensor Processing Unit (TPU) and / or a neural processing unit (NPU) specialized in handling computations in neural networks, may also be utilized to enhance Al operations, such as matrix multiplication, convolutions, and activation. Al module 140 is configured to handle and accelerate machine learning and the computational demands placed on them by RMIS 100 for the training and inference phases in machine learning, particularly on deep learning models. However, the artificial intelligence and machine learning functionality of Al module 140 may be implemented alternatively using cloud Al computing platforms, which could be more cost-effective.

[0064] Portable AMIVs 120 are strategically stationed at nearby storage sites 115, located judiciously throughout train rail network 105. As used herein, the term “portable” refers to an AMIV that is sufficiently small and lightweight to be readily transported and placed on or off theAttorney Docket No. 36409-100110rails using standard support equipment, such as pick-up trucks, small cranes, or hoists. In certain embodiments, a portable AMIV weighs less than about 800 pounds and has a compact structural form factor, for example approximately 66 inches (L) × 75 inches (W) × 25 inches (H), enabling safe lifting and transport. The AMIV may be constructed of framed steel or other suitable materials to provide durability while maintaining this transportability. The term “portable” pertains to offtrack handling and deployment and does not limit the AMIV’s autonomous, self-propelled operation on the rails. This configuration enables train crews to quickly deploy AMIVs 120 to specific rail segments 110, including those in n on-contiguous areas, preferably during gaps in train operations for minimal disruption to active rail traffic. After completing their MI tasks, AMIVs 120 may be lifted from the rails, loaded onto pick-up trucks, returned to their designated storage sites, and recharged.

[0065] As used in this context, a storage site refers to any accessible location along train rail network 105. This can include outdoor areas such as rail yards or compounds, storage garages and silos, as well as indoor facilities like underground yards, sheds, and equipment access sites. Rail segments 110 assigned to specific AMIVs 120 are preferably determined using ML models to minimize operational costs and reduce maintenance time.

[0066] Referring to Fig. 2, AMIV 120 includes a control processor module 210, a wireless communication module 215, an antenna 220, an audio module 225, a drive module 230 coupled to a power module 235, and a maintenance or inspection module (MIM) 205. MIMs 205 are preferably removably mounted and modular, allowing them to engage and disengage readily with the frame of the AMIVs. MIMs 205 may include specialized accessory attachments, such as brushes, scrapers, or wires, configured to clean a particular containment or precipitation off the tracks, such as ice, snow, or pectin. Alternatively, MIMs 205 may include specialized sensors to inspect and assess rail conditions, such sensors including cameras, lasers, and other diagnostic sensors. Still further, MIMs 205 may address vegetation growth on the track, and grinding tasks, or may even be tailored to apply sand or a high-friction coating, such as a specialized gel, to enhance wheel -to-rail adhesion. AMIVs 120 respond to queries from maintenance or train crew 155 through mobile web-based applications running on a mobile device 130, such queries including real-time maintenance statuses and / or reports, among other things. Authorized train crew can also manually override, if necessary, AMIV tasks and adjust operations, ensuring efficient and responsive maintenance and / or inspection management.Attorney Docket No. 36409-100110

[0067] Each AMIV 120 is a self-powered vehicle, preferably configured to receive instructions for autonomous operation for completing designated MI tasks. As such, each control processor module 210 includes a microcontroller 240 that controls the assigned MI tasks of the AMIV received from MICS 125 and controls communication to MICS 125 and other mobile devices 130. Microcontroller 240 may be an industrial single board computer (SBC) where all essential components like the CPU, RAM, GPU, and USB ports are integrated onto a single circuit board, and are readily available from Nvidia, Advantech, Gateworks, and Raspberry Pi, among others. If needed, an add-on Al board may be used to augment the SBC to run Al applications, reducing some of the workload of Al module 140. Such integration allows for real-time schedule adjustments based on changing conditions, enabling the scheduling model to refine task prioritization dynamically and account for unexpected rail obstructions or last-minute delays without relying on centralized processing.

[0068] Microcontroller 240 can be programmed and run web-based applications 245 to perform various tasks received from MICS 125, and may also control various external devices through its GPIO pins. Cellular, GPS / GNSS, and sensor modules may be readily connected to microcontroller 240, such as speedometers, cameras, altimeters, thermometers, metal detectors, range sensors, vibration sensors, and the like. Moreover, onboard software may periodically check sensor outputs for anomalies and recalibrate them automatically to maintain accuracy. Also, redundant sensors may be installed for critical functions, such as reporting the geocoordinates of the AMIVs and hazards along the tracks. Code and / or web-based applications 245 written in various languages (e.g., Python, Java, and C++) may be used to control the functionality of microcontroller 240 and its communication with mobile devices 130, and MICS 125.

[0069] In some embodiments, AMIVs may include a Global Navigation Satellite System (GNSS) module 280, including, for example, an integrated GPS board 285. Other suitable global navigation boards may include different navigation systems, such as Galileo, BeiDou Qzss, IRNSS, and GLONASS, which also provide global positioning and navigation.

[0070] In some embodiments, AMIV 120 may have a user interface (UI) 250 for train crew 155 to access some or all of its functionalities, and may include any number of devices, such as a keypad, touch display screen, and / or control switches UI 250 may be coupled to control processor module 210, allowing authorized train crew to activate or deactivate certain functions of the AMIV, as well as view diagnostics data, such as system status and alerts.Attorney Docket No. 36409-100110

[0071] Wireless communication module 215 includes a cellular unit 255 (e.g., a cellular, radio or satellite transmitter (Tx) and a receiver (Rx)) for communicating wirelessly with MICS 125 and other computer and / or mobile devices 130 to receive MI tasks, including the type of maintenance or inspection, the specific train line, or portion thereof it has been assigned.

[0072] Referring to Figs. 1-3, communication is provided between AMIVs 120 and MICS 125 through wireless communication links 165, which preferably access the Internet through a wireless network 170, either through cellular, Wi-Fi, or a satellite connection for regions without service. Wireless communication module 215 enables real-time control, telemetry, and diagnostic functionality. Authorized train crew 155 may attend MICS 125 through computer terminal or workstation 175.

[0073] In one embodiment, train crew 155 utilizes mobile devices 130 that include webbased applications to access some or all of MICS 125 functionality via wireless network 170. These mobile devices, such as smartphones, laptops, and tablets, monitor maintenance statuses in real-time and, if necessary, may reroute and / or reassign tracks to AMIVs based on data unavailable to MICS 125, ensuring efficient and responsive maintenance management.

[0074] RMIS 100 is envisioned to operate using wireless communication, preferably highspeed Wi-Fi over wireless network 170, such as Wi-Fi 5G. The current state-of-the-art protocol, however, is " Wi-Fi 6" (IEEE 802.1 lax) and is preferable to 5G since it offers significantly faster speeds than older protocols. Wi-Fi 6 features improved efficiency and reduced latency, making it ideal for real-time applications. However, these protocols are provided as examples and not as limitations. Alternative communication methods, including other Wi-Fi protocols, radio, microwave, LPWAN, private LTE and / or satellite, may also be employed with this system. Of course, redundant communication channels may be implemented to ensure continuous operation in case of signal loss.

[0075] AMIVs 120 may also include front and back diagnostic sensors 260-1, and 260-2, respectively, such as HD digital cameras, ultrasound sensors, or lasers that inspect and monitor the surrounding track environment. The number and positioning of the diagnostic sensors may be judiciously tailored to the specific needs of a given application, depending on the track environment and maintenance objectives. Front diagnostic sensor 260-1, and back diagnostic sensor 260-2 may use Al image analysis or LiDAR to aid navigation and inspection. Front diagnostic sensor 260-1, equipped with an infrared or thermal camera, or a laser, can pre-assessAttorney Docket No. 36409-100110the amount of a particular type of precipitation or contaminants on the rails to aid in scheduling MI tasks. In addition, a back diagnostic sensor 260-2, such as a camera or laser, can post-evaluate the level of cleanliness, for example, by Al image analysis and feature extraction, primarily based on the reflectance from the rails. Alternatively, diagnostic sensors 260-1 and / or 260-2 may use Raman spectroscopy to analyze the composition of the contaminants and the level of contamination. Those skilled in the art will readily note that such spectroscopy may be configured to scan optical wavelengths / frequencies of particular interest corresponding to components of the anticipated contaminants on the rails.

[0076] Some or all of these environmental images may be wirelessly transmitted to MICS 125 to aid and train AI / ML for later allocating MI tasks to AMIVs 120. Audio transmitted through audio module 225 may readily alert surrounding personnel or pedestrians of ongoing maintenance tasks or hazards on the track, along with corresponding warning lights 265, if needed. Audio module 225 may include an audio codec, amplifier, and potentially a microcontroller or DSP for signal processing, and can play audio through a speaker.

[0077] Drive module 230 propels the AMIV along its assigned rail segment 110. Drive module 230 may include any motor or engine 270 (gas or electric) capable of rotating the wheels to propel the AMIV to travel along its rail segment 110 at a desired speed. As implemented, a brushless electric motor is preferable, and in that latter embodiment drive module 230 may also include an electronic speed controller (ESC) 275 that controls the brushless motor's movement and speed by creating the appropriate current waveform to create a rotating magnetic field within the motor.

[0078] Control processor module 210 may include one or more software applications or hardware components configured for controlling or monitoring the operation of diagnostic sensors 260-1, 260-2, and GNSS module 280. Control processor module 210 includes hazard functionality that utilizes diagnostic sensors 260-1 and 260-2 (e.g., cameras and LiDAR) to detect obstructions, anomalies, and / or potential hazards on or near the tracks along its designated rail segment. These hazards may be mapped and their location recorded using GNSS module 280, which communicates with orbiting satellites to determine the hazards' geolocations or geocoordinates. AMIVs 120, in some embodiments, may autonomously stop or adjust its speed in response to detected hazards to prevent damage. AMIVs 120 may communicate with MICS 125 the geocoordinates of its location as well as those of the hazards over communication network 170.Attorney Docket No. 36409-100110

[0079] Referring to Fig. 4, a composite body shell 405 made of, for example, carbon fiber-reinforced plastic may be used to create the outer body of an AMIV 120. These materials are strong and lightweight, allowing the AMIVs to be readily mobile. In one embodiment, the shell may have a streamlined shape that reduces drag and uses less energy. This aerodynamic shape is somewhat teardrop-shaped, with a low profile, curved top, rounded front, and tapered rear, creating smooth airflow around the AMIV.

[0080] Figs. 5-6 illustrate the internal frame details of one embodiment of an AMIV 120, including a body having lower and upper rectangular frames 505 and 510, respectively, spaced vertically apart about 20". Each frame is made from, for example, a 2" box tube, 1 / 16" thick steel. Lower and upper frames 505 and 510, respectively, are vertically joined with similar 2" box tubings 515, for example by welding, yielding an overall dimension of about 66 (L) x 75 (W) x 25 (H) inches. Upper and lower frames 505 and 510 and interconnecting box tubings 515 form a framed enclosure region 520 for accepting and housing electronics and mechanical components, as described herein. A thin metal sheet covering the frame may be used as an outer body shell 518 of AMIV 120, with enclosure region 520 lying therein beneath. Alternatively, carbon fiber-reinforced plastic may be used, which is lightweight and strong. Front and back track wheels 525 and 530 are rotatably coupled to the underside of an outside perimeter frame 535 through front and back axles 540 and 545, respectively. Track wheels 525 and 530 are designed to slide axially along the axes, and they can be judiciously adjusted such that the distance between the wheels accommodates different track gauges. For example, the axes maybe spaced apart about 64.5 inches to accommodate the standard track width in the United States.

[0081] Train track wheels 525 and 530 are specially designed for use on railway tracks and act as a rolling component pressed fitted onto front axles 540, and 545, respectively, using taper bearings to minimize rolling friction. They may be made from high-strength steel alloys, such as carbon or low-alloy steel. The running surface of track wheels 525, and 530 is conical, which serves as the main means for keeping AMIV 120 aligned with the rails while in motion and improving stability. The conical shape creates a force that pulls the wheel back to the center of the track even when the train is turning, among other things. The inside of the conical wheel has a larger circumference than the outside. As such, when the AMIV rounds a curve, the outside wheel rides up to contact the rail at a larger diameter while the inside wheel drops down to contact the rail at a small diameter. This difference in distance traveled causes the wheels to follow the track.Attorney Docket No. 36409-100110

[0082] AMIV 120 may include drive wheels 550 that sit on and are in contact with the upper top surface of the rails to propel the AMIV along the rail. Drive wheels 550 may be made from natural or synthetic rubber, or some other petroleum-derived hydrocarbons, including synthetic materials. Power is preferably applied to drive wheels 550 instead of train wheels 525 and 530 to improve traction from the extra friction with the rails. However, in some embodiments, power may be applied directly to track wheels 525 and / or 530. An axle 555 is similarly rotatably coupled to drive wheels 550 along the underside of lower frame 505. Drive wheels 540 are press-fitted onto axle 555 using again taper bearings.

[0083] AMIV 120 may include one or more handrails 560 affixed to upper frame 510 surrounding its upper perimeter so that crew personnel can push, pull, or manually move AMIV 120. One or more lifting eyes 565 provide contact points for lifting AMIV 120 onto or from the train rails using a crane or other suitable machinery.

[0084] Each AMIV 120 includes drive module 230, which is used to propel the AMIV along its rail segment 110. Different drive mechanisms may be used to propel AMIVs 120. As mentioned, an electric brushless motor 270 is preferably used to power drive wheels 550 that sit on top of the rails. For example, sitting on lower frame 505 may be a neodymium, magnetic brushless motor with a total rating of about 1200 watts. An onboard diesel generator (not shown) may also be used to charge the batteries to allow the AMIV to cover longer distances if needed.

[0085] In one exemplary embodiment, motor 270 may be operationally coupled to drive axle 555 using a gear assembly having bevel gears for a 7:1 reduction, sufficient to propel the AMIVs at driving speeds up to about 20 MPH. However, most MI tasks are operated at around 6 - 10 MPH. Motor 270 may be operationally engaged by power module 235, such as a LiFe or LiPo battery, having about 2.5 kWh of battery capacity at greater than 14 Amp-hours. However, power module 235 may include any energy source to provide onboard electronics with electrical, mechanical, or solar power. Dry or wet-cell batteries such as nickel-cadmium, nickel metal hydrides, silicon-carbon, or lead-acid batteries may also be used. Furthermore, AMIVs 120 may include backup power systems to maintain the operation of essential electronics during power interruptions or failures. This is particularly important for GPS geocoordinate tracking and cellular communication. These backup systems can consist of ultracapacitors, auxiliary batteries, or similar technologies.Attorney Docket No. 36409-100110

[0086] In one embodiment, the RMIS of the present disclosure may utilize an energy management and optimization functionality (EMO) that intelligently regulates the AMIV's energy consumption based on its assigned Ml tasks and the system's overall operational demands. This latter functionality is performed by microcontroller 240. As AMIV 120 begins its task, the EMO evaluates the nature of the assigned task, rail conditions, battery levels, and weather conditions. If an AMIV is tasked with a high-energy-demand activity or extreme weather conditions, the system may adjust the cleaning intensity and propulsion speed to minimize energy consumption. If needed, any adjustments are communicated to MICS 125 for re-optimizing assigned tasks and additional deployment of AMIVs 120 to designated rail segments.

[0087] AMIV 120 may include a brake assembly (not shown) including one or more disc brakes, and a brake cylinder having linkage thereto. Engaging the brake cylinder causes the brake pads to squeeze against train wheels 525 and 530. Disc brakes work by using calipers to squeeze brake pads against the rotating disc of the train wheels - wheel or axle mounted - and located under the wheel carriage of the AMIV. Squeezing the brake pads creates friction, slowing train wheels 525 and 530 and, ultimately AMIV 120. The AMIV's kinetic energy is converted into heat dissipated by the disc, allowing the AMIV to slow down or stop.

[0088] In another embodiment, regenerative braking may instead be used where there is an energy recovery mechanism that slows down the moving AMIV 120 by converting its kinetic energy into a form of potential energy that can either be stored or used immediately, such as for recharging power module 235 (e.g., LiPo battery). Regenerative brakes operate by driving the electric motor in reverse to recapture energy otherwise lost as heat during braking.

[0089] Each AMIV 120 is preferably configured to accept a removable modular M1M 205 affixed preferably on either side of frame 515, and upward towards upper frame 510 and into some or all of framed enclosure region 520 beneath outer body shell 518. MIM 205 is preferably constructed of a durable, modular framework enclosure designed to accommodate a variety of accessory attachments, such as rail brushes, scrubbing wheels, sprayers, and specialized sensors. A modular framework enclosure provides a protective housing while allowing flexibility in configuration. Standardized mounting systems, such as a rail, slot, grid pattern, couplers, and other fittings may be utilized to secure and removably attach MIM 205 to the side frame of AMIV 120, such that the MI accessories may be positioned directly above the rails.Attorney Docket No. 36409-100110

[0090] Side or back access allows for quick installation, while integrated cable channels may be used to organize the routing of power and data connections. MIM 205 may be designed with sealing mechanisms, such as gaskets or weatherproofing elements, to protect internal electronics from environmental factors like dust, debris, moisture, or vibration, commonly encountered in outdoor rail environments. The modular framework of MIM 205 may be designed for adaptability to support a wide range of maintenance and inspection accessory attachments.

[0091] Some MIMs 205 are designed to handle different types of precipitation and track contaminants, such as pectin, black precipitate, snow, or ice. Others are designed with specialized sensors for track inspection. Still, others have specialized accessories for controlling vegetation growth on the tracks, performing rail grinding operations, or improving wheel-to-rail adhesion by applying sand or a high-friction coating, such as a specialized gel.

[0092] In one embodiment, Fig. 7 illustrates the details of a MIM 700 having three 10-inch diameter scrubbing wheels 705, preferably used for removing pectin and / or black precipitate from the rails. Each scrubbing wheel 705 is a rotating abrasive tool consisting of a metal hub with numerous stiff metal wires densely packed together in a cylindrical shape, designed to aggressively remove surface contaminants by applying friction through its spinning motion. Preferably, the scrubbing wheels should be positioned between 45 and 90 degrees to the longitudinal axis of the train rails for increased efficiency.

[0093] MIM 700 may be removably coupled to box tubings 515 of AMIV 120 using “U-shaped” members 710 attached to the distal ends of axle 715 supporting the components of the MIM. The sidewalls of box tubings 515 include a number of transverse vertical holes 720 distributed along and through the cross section of the length of box tubings 515. U-shaped member 710 may be secured in position along box tubings 515 by aligning holes 725 pierced through the sidewalls of U-shaped member 710 with one of the vertical holes 720, and passing pop pins 730 through the aligned holes 720, 725. This accommodates for different heights of different MIMs, and removably fixes MIM 700 to box tubings 515.

[0094] A bias actuator 735, preferably moving in a linear fashion extends shaft 740 outward from its lower end. The lower end of shaft 740 is attached to a bracket 745 welded to a cover plate 750. The shaft 740 is pivotally secured to bracket 745 by passing a fixed pin 755 through shaft 740 and aligned bracket holes 760. Bias actuator 735 may include a linear actuator having a cylinder and an internal lead screw, and a drive nut that works inside the cylinder at theAttorney Docket No. 36409-100110end of the shaft. Rotational motion is first generated by an electric motor and then reduced by a gearbox to increase the torque used to turn the leadscrew. The leadscrew then turns, resulting in the linear motion of the drive nut coupled to the leadscrew, thereby extending the shaft outward. However, other types of actuators may be used, which are categorized by their power source, such as mechanical, electromechanical, hydraulic, and piezoelectric actuators.

[0095] As shown, bias actuator 735 is secured at an inclined angle. The linear movement of shaft 740 about pin 755 permits the raising or lowering of MIM 700 to accommodate different applied forces to the rail. Bias actuator 735 further includes a U-shaped articulated member 765 at another distal end of bias actuator 735 coupled to upper frame 510 of AMIV 120. Similarly, U-shaped articulated member 765 is secured in position along upper frame 510 by aligning holes 770 with complementary holes 775 and passing a pop pin 780 through the aligned holes. Shaft 740 is capable of extending while MIM 700 is attached to the upper frame of AMIV 120, allowing scrubbing wheels 705 to apply a desired force onto the rail. Pop pins 730 and 780 may be readily removed, allowing MIM 700 to be removed from box tubings 515 and upper frame 510 of the AMIV. With the pop pins removed, MIM 700 may be disengaged, allowing U-shaped member 765 to pivot under the influence of a manual force, as U-shaped member 710 moves outward and away from box tubing 515. Of course, other coupling fasteners, such as a rod with flanged nuts, detents, cotter pins, or other quick release pins, may be used to facilitate removable attachment.

[0096] MIM 700 may include rotating mechanism 785, including a small-toothed gear 790, a large-toothed gear 791, a metal chain 792, and a brushless motor 793. Small-toothed gear 790 coupled to brushless motor 793 engages and rotates at about a 1:7 reduction with larger-toothed gear 791 through drive chain 792. Scrubbing wheels 705 may be removably attached to an axle 794 and larger-toothed gear 791 if axle 794 is partitioned into two or more sections, but connected using, for example, Oldham or spider couplers 795. Of course, other types of shaft couplers, adapters, and other fittings, may be used to allow for a quick disconnect. A 600-Watt electric brushless motor 793, for example, may be coupled to smaller gear 790 to indirectly rotate scrubbing wheels 705 between 1000 and 2000 rpm.

[0097] Preferably, bias actuator 735 is a force-sensing actuator equipped with an internal sensor (not shown) — such as a load cell or strain gauge — to measure and provide feedback on the force applied to shaft 740, and in turn to scrubbing wheels 705. Under the control of control processor module 210, bias actuator 735, preferably electric, biases scrubbing wheels 705 byAttorney Docket No. 36409-100110extending downward as described herein and with enough force to press scrubbing wheels 705 orthogonally against the rails. Bias actuator 735 — preferably a force-sensing actuator — incorporates a spring shock absorber, allowing scrubbing wheels 705 to adjust independently to variations in rail height or vertical inconsistencies. This setup enables scrubbing wheels 705 to maintain consistent contact with the rails, applying an approximately constant force of 50 - 80 lbs. As scrubbing wheels 705 rotate with this applied pressure, they effectively remove contaminants and precipitates — like pectin, ice or snow — as the AMIV 120 moves along the rails.

[0098] If the rails have a dip or depression, bias actuator 735 pushes scrubbing wheels 705 further downward to maintain contact. As scrubbing wheels 705 wear down, bias actuator 735 compensates by continuing to apply the same constant force. When necessary, the wheels can be easily replaced. However, excessive force may lead to premature wear of both the scrubbing wheels and the rail tracks.

[0099] Referring back to Fig. 2, as AMIV 120 moves along the track, diagnostic sensors 260-1 and 260-2 continuously monitor track conditions and feed this data to microcontroller 240 within a control loop. Microcontroller 240 processes the track conditions and determines the speed adjustments to ensure effective contaminant removal. It sends commands to drive module 230 to adjust motor speed and, if needed, applies braking force to slow down the AMIV. If contamination levels are low, the AMIV maintains its normal speed; otherwise, microcontroller 240 reduces speed proportional to the contamination level and may increase the scrubbing force applied to the rails.

[0100] Microcontroller 240 makes precise, real-time corrections by analyzing deviations from optimal cleanliness levels. Alternatively, AI / ML functionality may be used to predict and adjust speeds based on historical and real-time contamination data, providing a more adaptive response. Track conditions, contamination levels, and geographical coordinates may also be transmitted to MICS 125 for later review, helping schedule maintenance and inspection.

[0101] In one embodiment, diagnostic sensors 260-1 and 260-2 analyze rail cleanliness, with microcontroller 240 adjusting power, for example, to scrubbing wheel 705 and the speed of AMIV 120 to achieve the desired cleanliness level. The closed-loop control system adjusts power and speed based on the type and level of contamination detected by diagnostic sensors 260-1 and 260-2, such as moisture, pectin, snow, or ice. These diagnostic sensors, including cameras, LiDAR, infrared detectors, ultrasound, and friction sensors, continuously scan the rails in real-time.Attorney Docket No. 36409-100110

[0102] In another embodiment, diagnostic sensors 260-1 and 260-2 may be used to identify excessive wear on the rail tracks. The amount of excessive wear and its locations may then be transmitted to MICS 125 for later review and use in scheduling Ml tasks.

[0103] Scrubbing wheels 705 are particularly designed to remove pectin, carbonized debris, and / or corrosion. Other MIMs for cleaning and / or inspection may also be removably coupled to the frame of AMIV 120. Again, these modules include one or more coupling fasteners to facilitate removing the MIM. Similar components as those employing a scrubbing wheel may be employed, as described herein above.

[0104] With reference to Figs. 8 - 11, below herein are descriptions of various MIMs having specialized attachment accessories for a particular maintenance or inspection task with elements of Fig. 7 are labeled similarly.

[0105] Fig. 8 illustrates an MIM 800 using a rail brush 805 that may be energy efficient and also suited for removing snow. Rail brush 805 is made from spring steel tines 810 instead of scrubbing wheels. MIM 800 similarly may include a bias actuator 735, preferably a force-sensing actuator. Brushless motor 793 coupled to the brush head 805 may be used to better position spring steel tines 810 relative to the rail to remove the snow on the rails.

[0106] When AMIV 120 is transported, stored, or disengaged from maintenance and / or inspection tasks, the accessory attachments, such as scrubbing wheels 705 or rail brush 805 may be raised or positioned to have sufficient clearance off the ground. In operation, scrubbing wheels 705 or rail brush 805 is sufficiently lowered to engage the train rails for its particular task.

[0107] In another embodiment illustrated in Fig. 9, a MIM 900 may include a rail grinder 905 for minor rail surface corrections, such as removing minor irregularities or polishing rail heads caused by worn tracks from rail corrugation. Rail grinding removes a very small amount of metal from the rail's surface using, for example, three rotating grinding wheels 910 made of stone or the like and set at particular angles to restore the track rail to its correct profile or remove surface defects. Of course, larger rail grinder trains weighing a ton or more would be necessary for significant rail profile corrections.

[0108] It is contemplated that some MIMs may be designed to have specialized accessory attachments for controlling vegetation growth on or along the train rails that may cause safety hazards. Referring to Fig. 10, a MIM 1000 may alternatively include a specialized spray system having a tank 1005 and sprayer 1010 to spread and / or spray traction gel 1015 onto the head of theAttorney Docket No. 36409-100110rails. Gel 1015 is picked up by the train wheels and carried along, treating the rail head. The gel is typically a mixture of sand-like particles combined with a viscous liquid, providing abrasive action and adhesion enhancement. In that manner, the gel enhances the coefficient of friction at the wheelrail interface, and improves the grip between the train wheels and the rails, especially in conditions where adhesion might be low due to factors like wet weather, leaf fall, or frost. Furthermore, tank 1005 may alternatively contain herbicides 1020 that are sprayed along the tracks to help control invasive vegetation growth.

[0109] Fig. 11 illustrates a MIM 1105 mechanically coupled to the frame of an AMIV 1120 using coupling fasteners to facilitate removable attachment. However, MIM 1105 may include instead specialized sensors 1130, such as cameras, transducers, vibration sensors, ultrasound sensors, lasers, and the like to measure track geometry and inspect for flaws: ultrasound sensors to detect internal flaws, cracks, and discontinuities in the rail; eddy current sensors to detect surface and near-surface flaws by inducing an alternating current in the rail; magnetic sensors to detect non-ferromagnetic defects; x-ray sensors to detect damage in specific locations, such as bolt holes and welds; laser-based sensors to measures width, flatness, elevation in curves, gauge, and geometry irregularities such as curvature, cross-level, warp, twist, and alignment; and radar sensors to detect water damage or deterioration in the track foundation. Additionally, MIM 1105 may include rolling gauge readers to ensure the track is spaced at the standard distance. In the United States, the distance is 56.5 inches.

[0110] Likewise, if appropriate, a bias actuator may be used to allow the above-indicated sensors to maintain a distance from or in continuous contact with the rails. Of course, the applied bias force is adjusted for the particular application where it is necessary for the sensors to contact the rails. When AMIV 1120 is transported, stored, or disengaged from maintenance and / or inspection tasks, sensor(s) 1130 may be similarly raised or positioned to have sufficient clearance off the ground.

[0111] In one embodiment, AMIV may be configured explicitly with diagnostic sensors 260-1, 260-2, and / or sensors 1130 to inspect the tracks for assessing the rail conditions, including the tracks, railheads, ballast, vegetation growth, cross-ties, and / or fasteners. Sensor(s) 1130, such as cameras, may use any form of optical recording, including infrared cameras, LIDAR, and thermal imaging, to capture the visual information on or near the vicinity of the tracks for determining the rail integrity, surface wear, and thermal variations. Diagnostic sensors 260-1, 260-Attorney Docket No. 36409-1001102, as well as sensor(s) 1130 may also have a motorized lens for adjusting its focal length, as well as being mounted on a motorized gimbal to orient the camera to better capture the surroundings around the tracks. Al image analysis algorithms may be readily used to identify track defects, such as cracks and / or corrosion, as well as vegetation growth encroaching on the rails. Likewise, mapping these rail conditions may be automatically logged and geolocated using the coordinates from GNSS module 280, then transmitted to MICS 125 for later review, and used in scheduling MI tasks, including dispatching train crew to attend to immediate repairs.

[0112] In one embodiment, database servers 135 are augmented with simulated track conditions using physics-based or generative models for training, track and station layout, train schedules, generated optimal maintenance routes, and / or other information described herein. Database servers 135 may include cloud storage devices, with information stored therein that may be accessed by AMIVs 120, MICS 125, and / or user mobile devices 130.

[0113] RMIS 100 includes a plurality of AMIVs 120 with their identification and location having been registered therewith, and the AMIVs 120 receive optimal MI tasks from MICS 125. The design and functionality of the AMIVs 120 have been described above. As described herein, AMIVs 120 operate substantially autonomously on the rails once provided with optimal rail segments and their MI tasks to minimize train downtime and cost. In some embodiments an optimization algorithm may be employed to reduce downtime during service hours. The optimization algorithm may prioritize safety-related MI tasks, and delegate cost-reducing tasks to a lower priority level. AMIVs 120 are preferably configured to capture and / or generate sensor data relating to environmental conditions, and performance of the AMIVs. Environmental conditions may include weather, track containment conditions, track usage patterns, seasonal effects, and observed obstacles, including other vehicles, persons, or objects. Performance data may include battery levels, maintenance logs, vehicle usage, mileage, and cleaning intensity. MICS 125 may transmit any data to AMIVs 120, including optimal assigned rail segments for maintenance and / or inspection, task instructions, and any modified rail segments based on environmental data received from other AMIVs 120.

[0114] MICS 125 has AI / ML functionality to implement routing and analytics. Central processor 150 may receive all data from the train crew through web-based apps running on mobile devices 130. Database servers 135 may include a weather database server associated with and / or maintained by third-party weather services, such as the National Weather Service (NWS) thatAttorney Docket No. 36409-100110provide current weather conditions and forecasts affecting the geographical region where rail network 105 is located. Weather database servers may communicate with MICS 125, and / or crew mobile devices 130 to transmit and / or receive information.

[0115] Database servers 135 may also include a rail database server maintained by train rail network 105. This database server provides rail line topography overlaid with a geographical map of the region, providing MICS 125 with information and data regarding the paths of the different train lines, including their location, stops, train schedules, and rider traffic. MICS 125 may also have access to database servers of the Federal Railroad Administration (FRA) containing topography maps with geospatial resources for identifying stations and tracks for rail networks in the United States.

[0116] Al module 140 creates an efficient strategy for assigning AMIVs 120 and their assigned rail segment 110 for MI tasks. Examples of the types of Al models, and associated inputs and outputs, that may be used by the Al module 140 have been described above. AMIV 120 identification information and their associated storage sites, as well as sensor data from database servers 135 are used to partition train lines, allocate AMIVs 120 and assign their MI tasks. These tasks and rail segment assignments are developed with routing strategies to optimize the use of AMIVs 120 and lower operating costs and downtime. In some embodiments, the Al module 140 may develop strategies for the greatest maintenance system performance (e.g., cleaning) without prioritizing cost, or for minimizing cost and train downtime. Al module 140 may also create optimal MI tasks based on a cost analysis for each task, accounting for labor costs and commuter train schedules.

[0117] In partitioning the train lines, and assigning Ml tasks, Al module 140 may identify and prioritize train lines for maintenance and inspection while deprioritizing others with lower ridership. As such, Al module 140 may allocate more AMIVs 120 to specific train lines than others and transmit corresponding instructions to the AMIVs. Al module 140 may identify rail segments for contaminant inspection rather than maintenance in partitioning the train lines. Al module 140 may also instruct AMIVs 120 to travel along designated tracks to map out obstructions, anomalies, potential hazards, rail flaws, or track geometry.

[0118] Rail segments may also be prioritized based on track conditions, past, current and forecast weather conditions, rail contaminants, past maintenance and cleaning records, environmental data on local vegetation growth, and track usage frequency. Al module 140 may beAttorney Docket No. 36409-100110further configured to identify maintenance patterns or trends as historical data is collected with continued usage and stored in database servers 135 to augment prioritization. In one embodiment, Al module 140 identifies maintenance, inspection and / or repair patterns, and their corresponding rail locations. This allows Al module 140 to develop maintenance and / or inspection strategies to preemptively instruct AMIVs 120 to perform MI tasks for assigned rail segments, and preferably schedule them on off-peak train hours, such as during late night or early morning hours before heavy ridership occurs.

[0119] In an embodiment, MICS 125 may be linked and / or communicate with automatic train protection (ATP) and automatic train control (ATC) systems of many of today's modem train systems, where normal signaling operations, such as route setting and train regulation, are carried out by the ATP and / or the ATC system. These later systems work together to maintain a train's safe distance apart by marginally adjusting moving and station dwell time to ensure safe spacing between trains. These systems may be a source of external data, such as train position, current speed, track conditions (including speed limits, gradients, and switch positions), signal information, upcoming station stops, timetable data, and any relevant trackside information transmitted via trackside beacons (balises), which the train receives and uses to automatically adjust its speed and movement according to safety regulations and operational schedules.

[0120] Fig. 12 illustrates an exemplary flow diagram of method 1200 according to an embodiment of the present disclosure, which partitions the train lines into rail segments 110, determines and assigns a MI task for each rail segment 110, and allocates AMIVs 120 to partitioned rail segments 110 for specific MI tasks, aiming to reduce downtime and costs. In this embodiment, method 1200 is executed by MICS 125, utilizing AI / ML to improve the efficiency of MI tasks while optimizing their execution. MICS 125 is responsible for executing the Al and ML functionality using Al module 140 for various tasks, including routing, analytics, reporting, and system management. These operations may involve specialized instructions requiring a central processor 150 to work with additional coprocessors.

[0121] Method 1200 begins with step 1205, which involves retrieving from database servers 135 rail network information, including the number of lines and branches, the number of stations, the mileage of rail for each line, train schedules, the geolocations or geographic coordinates of the track lines, and any other available data relevant to the commuter service of train rail network 105. In step 1205, MICS 125 also accesses database server 135 to retrieve RMISAttorney Docket No. 36409-100110information, including the identification and storage locations of AMIVs 120, the operational status and capability of each AMIV, the personnel costs associated with operating each AMIV per mile of track, and historical data related to previous inspection and maintenance activities, available MI tasks, the track conditions, current and forecasted weather, task priorities, track assignment records, optimization algorithms, and training datasets.

[0122] At step 1210, based on the rail network 105 and RMIS information, the train lines are partitioned into rail segments 110. In step 1210, specified MI tasks are also determined and assigned for each rail segment 110 based on track and weather conditions, including the start and estimated completion times of the tasks. Database servers 135 contain specific details regarding the type of MI tasks required, which may depend on factors such as the type of contaminants, debris, or precipitation, along with the priority levels of the specific tasks. Extreme cold, snow, and ice can impact train rails that run at street level or above ground more than others, requiring snow and ice maintenance service on those corresponding rail segments. Other rail segments may only need routine inspection for debris or flaws, or rail maintenance to remove pectin or leaves because of fall foliage.

[0123] AMIVs are also equipped with optical sensor(s) and wireless communication systems, which can provide real-time updates on maintenance and / or inspection activities to both train crew and a maintenance and inspection computer server (MICS), which also has access to multiple database servers storing track and weather conditions. In step 1215, MICS 125 allocates AMIVs 120 to rail segments 110, preferably based on the proximity of their corresponding storage sites to the rail segments, ensuring that each AMIV 120 is assigned to the nearest geographical rail segment. Additionally, MICS 125 considers the train schedules of the trail rail network 105 and limits the time AMIVs 120 may operate, scheduling MI tasks preferably within train operation gaps to prevent service disruptions.

[0124] In steps 1210 and 1215, MICS 125 applies Al and ML techniques to analyze rail network information, RMIS information, and task definitions to generate an assignment of AMIVs 120, rail segments 110, and MI tasks and schedules. The process considers the number of AMIVs 120 required and their designated MI tasks while minimizing downtime and cost. Rail segments with the worst track conditions and / or high passenger traffic are assigned a higher priority, ensuring maintenance in those areas is completed before lower-priority segments. Rail segments may also be prioritized based on track conditions, current and forecast weather conditions, railAttorney Docket No. 36409-100110contaminants, past maintenance and cleaning records, environmental data, and track usage frequency. Al models use historical task performance data to refine assignments, applying reinforcement learning and deep neural networks to improve future scheduling. An optimization function may be modeled mathematically as minimizing the total operational cost and downtime, represented as the sum of individual AMIV costs and task execution times, weighted by a balancing factor.

[0125] Once rail segments and AMIV assignments are determined in steps 1210 and 1215, MICS 125 transmits the assigned rail segments 110 and tasks to the allocated AMIVs 120 and train crew, as in step 1220. Each AMIV 120 receives its assigned rail segment 110, the specific MI task to be performed, the scheduled time frame for execution, and any operational constraints based on train schedules. In step 1225, train crew 155 quickly deploy AMIVs 120 from nearby storage sites 115 to their assigned rail segments 110. Storage sites 115 may be situated above or below ground, but are readily accessible by train crew. These AMIVs 120 are fitted with an appropriate MIM for the assigned MI tasks. Preferably, AMIVs 120 are deployed during gaps in train operation for minimal disruption to active rail traffic. It should be recalled that AMIVs 120 are relatively mobile and lightweight, making for fast and safe on-and-off track deployment using pick-up trucks and small cranes. This ensures that maintenance activities are conducted efficiently with minimal impact on rail operations.

[0126] In some embodiments, while AMIVs 120 traverse along their assigned rail segments, real-time sensor data from AMIVs 120 and trackside sensors are monitored to update track conditions, weather hazards, and task completion status, as in step 1230. If unexpected track obstructions, severe weather conditions, or maintenance failures are determined in decision step 1235, an updated maintenance and inspection schedule is generated in step 1240, and AMIVs 120 may be added as needed. The method returns to step 1220, and any updated MI tasks, rail assignments, and scheduling times are transmitted to AMIVs 120. Train crew may be informed of additional AMIVs 120 required, and additional AMIVs 120 may be positioned in designated rail segments to effect any new optimal maintenance and / or inspection schedule. If it is determined in decision step 1245 that AMIVs 120 have not yet completed their assigned MI tasks, the present exemplary method continues until the tasks are completed. However, once the assigned MI tasks have been completed, MICS 125 in step 1250 alerts the train crew to load AMIVs 120 back onto pick-up trucks, return them to their designated storage sites 115, and recharge them. In step 1255,Attorney Docket No. 36409-100110MICS 125 then receives updated status on the MI tasks completed and any updated RMIS and rail network information, including rail conditions, maintenance and / or inspection status, and AMIV status.

[0127] Furthermore, the AI / ML functionality within MICS 125 analyzes historical task performance to refine future assignments and priorities. If a rail segment 110 repeatedly fails to meet cleanliness or maintenance standards during post-inspection, MICS 125 automatically flags it for additional AMIV 120 deployment in future schedules.

[0128] The described process establishes an Al adaptive maintenance strategy that optimizes AMIV task assignments using real-time and historical data. By leveraging machine learning for predictive maintenance scheduling, method 1200 effectively reduces downtime, lowers costs, and enhances the reliability of rail infrastructure while dynamically adapting to changing operational conditions.

[0129] The exemplary method above includes various processing steps. These executable instructions embody any one or more of the methods, operations, or functions of the present invention. These steps may be embodied in computer-executable instructions and stored on any computer-readable medium, such as solid-state memories, optical media, and magnetic media.

[0130] In some embodiments, the above method may use optimization mathematically modeled as the sum of individual AMIV costs and task execution times, weighted by a balancing factor. One mathematical model for the optimization in method 1200 may use the following set of notations and parameters, as shown in Table 1.TABLE 1Input Optimization Parameters for Mathematical ModelS: Total number of rail segments r.N: Total number of AMIVs.T: Total available time slots (off-peak hours).CjirOperational cost of AMIV j performing task i on rail segment r.Attorney Docket No. 36409-100110DJsrTime required for AMIV j to travel from storage location s to rail segment r.PirPriority level of task i on rail segment r based on weather, rail conditions, and ridership levels.RirIndicator for immediate need of task i on rail segment r (1 if required, 0 otherwise).H Set of high-priority rail segmentsL: Total track mileage.lrLength of rail segment r, satisfying:sItr=ldiDuration for any AMIV to complete task i.

[0131] The mathematical model may use the following decision variables, as shown in Table 2.TABLE 2Optimization Decision Variables for Mathematical Modelxijr = {1< if AMIV j is assigned to task i in rail segment r; 0, otherwisebi= Start time of task iei= End time of task iSj = Storage location of AMIV jAttorney Docket No. 36409-100110

[0132] The objective is to minimize the combined cost of task execution and total task completion time, ensuring that priority tasks are performed. The optimization function is given as:TABLE 3Optimization: Objective Functionmin Σ Cjirxijr+ λΣ (ei— bi)i,j,r iwhere A is a balancing factor that regulates the trade-off between cost and time.

[0133] The optimization of the mathematical model is subject to, but not limited to, the following constraints:TABLE 4Constraints1. Task Assignment: Each task must be assigned to at most one AMIV in a given rail segment:xijr< 1, Vi,rJThis ensures that no more than one AMIV is assigned to perform a specific task in a single rail segment.2. AMIV Availability: Each AMIV can only be assigned one task at a time:’xijr — 1'i,rAttorney Docket No. 36409-1001103. Total Available AMIV The total number of AMIV assigned at any time cannot exceed the fleet size:xijr< N, VrJ4 Time Constraints: Tasks must be scheduled within available operational hours:— Tstart' — 'i'end> Vi5 Priority Consideration: Higher priority tasks must be scheduled before lower priority ones:xijrPir> xikrPkr, if Pir> Pkr6. Vehicle Travel Time: An AMIV must have sufficient time to travel to a task location before starting the task:bi≥ sj+ Dj,r, ∀i, j, r7. Emergency Handling: Emergency repairs must take precedence over all other tasks:xijrRir≥ xikrRkr, if Rir= 18. Task Completion: Each task has a predefined duration:ei= bi+ di, ∀i9. Off-Peak Scheduling: Tasks must be scheduled when trains are not running:xijr = if trains are running10. High-Priority Rail Segments: Tasks in high-priority rail segments must be assigned first:xijr> 1, Vi, r E Hj

[0134] To implement the above mathematical model, MICS 125 may use an Al-based system that utilizes machine learning and constraint-solving techniques to generate optimal schedules and AMIV assignments. Such an Al-based system may be provided through Al module 140 or may be cloud-based. Several techniques may be employed, such as M1LP, RL, orAttorney Docket No. 36409-100110Metaheuristic Algorithms to find the best assignments for the AMIVs 120, their assigned tasks, and rail sections. MILP is a widely used optimization technique that reduces the scheduling problem as a set of linear constraints and decision variables. Al solvers such as Google OR-Tools, Gurobi, and CPLEX can readily process the objective function to determine optimal or near-optimal MI rail and task assignments.

[0135] The mathematical model representation involves defining binary decision variables that indicate whether a particular AMIV 120 is assigned to a specific MI task and rail segment. Constraints such as AMIV 120 availability, task priority, and time constraints are included with the optimization preprocessing input data, which may include environmental conditions, priority tasks, train schedules, and rail conditions, among other data. The MILP solver then generates an optimal maintenance plan by minimizing the total cost and task completion time, with a weighting factor balancing the two. These assignments may dynamically update the model in response to real-time data from AMIV sensors, train crew, changing weather conditions, and / or previously unreported repairs needing immediate attention.

[0136] Moreover, Reinforcement Learning (RL) is well suited for dynamic scheduling in environments that frequently change, such as unexpected weather events, track failures, or sudden vehicle malfunctions. RL allows MICS 125 to learn optimal scheduling from previous environmental conditions and adjust the strategies over time.

[0137] An RL framework may be structured around defining a state space that includes rail segments, available AMIVs 120, weather conditions, and task priorities. Al module 140 determines actions by assigning AMIVs 120 to rail segments 110 at specific times. A reward function is then designed to minimize cost and task completion time while ensuring that high-priority tasks are completed first. The training process for RL involves simulating various scheduling conditions, allowing the MICS 125 to learn optimal strategies through experience. With new data, the MICS 125 can dynamically adjust schedules, promptly addressing emergency tasks without disrupting overall efficiency. Metaheuristic approaches, such as Genetic Algorithms (GA), Ant Colony Optimization (ACO), and Simulated Annealing (SA), may also provide efficient, near-optimal solutions.

[0138] The Al-driven optimization system for the present RMIS requires real-time data from multiple sources, including weather forecasts, to prioritize the removal of snow, ice, or leaves and ensure that hazardous conditions are addressed. AMIV sensors can detect track flaws, andAttorney Docket No. 36409-100110vegetation overgrowth, which are essential for scheduling inspections and maintenance tasks. Train schedules determine available time slots for maintenance, allowing tasks to be executed during off-peak hours. Ridership volume also provides data for priority scheduling, and GPS helps generate estimated task durations. Of course, historical data offers insights from past performance, allowing Al models to make better predictions and improve decision-making.

[0139] The balancing factor 2 determines the trade-off between minimizing cost and minimizing total task completion time, and depends on several factors. However, one way to set the factor is to empirically test different values and see how the optimization results change. If A is too high, the model will strongly prioritize minimizing task duration over cost, likely leading to higher vehicle operating costs. If A is too low, the model will focus on cost minimization, possibly increasing the total time required to complete the tasks. Preferably, A should start with a small value, and gradually increase to see the trade-off.

[0140] To ensure that the cost and time are weighted fairly, A can be set based on their relative magnitudes. For example, if Cmaxis the maximum cost of an AMIV 120 performing a MI task, and Tmaxis the maximum time duration for all tasks, then A may be chosen as follows, ensuring that a unit increase in time is weighted proportionally the same as a unit increase in cost.c^maxA, — ” -TlmaxHowever, if minimizing cost is the top priority, then A should be chosen smaller, and if the task completion is more critical, then A should be chosen larger. Such a choice largely depends on cost and task constraints imposed by the administrator of the commuter rail network.

[0141] To ensure that the objective function does not simply avoid assigning MI tasks to minimize total cost, additional constraints or penalties may be imposed. In some embodiments, an added constraint may be to require that every task of high priority is assigned to an AMIV 120, ensuring that all high priority tasks are performed:xijr= 1, ∀i ∈ high priority tasksJ,rAttorney Docket No. 36409-100110

[0142] In some embodiments, the objective function may include a penalty for any high priority tasks not assigned:min ^,ijrCjtrxijr + b^) f (1xijr)If the M is sufficiently large, then high priority MI tasks will be assigned to AMIVs rather than leaving them unassigned to minimize cost.

[0143] In some embodiments, the objective function may include a term that rewards assigning a task, incorporating priority levels Pir.min ' CjirXijr-I- A. bi) (3 ’ Pirxijri,j,r i i,rwhere / ? is a positive weight ensuring that high-priority tasks are not ignored.

[0144] In some embodiments, the model may set a usage rate for the AMIVs 120 so that each AMIV 120 performs at least one task within the available hours:’xijr — Umin, ji,rwhere Uminis the minimum number of tasks per AMIV 120. Although most AMIVs 120 will be assigned one task, some may be assigned more than one if diagnostic sensors 260-1 and 260-2 are utilized to perform an additional task, such as simultaneously inspecting the tracks.

[0145] In some embodiments, the mathematical model may ensure that higher-priority MI tasks are maximized and completed. The objective function and constraints may be modified accordingly. Instead of only minimizing cost and task completion time, the model introduces a priority -weighted term to ensure that higher-priority tasks are assigned.max Pirxijr—Cjirxijr ~ (ei—^i) i,j,r i,j,r iAttorney Docket No. 36409-100110Piris the priority level of MI task z on rail segment r, ensuring high-priority MI tasks contribute more to the objective.Cjirrepresents the operational cost of assigning AMIV / to task z.(e, — bj) represents the total time taken for tasks, balancing efficiency.a and are weighting factors to balance cost minimization and time efficiency while prioritizing higher needed tasks.

[0146] In some embodiments, to determine the number of rail segments 110 when partitioning train lines, a simplistic approach may be used, dividing the total track length L by the desired segment length SEG, with the constraint that the number of rail segments does not exceed the number of available AMIVs 120.LS = - SEGS < Nwhere S’ is the number of rail segments, L is the total length of the train rail network, SEG is the desired length of each rail segment, and N is the number of available AMIVs 120.

[0147] In some embodiments, natural and operational constraints are considered in partitioning the train lines, such as breaking segments at access points where AMIVs 120 may be readily loaded and unloaded, or breaking segments to areas historically requiring frequent maintenance. Typically, MI tasks are performed during gaps in train operation, so partitioning would be based on the available maintenance time window so that rail segments could be serviced within that time. With the AMIVs 120 having a defined operational range, rail segment lengths should be small enough that an AMIV 120 could complete its MI task within its operational time and range.

[0148] As such, partitioning the train lines may use a more generalized approach as follows. The process of determining rail segments 110 for MI tasks begins with the initialization of segmentation points. The train rail network 105 is considered from a starting point, represented mathematically as x = 0, and a structured list is created to store the designated segment endpoints. These endpoints will later define the individual segments of the railway line that require MI tasks.Attorney Docket No. 36409-100110

[0149] Following initialization, mandatory segmentation points are identified, including key access points where AMIVs 120 may be readily loaded and unloaded, or areas historically requiring frequent Ml tasks, and may also include locations, such as stations and rail junctions. These fixed points serve as reference markers where segmentation is required to ensure AMIV 120 accessibility and to facilitate efficient scheduling. With mandatory segmentation points in place, the segment length is determined dynamically. An initial estimate for segment length, denoted as Sinitis calculated using the formula:Lc. —where L represents the total length of the track, and N is an estimated number of segments 110. The rail segment count N is derived from:Laavgwhere Savgis the average of the minimum Sminand the maximum permissible segment lengths, Smax. This ensures that segments 110 are neither excessively large nor too small. In this way, this constraint guarantees that each segment 110 does not exceed Smaxwhile also ensuring that no segment is shorter than Sminmaintaining operational feasibility.

[0150] Once the segment lengths are estimated, the segmentation is optimized based on the capabilities of the AMIVs 120. Each rail segment r is evaluated to determine whether an AMIV can complete its assigned maintenance within Tmaxthe operational time constraint. The required time for task execution is computed as:Dy — _r_i_ / - ' r ''workVwhere Drrepresents the segment length, v is the speed of the AMIV, and tworkaccount for the time needed to complete the Ml task. If Trexceeds Tmax, the rail segment is further divided into smaller rail segments to ensure that AMIVs can complete the MI tasks within their operational time limits. Also, an additional constraint is that the number of rail segments 110 does not exceed the number of available AMIVs 120.

[0151] In the partitioning process, each rail segment 110 may be structured to allow MI tasks to take place during off-peak hours, ensuring minimal train service disruption. If any railAttorney Docket No. 36409-100110segment is too large to be fully serviced within the allocated maintenance window, it is further subdivided to fit within the available time slots, with the constraint that more AMIVs 120 may be allocated.

[0152] This partitioning is followed by storing the confirmed segment breakpoints, and ensuring all rail segments 110 may be readily accessed by the AMIVs 120 with reference to the mandatory segmentation points. Once the railway line has been segmented, the total number of segments can be determined by counting the finalized segment breakpoints. The total number of rail segments is equal to the number of breakpoints minus one. Each rail segment 110 is then associated with a corresponding storage site, preferably the nearest one and an AMIV 120 allocated from that storage site for quick deployment.

[0153] To prioritize rail segments specifically for a given MI task, each rail segment 110 may be evaluated using a tailored priority scoring method. This approach allows the train crew to determine which rail segments 110 should be addressed first based on the urgency of conditions specific to the MI task.

[0154] The prioritization process may begin by defining the specific factors relevant to a particular MI task. Such factors may include current rail conditions such as contamination from ice, snow, pectin, rail wear, corrosion, or geometric irregularities. For example, if the MI task involves snow or ice removal, the relevant factors might include the current thickness of ice accumulation, forecasted snowfall, historical frequency of ice-related issues, ridership levels, and any emergency indicators that signal urgent maintenance needs. For other tasks, such as contamination removal, including pectin, relevant factors might instead include the current thickness of pectin, and for vegetation control or rail grinding, different factors, like vegetation growth rates, rail wear severity, historical inspection records, or the time since last maintenance would be considered. Also, real-time and forecasted weather conditions may be gathered, focusing on precipitation levels, temperature extremes, seasonal conditions like leaf fall, and other potential adverse weather. Such information may be readily retrieved from database servers 135, which have access to other real-time third-party databases, including the National Weather Service (NWS).

[0155] After identifying these relevant task-specific factors, each factor may be assigned a numerical score reflecting its severity or urgency, along with a corresponding weighting factorAttorney Docket No. 36409-100110representing its relative importance to the overall prioritization of the task. These weighted factors are combined into a single numerical value known as the task-specific priority score.[0156J In some embodiments, this prioritization can be expressed mathematically as follows:Pit,r1 = wcc> j. Ci r-I- wwvv £. Wii,ri + w rpFri -I- w nHj. Hic,ri + w CEii. Eii,ri + ws. S Ii, / rwhere:Pi ris the priority score of performing task i on rail segment r;Ciris the severity of conditions specifically relevant to task z on segment r (e.g., if task i is "ice removal," then Cirmeasures ice accumulation levels);Wi ris the real-time and forecasted weather conditions directly affecting task i on segment r (e.g., for snow removal, include snowfall intensity forecast);Firis the frequency or ridership importance on segment r relevant to task i; Hi rHistorical maintenance frequency or previous issues specifically for task i; Ei ris the emergency indicator specifically indicating urgent needs for task i on segment r (1 if an emergency; otherwise 0); andSi ris the time elapsed since last completion of task i on segment r.

[0157] The corresponding weights wc., ww, wF, wH, wE. and wsare chosen based on their importance, which can be set either empirically through expert knowledge or determined through historical data analysis.

[0158] For each rail segment r, the priority score for the given task is calculated by multiplying each factor's measured severity by its assigned weight, then summing these products. Rail segments with conditions indicating immediate attention, such as a combination of severe weather impacts, high contamination, high historical maintenance frequency, or emergency signals - will naturally produce higher scores. Once calculated, these scores allow rail segments to be ranked, highlighting which rail segments require immediate action for a specified MI task. Rail segments with the highest scores are assigned the highest priority, ensuring that critical tasks like urgent ice, snow, or pectin removal, vegetation management, or track repairs are addressed promptly.

[0159] This prioritization can be integrated into maintenance and / or inspection scheduling algorithms, allowing resources to be allocated efficiently and dynamically. And, by continuouslyAttorney Docket No. 36409-100110recalculating priority scores as new data becomes available, this method ensures the MI scheduling remains responsive to changing conditions.

[0160] In some embodiments, the prioritization may use ML models to predict the urgency of MI tasks, or the likelihood of rail defects based on historical track conditions, weather patterns, traffic intensity, and previous maintenance outcomes. Such a model would continuously learn and adjust the priority scoring accordingly. This prioritization may be integrated directly into assigning rail segments with specific MI tasks. MI tasks on rail segments with scores surpassing a critical threshold are marked as mandatory, guaranteeing their completion. Regular updates ensure that the priority scores reflect current conditions, allowing for timely responses to emerging maintenance and inspection needs.

[0161] Each AMIV 120 is preferably assigned to a rail segment based on the proximity of its storage site to the rail segment. In some embodiments, one method or algorithm for doing so is to compile geographical data on the locations of the rail segments 110 and the storage sites 115 for the AMIVs 120, which is readily available from database servers 135. For each rail segment 110 requiring a MI task, the distances from every storage site 115 to the midpoint of that rail segment 110 are calculated. Storage sites 115 are then ranked according to their proximity to the rail segment 110. Assignments are first attempted by selecting an AMIV 120 from the storage site 115 closest to the rail segment. If an AMIV 120 is unavailable from that storage site 115 due to other assignments or operational constraints, the next closest storage site is chosen, and so on. This iterative approach ensures that each rail segment 110 is serviced by an AMIV 120 from the nearest feasible storage site.

[0162] After initial assignments are made, operational constraints such as AMIV battery capacity, available operational hours, and maintenance window durations are verified. If any assignment proves infeasible, the algorithm recalculates assignments, selecting an AMIV 120 from the next closest storage site. The finalized assignment identifies which AMIV 120 services each rail segment 110. And if conditions change, the algorithm may dynamically reassess the AMIV 120 assignments to maintain optimal efficiency.

[0163] Figs. 13 and 14 depict exemplary displays or "screenshots" of the user interface of RMIS 100, highlighting some key functionalities of RMIS 100. Fig. 13 illustrates an exemplary " Home Page" 1300 and serves as the central entry point for accessing the functionalities of RMIS 100. " Homepage" 1300 presents the system's overall status, providing access to other availableAttorney Docket No. 36409-100110features. " Homepage" 1300 may be displayed at workstation 175 or on train crew mobile device 130, depicting the current locations of all active AMIVs 120 overlaid with a display 1305 of an exemplary train rail network 105. Note that the current locations of AMIVs 120 are displayed as a vehicle icon on the commuter rail map 1305. " Homepage" 1300 may also display a menu of icons and links 1210, allowing the user to access other functionalities.

[0164] In an exemplary embodiment, a menu of icons and links 1310 includes " AMIVs" 1320, " Maintenance / Inspection Schedule" 1330, " Assigned Rail" 1340, " Service Request" 1350, " Reports" 1360, and " Settings" (gear icon) 1370. The menu of icons and links 1310 could include fewer or additional icons and / or links for other functionalities of RMIS 100. The train crew upon selecting link " AMIVs" 1320, will be shown a list of active AMIVs 120 in the train system, and choosing one of the AMIVs 120 from a drop-down menu would display a second page 1400 displaying, but not limited to the following information for the selected AMIV 120, as also illustrated in Fig. 14.

[0165] TABLE 5AMIVIdentification Tag: 623Rail Segment: LO 0005Storage Site: WM 623Assigned task: Ice removalRail Mileage: 05Assigned start time: 2:30 AMEstimated end time: 4:40 AMStatus: In ProgressMile Post: RL 02Progress: 40% RemainingEstimated Time Remaining: 45 min

[0166] The format of the " Identification Tag" may have a unique three-digit corresponding to the AMIV vehicle I. D. number, whereas the " Rail Segment" may have the first two letters indicating the train line or branch code (e.g., LO for Lowell line), followed by the beginningAttorney Docket No. 36409-100110milepost and then the end milepost of the rail segment 110 to which the AMIV 120 has been assigned. The format of " Storage Site" may have the name of the nearest station (e.g., WM for West Medford) of the storage site for the particular AMIV, followed by three numbers corresponding to the vehicle ID. number. However, the locations of the storage sites 115, train stations, and assigned track segments 110 will be identified internally within database servers 135, preferably using more precise geospatial data sourced from the Federal Railroad Administration. Specific markers and reference points along the tracks will pinpoint the exact location of each assigned rail segment 110, AMIV 120, and train station within the commuter rail network.

[0167] Additional displayed information for the AMIV 120 may include the " Assigned Task," such as ice or snow removal, pectin removal, rail inspection, grinding, track geometry assessment, chemical spraying, and weeding, among other MI tasks. Further displayed are the assigned "rail mileage," "start time," and estimated "end time.". Also, hovering and holding a user's cursor over the vehicle icon on " Homepage "1300 will show the same latter information for the AMIV 120. Alongside the displayed information may be the status of the selected AMIV 120, such as " In Progress," " Inactive," or " Disabled," along with the task remaining shown in minutes and percentage, and its current location using the nearest milepost on the train line. This way, the train crew can readily see what task each AMIV 120 performs, the distance covered by each AMIV 120, and the type of maintenance or inspection.

[0168] Although not illustrated, when the " Maintenance / Inspection Schedule" link 1330 is selected, one or more pages may be displayed in tabular form, including the daily schedule for all AMIVs 120, such as rail segment assignments, assigned tasks, start and end times, mileage, and AMIV status. Also, it is contemplated that when the " Assigned Rails" link 1340 is selected, the rail conditions of all train lines under service and their associated MI tasks will be displayed, including scheduled start and end times, number of AMIVs assigned, their assigned rail segments and storage sites along with the assigned AMIV IDs, forecasted weather conditions, schedule track repairs, and priority settings of the tasks and repairs. When " Service" link 1350 is selected, the rain crew may communicate any repairs needed for the AMIVs 120. This link may also allow authorized train crew to add or schedule new tasks, which are transmitted to MICS 125 to check for any conflicting tasks.

[0169] When train crew 155 selects the " Settings" icon 1370, the user may be able to modify user-defined preferences for the AMIVs 120, display formats, and restrictions for specificAttorney Docket No. 36409-100110AMIVs 120 due to unreported needed repairs. If necessary, these preferences are communicated to MICS 125 to optimize rail and task assignments.

[0170] RM1S 100 may also display analytics when selecting the " Reports" link 1360. This allows a submenu of reports to be displayed, including forecasted weather conditions, estimated downtime for maintenance and repairs, detailed status of the AMIVs 120 and their onboard electronics, estimated labor costs, number of rail lines being serviced, performance and cleanliness levels, and track geometry measurements for the week, month, or year.

[0171] It should be emphasized that the above-described mathematical models, algorithms, and optimizations are merely examples of how the RMIS may utilize AI / ML to effect MI tasks on a train rail network. Other models, algorithms, and optimizations may readily be realized by those skilled in the art and who have been equipped with the understanding of the operation of the present RMIS as set forth in this disclosure.

[0172] Moreover, the corresponding structures and equivalents of all means plus function elements in the claims below are intended to include any structure performing the function in combination with other claimed elements as specifically claimed. Furthermore, the embodiments herein are merely illustrative of the principles of the invention. Various modifications may be made by those skilled in the art, which will embody the principles of the invention and fall within the scope thereof.

Claims

Attorney Docket No. 36409-100110CLAIMS1. A system for dynamically providing maintenance and inspection service to a train rail network, the system comprising:a plurality of autonomous vehicles configured to perform maintenance and inspection tasks on multiple train lines within the train rail network;a plurality of storage sites located throughout the train rail network along the multiple train lines, the plurality of autonomous vehicles being decentrally housed at the plurality of storage sites; anda computer server including a processor and a non-transitory memory, the computer server configured to perform operations comprising:partitioning the train lines into a plurality of discrete rail segments,acquiring track and weather condition data of each discrete rail segment, determining, based on the acquired track and weather condition data, a specific required maintenance and inspection task for each of the plurality of discrete rail segments, assigning at least one of the plurality of autonomous vehicles to each rail segment of the plurality of discrete rail segments based on the proximity of each of the plurality of autonomous vehicles decentrally housed in the plurality of storage sites to each of the plurality of discrete rail segments, andgenerating and transmitting to each assigned autonomous vehicle instructions to autonomously and concurrently proceed from its storage site to its assigned discrete rail segment and perform the required maintenance and inspection task.

2. The system of claim 1, wherein each of the plurality of autonomous vehicles is manually operable by a train crew member.

3. The system of claim 1, wherein each of the plurality of autonomous vehicles is portable, and configured to be readily deployed on and off an assigned rail segment.

4. The system of claim 1, wherein the plurality of discrete rail segments are prioritized for the assigned maintenance and inspection tasks based on at least one of: (i) track conditions,Attorney Docket No. 36409-100110(ii) past, current and forecast weather conditions, (iii) rail contaminants, (iv) past maintenance and cleaning records, (v) environmental data, and (vi) track usage.

5. The system of claim 1, wherein the computer server further includes an artificial intelligence module that includes an assignment model and a condition model, the artificial intelligence module configured to perform the steps of:partitioning the train lines into discrete rail segments;determining the specific required maintenance and inspection task for each of the plurality of discrete rail segments; andassigning at least one of the plurality of autonomous vehicles to each rail segment; andassessing the condition of a subsection of a discrete rail segment or its immediate surrounding environment, including but not limited to tunnels, vegetation, and obstacles.

6. The system of claim 1, wherein the step of assigning at least one of the plurality of autonomous vehicles to each rail segment is based upon an optimization algorithm for minimizing downtime and total cost for performing maintenance and inspection tasks on said train rail network.

7. The system of claim 7, wherein the optimization algorithm includes a constraint ensuring maintenance and inspection tasks having higher priority are assigned before lower priority maintenance and inspection tasks.

8. The system of claim 7, wherein the optimization algorithm includes a constraint assigning a priority weight to said rail segments or maintenance and inspection tasks so that higher priority maintenance and inspection tasks are assigned before lower priority maintenance and inspection tasks.Attorney Docket No. 36409-1001109. The system of claim 7, wherein the optimization algorithm partitions the multiple train lines based on at least one of: (i) key access points where said portable autonomous vehicles may be readily loaded and unloaded, (ii) train lines historically requiring frequent maintenance and inspection, and (iii) station locations and rail junction locations of the train rail network.

10. The system of claim 1, wherein each of the plurality of autonomous vehicles further comprises a modular maintenance and inspection module removably coupled to readily engage and disengage with each of the plurality of autonomous vehicles, said modular maintenance and inspection module including a specialized accessory attachment for performing a maintenance or inspection task assigned to said autonomous vehicle.

11. The system of claim 10, wherein the specialized accessory attachment includes brushes, scrapers, sprayers, or wires, configured to clean a particular contaminant or precipitate off said rail segments.

12. The system of claim 10, wherein the specialized accessory attachment includes diagnostic sensors to inspect and assess rail conditions.

13. The system of claim 12, wherein the diagnostic sensors includes at least one of: (i) a camera, (ii) a transducer, (iii) a vibration sensor, (iv) an ultrasound sensor, (v) a laser, (vi) an eddy current sensor, (vii) a magnetic sensor, (viii) a x-ray, (ix) a radar sensor, or (x) a rolling gauge reader.

14. The system of claim 10, wherein the specialized accessory attachment includes a grinder.

15. The system of claim 10, wherein the specialized accessory attachment includes a sprayer operable to spray herbicides on said rail segments.

16. The system of claim 10, wherein the specialized accessory attachment includes a sprayer operable to apply sand or a high-friction coating on the rails of said rail segments.Attorney Docket No. 36409-10011017. The system of claim 1 further comprising at least one database server containing the track and weather conditions including at least one of: (i) past, current and forecast weather conditions, (ii) topography data for the train rail network, (iii) rail line and station data, (iv) train schedules, (v) status of each of the portable autonomous vehicle, (vi) sensor data relating to the surface conditions of the track lines, (vii) the identification of contaminants, debris, corrosion and unevenness on the track lines, (viii) historical data on past maintenance and cleaning records, (ix) environmental data on local vegetation growth, and (x) track usage frequency.

18. The system of claim 1, wherein the computer server is further configured for generating analytics and reports directed to an operating status of the system.

19. The system of claim 1, wherein each of the plurality of autonomous vehicles includes one or more diagnostic sensors for providing real-time updates on the track conditions of its assigned rail segment.

20. The system of claim 1, wherein each of the plurality of autonomous vehicles includes one or more diagnostic sensors to inspect or monitor the surrounding track environment of its assigned rail segment.

21. The system of claim 20, wherein the one or more diagnostic sensors include at least one of a camera, laser, and LiDAR.

22. The system of claim 20, wherein the assigned maintenance and inspection tasks are modified based on the track environment surrounding the assigned rail segments.

23. The system of claim 1 further comprising a wireless communication network configured to enable communication between the plurality of autonomous vehicles and the computer server.Attorney Docket No. 36409-10011024. The system of claim 1, wherein each of the plurality of autonomous vehicles further includes a control processor module including a microcontroller that controls the assigned maintenance and inspection task and controls communication with the computer server.

25. The system of claim 1, wherein each of the plurality of autonomous vehicles further include a global navigation satellite system operable to calculate geocoordinates of the autonomous vehicle.

26. The system of claim 1 further comprising one or more mobile devices configured for train crew to query the plurality of autonomous vehicles regarding the operating status of the system.

27. The system of claim 26, wherein the one or more mobile devices are further configured to assign a maintenance and / or inspection task to one or more of the plurality of autonomous vehicles.

28. A method for directing the maintenance and / or inspection of a train rail network having a plurality of train lines, the method comprising the steps of:stationing a plurality of storage sites along the train lines of the train rail network; housing a plurality of portable autonomous vehicles at the plurality of storage sites such that the plurality of portable autonomous vehicles are decentrally located along the train rail network, each of said portable autonomous vehicles including a specialized accessory attachment for performing an assigned maintenance and / or inspection task; partitioning each of said train lines into one or more rail segments;assigning each of the one or more rail segments an assigned portable autonomous vehicle from among the plurality of portable autonomous vehicles based on the location of the storage site housing said assigned autonomous vehicle; andassigning an assigned maintenance and / or inspection task to each assigned portable autonomous vehicle based on track conditions of the assigned rail segment.Attorney Docket No. 36409-10011029. The method of claim 28 further comprising the step of prioritizing the assignment of maintenance and / or inspection tasks the one or more rail segments based on at least one of: (i) track conditions, (ii) past, current and forecast weather conditions, (iii) rail contaminants, (iv) past maintenance and cleaning records, (v) environmental data, and (vi) track usage.

30. The method of claim 28, wherein the step of assigning an assigned maintenance and / or inspection task is additionally based upon an optimization algorithm for minimizing downtime and total cost for performing maintenance and / or inspection tasks on said train rail network.

31. The method of claim 30, wherein the optimization algorithm includes a constraint ensuring maintenance and / or inspection tasks having higher priority are assigned before lower priority maintenance and / or inspection tasks.

32. The method of claim 30, wherein the optimization algorithm includes a constraint assigning a priority weight to said rail segments or the maintenance and / or inspection tasks so that higher priority maintenance and / or inspection tasks are assigned before lower priority maintenance and / or inspection tasks.

33. The method of claim 30, wherein the step of partitioning each of said train lines into one or more rail segments is performed by the optimization algorithm based on at least one of: (i) key access points where said portable autonomous vehicles may be readily loaded and unloaded, (ii) train lines historical requiring frequent maintenance and / or inspection, and (iii) station locations and rail junction locations of the train rail network, and (iv) track usage.

34. The method of claim 28, wherein the track conditions include at least one of: (i) past, current and forecast weather conditions, (ii) topography data for said train rail network, (iii) rail line and station data, (iv) train schedules, (v) status of each of said portable autonomous vehicles, (vi) sensor data relating to the surface conditions of the track lines,Attorney Docket No. 36409-100110(vii) the identification of contaminants, debris, corrosion and unevenness on said track lines, (viii) historical data on past maintenance and cleaning records, (ix) environmental data on local vegetation growth, and (x) track usage frequency.

35. The method of claim 28 further comprising the step of generating analytics and reports directed to the operating status of the train system.

36. A non-transitory computer-readable medium storing instructions that, when executed on a maintenance and inspection computer server, cause the maintenance and inspection computer server to perform steps comprising:partitioning one or more train lines into a plurality of rail segments; assigning each of said rail segments an assigned autonomous vehicle from the plurality of autonomous vehicles based on the location of the storage site of said selected autonomous vehicle;acquiring track and weather condition data of each of the plurality of rail segments;assigning an assigned maintenance and inspection task to each assigned autonomous vehicle based on the acquired track and weather condition data; and generating a task instruction comprising the assigned rail segment and the assigned maintenance and inspection task corresponding to each assigned autonomous vehicle; andtransmitting the task instruction via a wireless communication link corresponding to each assigned autonomous vehicle.

37. The non-transitory computer-readable medium of claim 36, wherein said instructions when executed further cause the computer server to perform the step of prioritizing said rail segments for assigned maintenance and inspection tasks based on at least one of: (i) track conditions, (ii) past, current and forecast weather conditions, (iii) rail contaminants, (iv) past maintenance and cleaning records, (v) environmental data, and (vi) track usage.Attorney Docket No. 36409-10011038. The non-transitory computer-readable medium of claim 36, wherein said assigned maintenance and inspection tasks are based upon an optimization algorithm for minimizing downtime and total cost associated with performing maintenance and inspection tasks on said train rail network.

39. The non-transitory computer-readable medium of claim 38, wherein said optimization algorithm includes a constraint ensuring maintenance and inspection tasks having higher priority are assigned before lower priority maintenance and inspection tasks.

40. The non-transitory computer-readable medium of claim 38, wherein said optimization algorithm includes a constraint assigning a priority weight to said rail segments or the maintenance and inspection tasks so that higher priority maintenance and inspection tasks are assigned before lower priority maintenance and inspection tasks.

41. The non-transitory computer-readable medium of claim 38, wherein the step of partitioning one or more train lines is executed by the optimization algorithm based on at least one of: (i) key access points where said portable autonomous vehicles may be readily loaded and unloaded, (ii) train lines historical requiring frequent maintenance and inspection, (iii) station locations and rail junction locations of the train rail network, and (iv) track usage.

42. The non-transitory computer-readable medium storing instructions of claim 38, wherein said track conditions include at least one of: (i) past, current or forecast weather conditions, (ii) topography data for said train rail network, (iii) rail line and station data, (iv) train schedules, (v) status of each of said portable autonomous vehicles, (vi) sensor data relating to the surface conditions of the track lines, (vii) the identification of contaminants, debris, corrosion and unevenness on said track lines, (viii) historical data on past maintenance and cleaning records, (ix) environmental data on local vegetation growth, and (x) track usage frequency.Attorney Docket No. 36409-10011043. The non-transitory computer-readable medium storing instructions of claim 38, wherein said instructions when executed further cause the computer server to perform the step of generating analytics and reports directed to the operating status of the system.

44. A maintenance and inspection vehicle for a train rail network having one or more rails, the maintenance and inspection vehicle comprising;a vehicle frame having an enclosure that extends at least over said rails when the maintenance and inspection vehicle is propelled along said rails of the train rail network;a plurality of wheels coupled to the vehicle frame;a gear assembly;a motor coupled to the vehicle frame and the gear assembly and operable to power the plurality of wheels and propel the maintenance and inspection vehicle along said rails of the train rail network;a modular maintenance and inspection module removably coupled within said enclosure to the vehicle frame, wherein the modular maintenance and inspection module includes a specialized accessory attachment for performing an assigned maintenance or inspection task;one or more processors; anda memory storing program instructions that when executed by said one or more processors causes the maintenance and inspection vehicle to autonomously propel itself along the rail network and causes the specialized accessory attachment to operably engage the rails to perform the assigned maintenance and inspection task.

45. The maintenance and inspection vehicle of claim 44, wherein the specialized accessory attachment includes one or more brushes, scrapers, or wires, configured to clean a particular containment or precipitation off said rails.

46. The maintenance and inspection vehicle of claim 44, wherein the specialized accessory attachment includes one or more diagnostic sensors to inspect and assess rail conditions.Attorney Docket No. 36409-10011047. The maintenance and inspection vehicle of claim 46, wherein the one or more diagnostic sensors includes: (i) a camera, (ii) a transducer, (iii) a vibration sensor, (iv) an ultrasound sensor, (v) a laser, (vi) an eddy current sensor, (vii) a magnetic sensor, (viii) an x-ray, (ix) a radar sensor, or (x) a rolling gauge reader.

48. The maintenance and inspection vehicle of claim 45, wherein the specialized accessory attachment includes a grinder.

49. The maintenance and inspection vehicle of claim 44, wherein the specialized accessory attachment includes a sprayer operable to spray herbicides on and around said rails.

50. The maintenance and inspection vehicle of claim 44, wherein the specialized accessory attachment includes a sprayer operable to apply sand or a high-friction coating to said rails.

51. The maintenance and inspection vehicle of claim 44, wherein the maintenance and inspection vehicle includes a control processor module including a microcontroller configured to control the maintenance and inspection vehicle to perform the assigned maintenance and inspection task and to control communication to a computer server and track maintenance crew.

52. The maintenance and inspection vehicle of claim 44, further comprising diagnostic sensors configured to provide real-time updates on the track conditions of an assigned rail segments.

53. The maintenance and inspection vehicle of claim 51, wherein the microcontroller is operable to continuously monitor rail conditions and adjust the speed of the vehicle based on said rail conditions to optimally perform the assigned maintenance and inspection task.Attorney Docket No. 36409-10011054. The maintenance and inspection vehicle of claim 44, wherein the maintenance and inspection vehicle further comprises a global navigation satellite system operable to calculate the geocoordinates of the vehicle.

55. The maintenance and inspection vehicle of claim 44, wherein the motor is an electric motor.

56. The maintenance and inspection vehicle of claim 44, the vehicle further comprising a power module operable to supply energy to the motor, the one or more processors, and the maintenance and inspection module.

Images