A loader work path optimization method based on a multi-agent system

Abstract

This invention discloses a loader operation path optimization method based on a multi-agent system, comprising the following steps: acquiring the loader status, operation target, and regional resources, and constructing a loader operation resource map; converting the material picking and unloading posture constraints to generate operation end posture anchor points; generating operation phase chains and hinge coupling occupancy zones based on the end posture anchor points; extracting resource end posture accommodating windows and performing embedding matching to generate end posture accommodating paths; identifying end posture occupancy conflicts and transferring them to generate conflict transfer results; and reusing end posture inheritance passage windows to generate collaborative operation path results. This invention utilizes end posture anchor points and hinge coupling occupancy zones to optimize the loader collaborative path, possessing advantages such as accurate posture matching, stable conflict transfer, and strong operation continuity.

Description

Technical Field

[0001] This invention relates to the field of mobile equipment path planning technology, and in particular to a method for optimizing the operation path of a loader based on a multi-agent system. Background Technology

[0002] Current loader operation path optimization methods are mostly based on the work area map, vehicle's current location, material pick-up point, unloading point, and obstacle information. Driving trajectories are generated through path search, time window scheduling, or multi-vehicle priority avoidance. In multi-device collaborative operation scenarios, loaders or transport vehicles are typically treated as ordinary moving bodies. Path planning and conflict resolution are performed based on path length, travel time, empty / loaded status, or resource occupancy order to reduce congestion at intersections, narrow passages, waiting areas, and unloading areas.

[0003] However, the aforementioned methods typically only focus on vehicle center trajectory, road resource occupation, or time conflicts, making it difficult to accurately represent the posture constraints of loaders during material picking, shoveling, heavy-load transfer, unloading alignment, and reversing to avoid obstacles. In particular, loaders have special operational characteristics such as articulated outward swing, bucket forward extension, bucket lifting, heavy-load safety expansion, and end posture recovery. Existing methods are prone to problems such as unexecutable material picking posture, interrupted unloading alignment, frequent yielding to heavy-load vehicles, and the inability of subsequent vehicles to reuse the verified passage conditions of preceding vehicles, affecting the efficiency of continuous multi-loader operations and the stability of path execution.

[0004] Therefore, how to provide a loader operation path optimization method based on a multi-agent system is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0005] One objective of this invention is to propose a loader operation path optimization method based on a multi-agent system. This invention utilizes end-position anchor points and the bucket coupling occupancy zone to optimize the loader's collaborative path, which has the advantages of accurate posture matching, stable conflict transfer, and strong operation continuity.

[0006] A loader operation path optimization method based on a multi-agent system according to an embodiment of the present invention includes the following steps:

[0007] The system acquires the operating status of multiple loaders, material handling targets, unloading targets, and resource status of the work area, and constructs a loader operation resource map with each loader as a loader agent.

[0008] Based on the loader's operational resource map and loader's operating status, attitude constraint transformations are performed on the material handling target and the unloading target respectively, generating the material handling end attitude anchor point set and the unloading end attitude anchor point set;

[0009] Based on the set of end posture anchor points for material handling and the set of end posture anchor points for unloading, generate the operation phase chain for each loader agent, and generate the set of articulated bucket coupling occupancy zones based on the operation phase chain;

[0010] Extract the resource end posture accommodation window set from the loader operation resource map, perform interlocking matching with the articulated bucket occupancy zone set, and generate the end posture accommodation path set and the interlocking failure reason set;

[0011] Based on the set of end pose accommodation paths and the set of embedding failure reasons, end pose occupancy conflicts between multiple loaders are identified, an end pose interruption cost conflict map is constructed and conflict transfer is performed, and conflict transfer results are generated.

[0012] Identify the end-position inheritance passage window formed after the preceding loader passes through the resource node, and generate multi-loader collaborative operation path results based on the end-position accommodate path set, conflict transfer results, and end-position inheritance passage window.

[0013] Optionally, the generation of the loader operation resource map specifically includes:

[0014] The operating status of multiple loaders is aligned according to a unified operating time, and a set of loader status frames is formed based on the position, heading, speed, articulation angle, bucket height, bucket angle, load weight, and driving gear of each loader.

[0015] The material handling operation target is converted into material handling resource nodes, and the unloading operation target is converted into unloading resource nodes, and then associated with the corresponding loader agent to form a set of operation target nodes;

[0016] Based on the resource status of the work area, the areas where roads, intersections, narrow passages, waiting spaces, yielding spaces, and dynamic obstacles are located are divided into resource nodes to form a set of work resource nodes;

[0017] Based on the spatial connectivity, travel direction, travel boundary and adjacent occupation relationship between each resource node in the task resource node set, resource connection edges are constructed to form the task resource topology skeleton;

[0018] The loader status frame set, the target node set, the resource node set, and the resource topology skeleton are bound together. For each loader agent, a current location node, target node, reachable resource node, and prohibited resource node are established to form a loader operation resource map.

[0019] Optionally, the generation of the set of anchor points for the material handling end and the set of anchor points for the unloading end specifically includes:

[0020] Perform material handling posture constraint transformation on the material handling resource node in the loader operation resource map to generate a candidate set of material handling posture constraints;

[0021] Perform unloading posture constraint transformation on the unloading resource node in the loader operation resource map to generate a candidate set of unloading posture constraints;

[0022] Perform end-point reachability verification on the candidate sets of material picking posture constraints and unloading posture constraints according to reachable resource nodes and forbidden resource nodes, and generate end-point anchor point screening results;

[0023] The candidate material handling posture constraints and candidate unloading posture constraints corresponding to the same loader agent in the end posture anchor point screening results are merged and encapsulated to generate the end posture anchor point set for material handling operation and the end posture anchor point set for unloading operation.

[0024] Optionally, the generation of the hinge coupling occupancy zone set specifically includes:

[0025] According to the associated loader intelligent agent, the set of material picking end posture anchor points and the set of material unloading end posture anchor points are paired based on the same source to form a set of picking and unloading end posture anchor point pairs.

[0026] Based on the set of anchor points at the pick-up and unloading ends, a set of operation phase chains is constructed, including the following sequentially connected phases: unloaded approach phase, shovel entry posture locking phase, pick-up and release phase, heavy load transfer phase, unloading alignment locking phase, unloading exit phase, and unloaded return phase.

[0027] Perform attitude handover calibration on adjacent operation phases in the operation phase chain set to form a phase connection constraint set;

[0028] Based on the phase connection constraint set and the loader's operating status, the occupied area of ​​each operation phase is expanded to generate a phase segmented occupied zone set;

[0029] The phase segment occupancy zone set is continuously spliced ​​and overlapped and merged according to the operation phase chain sequence to generate the articulated bucket coupled occupancy zone set, which is then bound to the unloading end posture anchor point set and the loader intelligent agent.

[0030] Optionally, the occupancy area expansion generates a front frame sweep area, a rear frame sweep area, an articulated swing area, a bucket forward extension area, a bucket lifting area, a heavy-load safety expansion area, and a reversing rearward protection area for each operating phase. The unloaded approach phase and the unloaded return phase retain the front frame sweep area, the rear frame sweep area, and the articulated swing area. The entry posture locking phase adds a bucket forward extension area, the heavy-load transfer phase adds a heavy-load safety expansion area, the unloading alignment locking phase adds a bucket lifting area, and the operating phase with a reversing gear adds a reversing rearward protection area. The occupancy area corresponding to each operating phase is written into the phase segment occupancy zone set.

[0031] Optionally, the generation of the end pose accommodating path set and the fitting failure reason set specifically includes:

[0032] Based on the node type of each resource node in the loader operation resource map, the passage boundary, passage direction and resource connection edge connected to the current resource node are extracted. The occupancy status of each resource node is determined according to the resource status of the operation area, the loader status frame set, the area where dynamic obstacles are located and the prohibited resource nodes, forming a basic set of resource accommodation.

[0033] Map the resource containment base set to the containment boundary of the material handling operation end pose anchor point set, the material unloading operation end pose anchor point set, and the articulated bucket coupling occupancy zone set, respectively, to generate the resource end pose containment window set.

[0034] The set of articulated coupling occupancy zones is split sequentially according to the associated loader agent, operation phase chain, and resource nodes to form the set of end pose fitting requests;

[0035] Perform node-by-node pose fitting matching between the pose fitting request set and the resource pose accommodating window set to generate pose fitting matching results;

[0036] For end pose chiming requests that fail node-by-node chiming matching in the end pose chiming matching results, failure reasons are attributed and a set of chiming failure reasons is generated.

[0037] The end pose fitting requests that are matched node by node in the end pose fitting results are concatenated according to the operation phase chain order to generate an end pose accommodating path set.

[0038] Optionally, the generation of the conflict transfer result specifically includes:

[0039] Resource usage is compared between the end pose accommodation path set and the set of embedding failure reasons to identify end pose occupancy conflicts between multiple loaders and generate an end pose occupancy conflict set.

[0040] Take the conflicting loader agents in the pose occupancy conflict set as graph nodes, establish conflict relationship edges between two loader agents with pose occupancy conflicts in the same resource node, and construct the pose interruption cost conflict graph.

[0041] For each conflicting edge in the attitude interruption cost conflict graph, calculate the total attitude interruption cost of each conflicting loader agent and generate an attitude interruption cost table.

[0042] According to the end pose interruption cost table, the loader agents on the same conflict relationship edge are judged to determine the transfer object, and a set of transfer objects is generated, including the loader agent to which the transfer is made and the loader agent to which the transfer is made.

[0043] Generate the transfer action result based on the set of transfer objects and the pose interruption cost conflict graph;

[0044] The process involves encapsulating the transfer of the loader agent, the retention of the loader agent, the conflicting resource nodes, the type of pose occupancy conflict, the total cost of pose interruption, the transfer action result, the adjusted plan entry time, and the adjusted plan exit time to generate the conflict transfer result.

[0045] Optionally, the transfer object determination is made among the conflicting loader agents corresponding to the same conflict relationship edge. The loader agent with the smaller total cost of end posture interruption and the existence of a connectable preceding resource node or a connectable following resource node is selected as the transfer loader agent, and the other loader agent is selected as the holding loader agent. When the conflicting loader agent is in the working phase corresponding to the shovel attitude locking phase, the unloading alignment locking phase, or the reverse driving gear, the loader agent is excluded from the transfer loader agent candidate.

[0046] Optionally, the generation of the multi-loader collaborative operation path result specifically includes:

[0047] Extract the loader agent that has passed the resource node from the end pose accommodation path set and conflict transfer results, and generate the previous pass record set;

[0048] Based on the previous record set, construct an end-pose inheritance passage window set for the resource nodes that have been passed;

[0049] Extract loader agents from the end pose accommodating path set that plan to pass through the same window resource node and whose planned entry time is later than the window start time, and generate a subsequent set of chimerism candidates.

[0050] The subsequent set of chimerism candidates is matched with the set of end pose inheritance passage windows to generate end pose inheritance matching results.

[0051] Based on the end pose inheritance matching results, the end pose accommodation path of the subsequent loader agent is adjusted for window reuse, and the end pose inheritance passage window reuse results are generated.

[0052] The end pose containment path set, conflict transfer results, and end pose inheritance passage window reuse results are merged to generate multi-loader collaborative operation path results.

[0053] The beneficial effects of this invention are:

[0054] This invention transforms the objectives of material handling and unloading operations into sets of end-position anchor points for material handling and unloading operations, respectively. This allows the loader's operating path to no longer be planned solely based on spatial location or road nodes, but rather on attitude constraints that enable material handling, shovel entry, unloading alignment, and exit actions. Through an operational phase chain, it continuously expresses unloaded approach, shovel entry attitude locking, material handling separation, heavy-load transfer, unloading alignment locking, unloading exit, and unloaded return. Furthermore, it generates a set of articulated bucket coupling occupancy zones, incorporating the front frame sweep area, rear frame sweep area, articulated swing area, bucket forward extension area, bucket lifting area, heavy-load safety extension area, and reversing rearward protection area into the path optimization process. This improves the consistency between the loader's operating path and its actual operating posture, load status, and spatial occupancy range.

[0055] This invention, through the interlocking and matching of the resource end-position accommodating window set and the articulated bucket coupling occupancy zone set, can simultaneously determine the passage boundary, accommodating time period, material retrieval accommodating boundary, unloading accommodating boundary, occupancy area accommodating boundary, and occupancy status of resource nodes during the path generation stage. This reduces material retrieval space conflicts, unloading space conflicts, articulated outward swing conflicts, and heavy-load outward expansion conflicts caused by planning solely based on the path centerline or ordinary time windows. The end-position occupancy conflicts between multiple loaders are expressed through an end-position interruption cost conflict diagram, and conflict relinquishment is performed based on end-position reconstruction cost, heavy-load braking cost, articulated return-to-center cost, and cycle time blocking cost. This reduces the probability of unreasonable interruptions to the shovel attitude locking phase and unloading alignment locking phase, and reduces the operation cycle time loss caused by heavy-load braking, repeated alignment, and avoidance waiting. Furthermore, the end posture inheritance passage window can reuse the entry direction, exit direction, inherited posture boundary, and inherited occupied area boundary formed when the preceding loader has passed through the resource node for subsequent loader paths, making the path connection of continuous material picking, continuous material unloading, and unidirectional passage through intersections or narrow passages more compact, thereby improving the safety, continuity, and execution efficiency of multi-loader collaborative operation paths. Attached Figure Description

[0056] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0057] Figure 1 This is a flowchart of a loader operation path optimization method based on a multi-agent system proposed in this invention;

[0058] Figure 2 This is a flowchart illustrating the generation of the work phase chain and the bucket coupling occupancy zone in a loader operation path optimization method based on a multi-agent system proposed in this invention.

[0059] Figure 3This is a flowchart illustrating the generation of the work phase chain and the bucket coupling occupancy zone in a loader operation path optimization method based on a multi-agent system proposed in this invention. Detailed Implementation

[0060] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, illustrating only the basic structure of the invention, and therefore only show the components relevant to the invention.

[0061] refer to Figures 1-3 A method for optimizing the operation path of a loader based on a multi-agent system includes the following steps:

[0062] The system acquires the operating status of multiple loaders, material handling targets, unloading targets, and resource status of the work area, and constructs a loader operation resource map with each loader as a loader agent.

[0063] Based on the loader's operational resource map and loader's operating status, attitude constraint transformations are performed on the material handling target and the unloading target respectively, generating the material handling end attitude anchor point set and the unloading end attitude anchor point set;

[0064] Based on the set of end posture anchor points for material handling and the set of end posture anchor points for unloading, generate the operation phase chain for each loader agent, and generate the set of articulated bucket coupling occupancy zones based on the operation phase chain;

[0065] Extract the resource end posture accommodation window set from the loader operation resource map, perform interlocking matching with the articulated bucket occupancy zone set, and generate the end posture accommodation path set and the interlocking failure reason set;

[0066] Based on the set of end pose accommodation paths and the set of embedding failure reasons, end pose occupancy conflicts between multiple loaders are identified, an end pose interruption cost conflict map is constructed and conflict transfer is performed, and conflict transfer results are generated.

[0067] Identify the end-position inheritance passage window formed after the preceding loader passes through the resource node, and generate multi-loader collaborative operation path results based on the end-position accommodate path set, conflict transfer results, and end-position inheritance passage window.

[0068] In this embodiment, the generation of the loader operation resource map specifically includes:

[0069] The operating status of multiple loaders is aligned according to a unified operating time, and a set of loader status frames is formed based on the position, heading, speed, articulation angle, bucket height, bucket angle, load weight, and driving gear of each loader.

[0070] The material handling operation target is converted into material handling resource nodes, and the unloading operation target is converted into unloading resource nodes, and then associated with the corresponding loader agent to form a set of operation target nodes;

[0071] Based on the resource status of the work area, the areas where roads, intersections, narrow passages, waiting spaces, yielding spaces, and dynamic obstacles are located are divided into resource nodes to form a set of work resource nodes;

[0072] Based on the spatial connectivity, travel direction, travel boundary and adjacent occupation relationship between each resource node in the task resource node set, resource connection edges are constructed to form the task resource topology skeleton;

[0073] The loader status frame set, the target node set, the resource node set, and the resource topology skeleton are bound together. For each loader agent, a current location node, target node, reachable resource node, and prohibited resource node are established to form a loader operation resource map.

[0074] In this embodiment, the generation of the set of anchor points for the material handling end and the set of anchor points for the material unloading end specifically includes:

[0075] Perform material handling posture constraint transformation on the material handling resource node in the loader operation resource map to generate a candidate set of material handling posture constraints;

[0076] The material handling posture constraint transformation takes the passage boundary of the material handling resource node, the adjacent resource connection edge, and the current position node of the corresponding loader agent as the constraint source to determine the material handling entry direction, material handling stopping position, material handling hinge angle range, material handling bucket height range, material handling bucket angle range, and material handling exit direction when the loader enters the material handling resource node. It is then bound to the corresponding loader agent to form a candidate set of material handling posture constraints.

[0077] Perform unloading posture constraint transformation on the unloading resource node in the loader operation resource map to generate a candidate set of unloading posture constraints;

[0078] The unloading posture constraint transformation takes the passage boundary of the unloading resource node, the adjacent resource connection edge and the target node of the corresponding loader agent as the constraint source to determine the unloading entry direction, unloading stopping position, unloading articulation angle range, unloading bucket height range, unloading bucket angle range and unloading exit direction when the loader enters the unloading resource node, and binds them with the corresponding loader agent to form a candidate set of unloading posture constraints.

[0079] Perform end-point reachability verification on the candidate sets of material picking posture constraints and unloading posture constraints according to reachable resource nodes and forbidden resource nodes, and generate end-point anchor point screening results;

[0080] The end pose reachability verification matches the candidate set of material picking posture constraints and the candidate set of material unloading posture constraints with the current position node, reachable resource node and prohibited resource node of the corresponding loader agent, respectively. It retains the candidate set of material picking posture constraints and the candidate set of material unloading posture constraints that can be reached from the current position node through the reachable resource node without passing through the prohibited resource node. The retained candidate set of material picking posture constraints and the candidate set of material unloading posture constraints are written into the end pose anchor point filtering results.

[0081] The candidate material handling posture constraints and candidate unloading posture constraints corresponding to the same loader agent in the end posture anchor point screening results are merged and encapsulated to generate the end posture anchor point set for material handling operation and the end posture anchor point set for unloading operation.

[0082] The material handling end posture anchor point consists of the material handling resource node, material handling entry direction, material handling stopping position, material handling hinge angle range, material handling bucket height range, material handling bucket angle range, material handling exit direction, and associated loader intelligent body. The unloading end posture anchor point consists of the unloading resource node, unloading entry direction, unloading stopping position, unloading hinge angle range, unloading bucket height range, unloading bucket angle range, unloading exit direction, and associated loader intelligent body.

[0083] In this embodiment, the generation of the hinge coupling occupancy zone set specifically includes:

[0084] According to the associated loader intelligent agent, the set of material picking end posture anchor points and the set of material unloading end posture anchor points are paired based on the same source to form a set of picking and unloading end posture anchor point pairs.

[0085] The same-source pairing performs consistency matching between the associated loader agents in the material picking operation end posture anchor points and the unloading operation end posture anchor points. The material picking operation end posture anchor points and the unloading operation end posture anchor points corresponding to the same loader agent are combined into a picking and unloading end posture anchor point pair. The material picking entry direction, material picking stopping position, material picking exit direction, unloading entry direction, unloading stopping position, and unloading exit direction are retained to form a set of picking and unloading end posture anchor point pairs that limit a single material picking to unloading operation cycle.

[0086] Based on the set of anchor points at the pick-up and unloading ends, a set of operation phase chains is constructed, including the following sequentially connected phases: unloaded approach phase, shovel entry posture locking phase, pick-up and release phase, heavy load transfer phase, unloading alignment locking phase, unloading exit phase, and unloaded return phase.

[0087] The phase is defined by the material entry direction and material parking position to determine the termination posture of the no-load approach phase; the material entry posture locking phase is defined by the material articulation angle range, material bucket height range, and material bucket angle range; the material exit direction to determine the material exit phase; the material unloading entry direction to determine the termination direction of the heavy-load transfer phase; the material unloading parking position, material unloading articulation angle range, material unloading bucket height range, and material unloading bucket angle range to determine the material unloading alignment locking phase; and the material unloading exit direction to determine the material unloading exit phase. The material unloading exit phase is connected with the next round of no-load return phase to form a set of operation phase chains for each loader agent.

[0088] Perform attitude handover calibration on adjacent operation phases in the operation phase chain set to form a phase connection constraint set;

[0089] The attitude transition calibration uses the termination direction, termination position, termination hinge angle, termination bucket height, termination bucket angle, and termination load state of the previous operation phase as the starting constraints for the next operation phase. It generates phase connection constraints between the entry attitude locking phase and the material removal phase, between the heavy load transfer phase and the unloading alignment locking phase, and between the unloading exit phase and the empty load return phase.

[0090] Based on the phase connection constraint set and the loader's operating status, the occupied area of ​​each operation phase is expanded to generate a phase segmented occupied zone set;

[0091] The phase segment occupancy zone set is continuously spliced ​​and overlapped and merged according to the operation phase chain sequence to generate the articulated bucket coupled occupancy zone set, which is then bound to the unloading end posture anchor point set and the loader intelligent agent;

[0092] The continuous splicing and overlapping are combined with the phase connection constraint set as the splicing boundary. The phase segment occupancy zone of the same loader agent is connected in the order of no-load approach phase, shovel attitude locking phase, material picking and unloading phase, heavy-load transfer phase, unloading alignment locking phase, unloading exit phase and no-load return phase. The front frame sweep area, rear frame sweep area, articulated swing area, bucket forward extension area, bucket lifting area, heavy-load safety expansion area and reversing rear protection area that overlap in the same resource node of adjacent operation phases are merged to obtain the articulated bucket coupling occupancy zone set corresponding to each loader agent.

[0093] In this embodiment, the occupied area expansion generates a front frame sweep area, a rear frame sweep area, an articulated swing area, a bucket forward extension area, a bucket lifting area, a heavy-load safety expansion area, and a reversing rear protection area for each working phase. The empty approach phase and the empty return phase retain the front frame sweep area, the rear frame sweep area, and the articulated swing area. The entry posture locking phase adds a bucket forward extension area, the heavy-load transfer phase adds a heavy-load safety expansion area, the unloading alignment locking phase adds a bucket lifting area, and the working phase with a reversing gear adds a reversing rear protection area. The occupied area corresponding to each working phase is written into the phase segment occupied zone set.

[0094] The front frame sweep area is the ground projection area formed by the continuous movement of the outer contour of the loader's front frame along the planned travel trajectory within the corresponding working phase. The rear frame sweep area is the ground projection area formed by the continuous movement of the outer contour of the loader's rear frame along the planned travel trajectory within the corresponding working phase. The articulated outward swing area is the area occupied by the folding and outward expansion between the front and rear frame sweep areas due to changes in the articulation angle. The bucket forward extension area is the ground projection area located in front of the front frame during the loader's bucket in the entry attitude locking phase, material removal phase, and unloaded approach phase. The bucket lifting zone is the projected boundary on the ground of the front upper working area occupied by the loader during the unloading alignment locking phase and the unloading exit phase due to the lifting and tilting of the bucket. The heavy-load safety expansion zone is the safety occupied area added outside the front frame sweeping zone, rear frame sweeping zone, articulated swing zone, bucket extension zone and bucket lifting zone when the loader is in the heavy-load transfer phase and the unloading alignment locking phase. The reversing rear protection zone is the rear protection occupied area formed at the rear end of the rear frame when the loader is in the reverse driving gear or reversing to avoid an obstacle.

[0095] In this embodiment, the generation of the end pose accommodating path set and the set of fitting failure reasons specifically includes:

[0096] Based on the node type of each resource node in the loader operation resource map, the passage boundary, passage direction and resource connection edge connected to the current resource node are extracted. The occupancy status of each resource node is determined according to the resource status of the operation area, the loader status frame set, the area where dynamic obstacles are located and the prohibited resource nodes, forming a basic set of resource accommodation.

[0097] The occupied state refers to the state in which a resource node is occupied by a loader agent, a dynamic obstacle area, or a prohibited resource node at a unified operation time.

[0098] Map the resource containment base set to the containment boundary of the material handling operation end pose anchor point set, the material unloading operation end pose anchor point set, and the articulated bucket coupling occupancy zone set, respectively, to generate the resource end pose containment window set.

[0099] The boundary mapping uses the access boundary and access direction of the resource node as the source of the window boundary, the material entry direction, material stopping position, material hinge angle range, material bucket height range, material bucket angle range, and material exit direction in the material picking operation end posture anchor point as the source of the material picking end posture, the material entry direction, material stopping position, material hinge angle range, material bucket height range, material bucket angle range, and material exit direction in the unloading operation end posture anchor point as the source of the unloading end posture, and the material entry direction, material stopping position, material hinge angle range, material bucket height range, material bucket angle range, and material exit direction in the articulated bucket coupling occupancy zone set as the source of the occupancy area, forming a resource end posture occupancy window set consisting of window resource node, occupancy time period, occupancy access direction, material picking occupancy boundary, unloading occupancy boundary, occupancy area occupancy boundary, preceding connected resource node, following connected resource node, and occupancy status;

[0100] The set of articulated coupling occupancy zones is split sequentially according to the associated loader agent, operation phase chain, and resource nodes to form the set of end pose fitting requests;

[0101] Each end posture fitting request in the end posture fitting request set consists of the associated loader agent, the current operation phase, the resource node to be entered, the planned entry time, the planned exit time, the entry direction, the exit direction, the front frame sweep area, the rear frame sweep area, the articulated swing area, the bucket forward extension area, the bucket lifting area, the heavy load safety expansion area, and the reversing rear protection area, and is bound to the corresponding unloading end posture anchor point pair;

[0102] Perform node-by-node pose fitting matching between the pose fitting request set and the resource pose accommodating window set to generate pose fitting matching results;

[0103] The node-by-node splicing matching process involves: matching the resource node to be entered in the end-position splicing request with the window resource node in the resource end-position accommodating window set; matching the planned entry time and planned exit time with the accommodating time period; matching the entry direction and exit direction with the accommodating passage direction; matching the articulated swing area with the boundary defined by the material picking accommodating boundary or unloading accommodating boundary; matching the bucket forward extension area with the boundary defined by the material picking bucket height range, material picking bucket angle range, and passage boundary in the material picking accommodating boundary; matching the bucket lifting area with the boundary defined by the unloading bucket height range, unloading bucket angle range, and passage boundary in the unloading accommodating boundary; and matching the front frame sweep area, rear frame sweep area, heavy-load safety expansion area, and reversing rear protection area with the occupancy area accommodating boundary and occupancy status. When all matches are successful, the corresponding end-position splicing request is written into the end-position splicing matching result.

[0104] For end pose chiming requests that fail node-by-node chiming matching in the end pose chiming matching results, failure reasons are attributed and a set of chiming failure reasons is generated.

[0105] The failure attribution is determined by the failed matching items, including time period matching failure, direction matching failure, attitude matching failure, material picking space matching failure, unloading space matching failure, occupied area matching failure, occupied state matching failure, and connection failure of preceding and following resources. Each matching failure reason is bound to the corresponding loader agent, current operation phase, resource node to be entered, and picking and unloading end attitude anchor point.

[0106] The end pose fitting requests that are matched node by node in the end pose fitting results are concatenated according to the operation phase chain order to generate an end pose accommodating path set.

[0107] In this embodiment, the generation of the conflict transfer result specifically includes:

[0108] Resource usage is compared between the end pose accommodation path set and the set of embedding failure reasons to identify end pose occupancy conflicts between multiple loaders and generate an end pose occupancy conflict set.

[0109] The resource occupancy comparison uses the resource nodes to be entered, planned entry time, planned exit time, entry direction, exit direction, and operation phase chain in the end posture accommodating path set as the comparison objects. It combines the posture occupancy failure, material picking space occupancy failure, material unloading space occupancy failure, occupied area occupancy failure, occupied state occupancy failure, and failure of connection between front and rear resources in the occupancy failure reason set to determine the end posture occupancy conflict type, including shovel end posture conflict, unloading end posture conflict, articulated outward swing conflict, heavy load outward expansion conflict, reversing protection conflict, intersection occupancy conflict, narrow passage squeezing conflict, and phase continuity break conflict.

[0110] Take the conflicting loader agents in the pose occupancy conflict set as graph nodes, establish conflict relationship edges between two loader agents with pose occupancy conflicts in the same resource node, and construct the pose interruption cost conflict graph.

[0111] The conflict relationship edge records the conflict resource node, conflict operation phase, conflict start time, conflict end time, end pose occupancy conflict type, corresponding chiseling failure reason, involved take-off and unload end pose anchor point pairs, and involved hinge coupling occupancy zone.

[0112] For each conflicting edge in the attitude interruption cost conflict graph, calculate the total attitude interruption cost of each conflicting loader agent and generate an attitude interruption cost table.

[0113] The total cost of attitude interruption is obtained by adding the attitude reconstruction cost, heavy-load braking cost, articulation return cost, and cycle blockage cost.

[0114] The end-position reconstruction cost is the attitude recovery time required for the transferred loader agent to regain the end-position anchor point for material handling or unloading operations from the interrupted position. The heavy-load braking cost is the speed recovery time required for the transferred loader agent to decrease from the current speed to the waiting speed and then recover to the planned speed before the transfer in the heavy-load transfer phase or unloading alignment lock phase. The articulation return cost is the articulation angle adjustment time required for the transferred loader agent to re-enter the material handling articulation angle range or unloading articulation angle range from the current articulation angle. The cycle blockage cost is the time that the transfer action causes the material handling resource node, unloading resource node, waiting position, or avoidance position to be occupied for an extended period. The end-position interruption cost table records the end-position reconstruction cost, heavy-load braking cost, articulation return cost, cycle blockage cost, and total end-position interruption cost corresponding to each conflicting loader agent.

[0115] According to the end pose interruption cost table, the loader agents on the same conflict relationship edge are judged to determine the transfer object, and a set of transfer objects is generated, including the loader agent to which the transfer is made and the loader agent to which the transfer is made.

[0116] Generate the transfer action result based on the set of transfer objects and the pose interruption cost conflict graph;

[0117] The transfer action results are determined according to the type of end posture occupancy conflict and the reason for the coupling failure. These actions include waiting for the preceding connecting resource node, switching to the subsequent connecting resource node, entering the avoidance position, delaying the planned entry time, reselecting the resource end posture accommodating window, and short-distance detour. For the shovel end posture conflict and the unloading end posture conflict, the corresponding material picking end posture anchor point or unloading end posture anchor point of the loader intelligent body is kept unchanged. For the articulated outward swing conflict, heavy load outward expansion conflict, and reversing protection conflict, the loader intelligent body is transferred to a position that does not intrude into the articulated coupling occupancy zone of the loader intelligent body.

[0118] The process involves encapsulating the transfer of the loader agent, the retention of the loader agent, the conflicting resource nodes, the type of pose occupancy conflict, the total cost of pose interruption, the transfer action result, the adjusted plan entry time, and the adjusted plan exit time to generate the conflict transfer result.

[0119] In this embodiment, the loader agent with the smaller total cost of end posture interruption and the existence of a preceding or following resource node is selected as the loader agent to be transferred from the conflicting loader agents corresponding to the conflicting edge. The other loader agent is selected as the holding loader agent. When the conflicting loader agent is in the working phase corresponding to the shovel attitude locking phase, the unloading alignment locking phase, or the reverse driving gear, the loader agent is excluded from the candidate loader agents to be transferred.

[0120] In this embodiment, the generation of the multi-loader collaborative operation path result specifically includes:

[0121] Extract the loader agent that has passed the resource node from the end pose accommodation path set and conflict transfer results, and generate the previous pass record set;

[0122] The preceding pass record set consists of loader agents that have completed end pose fitting and matching in the corresponding resource node and whose planned departure time is earlier than the planned entry time of other loader agents. Each preceding pass record includes the preceding loader agent, the resource node passed, the pass operation phase, the entry direction, the exit direction, the planned entry time, the planned departure time, the pick-up and unload end pose anchor point pair, the hinge coupling occupancy zone, and the corresponding conflict transfer result.

[0123] Based on the previous record set, construct an end-pose inheritance passage window set for the resource nodes that have been passed;

[0124] The construction of the end posture inheritance window uses the passed resource nodes as window resource nodes, the planned departure time of the preceding loader agent as the window start basis, the next occupancy state of the same window resource node as the window end basis, the entry and exit directions of the preceding loader agent as the inheritance direction source, the hinge angle interval, bucket height interval, and bucket angle interval in the corresponding unloading end posture anchor point pair of the preceding loader agent as the inheritance posture source, and the hinge coupling occupancy zone of the preceding loader agent as the inheritance occupancy area source. This forms an end posture inheritance passage window set consisting of window resource nodes, window start time, window end time, inheritance entry direction, inheritance exit direction, inheritance operation phase, inheritance posture boundary, inheritance occupancy area boundary, and associated preceding loader agent.

[0125] Extract loader agents from the end pose accommodating path set that plan to pass through the same window resource node and whose planned entry time is later than the window start time, and generate a subsequent set of chimerism candidates.

[0126] The subsequent loader agents in the subsequent embedding candidate set are those loader agents that have not yet passed the corresponding window resource node in the end posture accommodating path set, and whose end posture accommodating path needs to pass through the window resource node. Each subsequent embedding candidate is bound to the corresponding loader agent's pick-up and unload end posture anchor point pair, operation phase chain, end posture embedding request, planned entry time, planned exit time, and articulated coupling occupancy zone.

[0127] The subsequent set of chimerism candidates is matched with the set of end pose inheritance passage windows to generate end pose inheritance matching results.

[0128] The inheritance matching process involves: matching the resource nodes to be entered by subsequent chimera candidates with the window resource nodes in the end-pose inheritance access window set; matching the planned entry time of subsequent chimera candidates with the window start time and window end time; matching the entry and exit directions of subsequent chimera candidates with the inherited entry and exit directions; matching the current operation phase of subsequent chimera candidates with the inherited operation phase; matching the pick-up and unload end-pose anchor point pairs of subsequent chimera candidates with the inherited attitude boundary; matching the articulated coupling occupancy zone of subsequent chimera candidates with the inherited occupancy area boundary; and writing the subsequent chimera candidates that pass all the matching into the end-pose inheritance matching results.

[0129] Based on the end pose inheritance matching results, the end pose accommodation path of the subsequent loader agent is adjusted for window reuse, and the end pose inheritance passage window reuse results are generated.

[0130] Among them, the window reuse adjustment will adjust the planned entry time of the subsequent loader agent on the window resource node to the end posture inheritance passage window, and bind the entry direction, exit direction, operation phase and articulated coupling occupancy zone of the corresponding end posture embedding request to the end posture inheritance passage window. For subsequent embedding candidates that fail to pass the inheritance matching, their original end posture accommodation path and conflict transfer result will be maintained.

[0131] The end pose containment path set, conflict transfer results, and end pose inheritance passage window reuse results are merged to generate multi-loader collaborative operation path results;

[0132] The results of the multi-loader collaborative operation path include the pick-up and unloading end posture anchor point pairs of each loader agent, operation phase chain, resource node passage order, articulated bucket coupling occupancy zone, planned entry time, planned departure time, transfer action results, end posture inheritance passage window, and reused end posture containment path.

[0133] Example 1: To verify the feasibility of this invention in practice, it was applied to a loader collaborative operation scenario at a sand and gravel aggregate stockpile. The stockpile includes a material collection pile, a transport vehicle parking area, an unloading port, intersections, narrow passages, waiting areas, and yielding areas. Multiple loaders operate within the same work area, performing material collection, heavy-load transfer, unloading, and empty return trips. The existing on-site operation mainly relied on fixed routes and manual experience for yielding. When transport vehicles arrived in concentrated numbers, the unloading port experienced short-term congestion, or oncoming traffic occurred in narrow passages, issues arose such as loaders repeatedly adjusting their posture before entering the shovel, heavy-load vehicles waiting at intersections, interrupted unloading alignment, and empty vehicles occupying the passage space for heavy-load vehicles, leading to unstable work rhythms.

[0134] In this scenario, the location, heading, speed, articulation angle, bucket height, bucket angle, load weight, and driving gear of each loader are first collected. This data is then combined with information on material pile boundaries, unloading port locations, transport vehicle parking areas, road boundaries, intersections, narrow passages, waiting areas, avoidance areas, and areas containing dynamic obstacles to construct a loader operation resource map. For each loader, instead of directly using the pick-up and unloading points as ordinary path endpoints, the pick-up operation target is converted into a pick-up operation end posture anchor point, and the unloading operation target is converted into an unloading operation end posture anchor point. This ensures that the loader has a clear entry direction, parking position, articulation angle range, bucket height range, bucket angle range, and exit direction when entering pick-up and unloading resource nodes. In this way, the path generation stage can preemptively eliminate paths that only reach the location but cannot complete the loading or unloading alignment.

[0135] During operation, based on the material handling end posture anchor points and unloading end posture anchor points, the following phases are generated: no-load approach phase, shovel entry posture locking phase, material handling separation phase, heavy-load transfer phase, unloading alignment locking phase, unloading exit phase, and no-load return phase. A bucket coupling occupancy zone is then generated according to the phase connection relationship. This occupancy zone includes not only the front and rear frame sweep areas, but also the articulated outward swing area, bucket forward extension area, bucket lifting area, heavy-load safety expansion area, and reversing rearward protection area. Therefore, when the loader passes through intersections, enters material piles, approaches the unloading port, or reverses to avoid obstacles, the occupancy range is no longer simplified to the vehicle centerline, but is consistent with the bucket posture, articulated state, and load state.

[0136] The resource end-position accommodating window is formed by the passage boundaries, passage directions, occupancy status, and end-position accommodating boundaries of roads, intersections, narrow passages, material picking resource nodes, and unloading resource nodes. Before the loader's bucket coupling occupancy zone enters a resource node, it will perform a matching process with the resource end-position accommodating window. The matching content includes the entry time, entry direction, exit direction, material picking space, unloading space, occupancy area, and occupancy status. If the matching fails, a reason for the matching failure is formed, which helps to distinguish whether it is due to the bucket extending beyond the boundary, insufficient unloading lifting space, heavy-load outward expansion occupancy conflict, or inability to connect resources in front and behind. When multiple loaders apply for the same intersection, narrow passage, or unloading port at the same time, the conflict transfer object is determined based on the end-position interruption cost conflict diagram, so that the entry attitude locking phase, unloading alignment locking phase, and heavy-load transfer phase are given priority protection, and loaders returning empty or with unlocked end positions enter the waiting position or the avoidance position.

[0137] To demonstrate the application's effectiveness, during continuous production shifts at an aggregate stockpile in East China, logs were maintained on-site, including loader trajectory logs, bucket attitude records, articulation angle records, resource node occupancy records, conflict transfer records, and end-position inheritance passage window reuse records. The operation logs showed that the loader's attitude adjustment process before entering the stockpile was more concentrated, unloading port alignment interruptions were reduced, temporary stops of heavily loaded loaders in narrow passages and intersections were suppressed, and unloaded loaders completed transfers more frequently through waiting or yielding positions. The end-position inheritance passage window formed after the preceding loader passes through material picking resource nodes, unloading resource nodes, or intersections can be reused by subsequent loaders that meet the direction, phase, and occupancy zone conditions, making the connection between continuous material picking, continuous unloading, and same-direction passages more compact. The aforementioned operation logs and on-site scheduling records indicate that this method can improve the attitude executability, conflict handling stability, and continuous operation efficiency of multi-loader collaborative operation paths without altering the loader's mechanical structure.

[0138] Table 1. Comprehensive Comparison of Multi-Loader Collaborative Path Optimization under Attitude Constraints

[0139] Comparison Algorithm Hybrid A* Fixed Priority Path Planning Algorithm CBS Conflict Search Multi-Agent Path Planning Algorithm SIPP safety interval path planning algorithm MAPPO Multi-Agent Reinforcement Learning Path Planning Algorithm Method of the present invention Average single-cycle operation time (min) 7.82 7.66 7.58 7.49 7.21 Average waiting time (s / cycle) 51.6 48.3 44.7 46.1 39.2 Number of heavy-load temporary stops (times / 10 cycles) 5.4 4.9 4.5 4.7 3.8 Secondary adjustment rate of material handling end posture (%) 10.6 10.1 9.4 9.8 7.7 Unloading alignment interruption rate (%) 7.8 7.2 6.9 7.1 5.8 Time taken to handle conflicts at intersections and narrow passages (s / time) 18.9 17.4 16.1 16.8 14.3 Total effective cycle count for multiple loaders (cycles / hour) 31.4 32.1 32.6 32.4 34.0 Fuel consumption per effective cycle (L / cycle) 0.84 0.82 0.81 0.82 0.78

[0140] Table 1 uses the collaborative operation records of multiple loaders at the same sand and gravel aggregate stockpile as the statistical basis. The operation area includes the material picking pile, unloading port, intersection, narrow passage, waiting position, and avoidance position. Each algorithm performs path generation and conflict handling under the same number of loaders, the same arrival rules for transport vehicles, and the same picking and unloading operation objectives. As shown in Table 1, the HybridA* fixed priority path planning algorithm has an average single-cycle operation time of 7.82 min, an average waiting time of 51.6 s / cycle, and 5.4 times / 10 cycles for heavy-load temporary stops. This indicates that although the method can generate a basic passage path, it mainly relies on geometric paths and fixed priority avoidance, and does not adequately protect the continuity of material picking end posture, unloading alignment, and heavy-load transfer stages. The CBS conflict search multi-agent path planning algorithm reduces the average single-cycle operation time to 7.66 minutes and the conflict handling time to 17.4 seconds per cycle, indicating that the conflict search mechanism can reduce some resource contention. However, its conflict objects are still mainly nodes, edges, or time occupancy, and do not fully express the actual operation occupancy caused by the loader's articulated outward swing, bucket extension, and bucket lifting.

[0141] The SIPP safety interval path planning algorithm performs well in terms of average waiting time, with a value of 44.7s / cycle, which is better than the Hybrid A* fixed priority path planning algorithm and the CBS conflict search multi-agent path planning algorithm, indicating that the safety interval mechanism can improve the entry timing of channels and intersections. However, the SIPP safety interval path planning algorithm still has a secondary adjustment rate of 9.4% for the picking end posture and a 6.9% interruption rate for unloading alignment, indicating that it mainly solves the problem of time-based entry, rather than whether the loader can complete the operation at the picking and unloading resource nodes with appropriate articulation angle, bucket height, bucket angle, and exit direction. The MAPPO multi-agent reinforcement learning path planning algorithm has an average single-cycle operation time of 7.49min and a total effective number of cycles per hour for multiple loaders of 32.4 times / hour, showing a certain degree of collaborative decision-making ability. However, its secondary adjustment rate of picking end posture, interruption rate for unloading alignment, and number of heavy-load temporary stops do not reach the level of the method of this invention, indicating that it is difficult to stably lock the loader's working end posture and the articulation coupling occupancy boundary by simply relying on the avoidance action obtained by policy learning.

[0142] The average single-cycle operation time of the method of this invention is 7.21 min, which is 3.7% lower than the better-performing MAPPO multi-agent reinforcement learning path planning algorithm and 7.8% lower than the Hybrid A* fixed priority path planning algorithm; the average waiting time is 39.2 s / cycle, which is 12.3% lower than the SIPP safety interval path planning algorithm; and the number of heavy-load temporary stops is 3.8 times / 10 cycles, which is 15.6% lower than the SIPP safety interval path planning algorithm. These data show that the method of this invention does not simply shorten the path length, but rather locks the operation posture in advance by using the set of end posture anchor points for material picking and unloading operations. It maintains the continuous relationship between empty approach, shovel posture locking, material picking and unloading, heavy-load transfer, unloading alignment locking, unloading exit, and empty return trip through the operation phase chain, thereby reducing the number of times the loader is forced to stop, restart, and readjust the path under heavy load.

[0143] In terms of the secondary adjustment rate of the material handling end posture and the interruption rate of unloading alignment, the method of this invention is 7.7% and 5.8% respectively, both lower than the comparative algorithms. This result is directly related to the resource end posture accommodation window and the articulated bucket coupling occupancy zone matching mechanism of this invention. The resource end posture accommodation window not only determines whether the resource node is idle, but also incorporates the material handling accommodation boundary, unloading accommodation boundary, occupancy area accommodation boundary, and occupancy status into the judgment; the articulated bucket coupling occupancy zone incorporates the front frame sweep area, rear frame sweep area, articulated swing area, bucket forward extension area, bucket lifting area, heavy load safety expansion area, and reversing rearward protection area into the path object. Thus, before the loader enters the material handling area or unloading area, the matching at the posture and occupancy area level has been completed, reducing the need for secondary correction of the bucket posture, articulation angle, or unloading alignment posture after reaching the position.

[0144] Regarding the time consumption for handling conflicts at intersections and narrow passages, the method of this invention achieves 14.3 s / cycle, lower than the SIPP safety interval path planning algorithm's 16.1 s / cycle and the MAPPO multi-agent reinforcement learning path planning algorithm's 16.8 s / cycle. This improvement stems from the synergistic effect of the end-position interruption cost conflict map and the conflict relinquishment result. The end-position interruption cost conflict map not only identifies ordinary path overlaps but also distinguishes between shovel end-position conflicts, unloading end-position conflicts, articulated outward swing conflicts, heavy-load outward expansion conflicts, reversing protection conflicts, intersection occupancy conflicts, narrow passage squeezing conflicts, and phase continuity break conflicts. Based on end-position reconstruction costs, heavy-load braking costs, articulated return costs, and cycle blockage costs, it determines which loader agent to relinquish and which to maintain. Therefore, when a conflict occurs, loaders in the shovel end-position locking phase, unloading alignment locking phase, or heavy-load transfer phase can obtain more stable continuous passage protection, while loaders on empty return trips or with upstream and downstream resource connection conditions undertake actions such as waiting, yielding, or reselecting resource end-position accommodation windows.

[0145] The method of this invention achieves a total effective cycle count of 34.0 times / hour for multiple loaders, with a fuel consumption of 0.78L / cycle per effective cycle. The increased effective cycle count is related to the reuse mechanism of the end-position inheritance passage window. After a preceding loader passes a resource node, its entry direction, exit direction, inherited operation phase, inherited posture boundary, and inherited occupied area boundary are converted into an end-position inheritance passage window that can be reused by subsequent loaders. Subsequent loaders can directly enter the window for reuse and adjustment when the resource node to be entered, planned entry time, direction, operation phase, pick-up / unload end-position anchor point pair, and hinge coupling occupied area all meet the inheritance matching conditions. This mechanism reduces downtime during continuous material picking, continuous unloading, and unidirectional passage through intersections or narrow passages, making path connections more compact, while reducing fuel consumption caused by heavy-load waiting, repeated alignment, and ineffective detours.

[0146] The method of this invention, through continuous processing of the loading and unloading end posture anchor point, operation phase chain, articulated bucket coupling occupancy zone, resource end posture accommodation window, end posture interruption cost conflict map, and end posture inheritance passage window, elevates the loader operation path optimization from the ordinary vehicle center trajectory and time window avoidance to a path optimization process oriented towards material picking posture, unloading posture, articulated outward swing, bucket operation range, heavy load safety expansion, and multi-loader collaborative cycle time. It can reduce heavy load temporary stops, repeated end posture adjustments, unloading alignment interruptions, and resource waste while ensuring path executability. It has comprehensive advantages such as accurate operation posture matching, stable conflict transfer, high continuous operation efficiency, and low unit cycle energy consumption.

[0147] The above are merely preferred embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A method for optimizing the operation path of a loader based on a multi-agent system, characterized in that, Includes the following steps: The system acquires the operating status of multiple loaders, material handling targets, unloading targets, and resource status of the work area, and constructs a loader operation resource map with each loader as a loader agent. Based on the loader's operational resource map and loader's operating status, attitude constraint transformations are performed on the material handling target and the unloading target respectively, generating the material handling end attitude anchor point set and the unloading end attitude anchor point set; Based on the set of end posture anchor points for material handling and the set of end posture anchor points for unloading, generate the operation phase chain for each loader agent, and generate the set of articulated bucket coupling occupancy zones based on the operation phase chain; Extract the resource end posture accommodation window set from the loader operation resource map, perform interlocking matching with the articulated bucket occupancy zone set, and generate the end posture accommodation path set and the interlocking failure reason set; Based on the set of end pose accommodation paths and the set of embedding failure reasons, end pose occupancy conflicts between multiple loaders are identified, an end pose interruption cost conflict map is constructed and conflict transfer is performed, and conflict transfer results are generated. Identify the end-position inheritance passage window formed after the preceding loader passes through the resource node, and generate multi-loader collaborative operation path results based on the end-position accommodate path set, conflict transfer results, and end-position inheritance passage window.

2. The loader operation path optimization method based on a multi-agent system according to claim 1, characterized in that, The generation of the loader operation resource map specifically includes: The operating status of multiple loaders is aligned according to a unified operating time, and a set of loader status frames is formed based on the position, heading, speed, articulation angle, bucket height, bucket angle, load weight, and driving gear of each loader. The material handling operation target is converted into material handling resource nodes, and the unloading operation target is converted into unloading resource nodes, and then associated with the corresponding loader agent to form a set of operation target nodes; Based on the resource status of the work area, the areas where roads, intersections, narrow passages, waiting spaces, yielding spaces, and dynamic obstacles are located are divided into resource nodes to form a set of work resource nodes; Based on the spatial connectivity, travel direction, travel boundary and adjacent occupation relationship between each resource node in the task resource node set, resource connection edges are constructed to form the task resource topology skeleton; The loader status frame set, the target node set, the resource node set, and the resource topology skeleton are bound together. For each loader agent, a current location node, target node, reachable resource node, and prohibited resource node are established to form a loader operation resource map.

3. The loader operation path optimization method based on a multi-agent system according to claim 1, characterized in that, The generation of the set of end posture anchor points for material handling and the set of end posture anchor points for unloading specifically includes: Perform material handling posture constraint transformation on the material handling resource node in the loader operation resource map to generate a candidate set of material handling posture constraints; Perform unloading posture constraint transformation on the unloading resource node in the loader operation resource map to generate a candidate set of unloading posture constraints; Perform end-point reachability verification on the candidate sets of material picking posture constraints and unloading posture constraints according to reachable resource nodes and forbidden resource nodes, and generate end-point anchor point screening results; The candidate material handling posture constraints and candidate unloading posture constraints corresponding to the same loader agent in the end posture anchor point screening results are merged and encapsulated to generate the end posture anchor point set for material handling operation and the end posture anchor point set for unloading operation.

4. The loader operation path optimization method based on a multi-agent system according to claim 1, characterized in that, The generation of the hinge coupling occupancy band set specifically includes: According to the associated loader intelligent agent, the set of material picking end posture anchor points and the set of material unloading end posture anchor points are paired based on the same source to form a set of picking and unloading end posture anchor point pairs. Based on the set of anchor points at the pick-up and unloading ends, a set of operation phase chains is constructed, including the following sequentially connected phases: unloaded approach phase, shovel entry posture locking phase, pick-up and release phase, heavy load transfer phase, unloading alignment locking phase, unloading exit phase, and unloaded return phase. Perform attitude handover calibration on adjacent operation phases in the operation phase chain set to form a phase connection constraint set; Based on the phase connection constraint set and the loader's operating status, the occupied area of ​​each operation phase is expanded to generate a phase segmented occupied zone set; The phase segment occupancy zone set is continuously spliced ​​and overlapped and merged according to the operation phase chain sequence to generate the articulated bucket coupled occupancy zone set, which is then bound to the unloading end posture anchor point set and the loader intelligent agent.

5. The loader operation path optimization method based on a multi-agent system according to claim 4, characterized in that, The occupancy area expansion generates a front frame sweep area, a rear frame sweep area, an articulated swing area, a bucket forward extension area, a bucket lifting area, a heavy-load safety expansion area, and a reversing rear protection area for each operating phase. Among them, the no-load approach phase and the no-load return phase retain the front frame sweep area, the rear frame sweep area, and the articulated swing area. The entry posture locking phase adds a bucket forward extension area, the heavy-load transfer phase adds a heavy-load safety expansion area, the unloading alignment locking phase adds a bucket lifting area, and the operating phase with a reversing gear adds a reversing rear protection area. The occupancy area corresponding to each operating phase is written into the phase segment occupancy zone set.

6. The loader operation path optimization method based on a multi-agent system according to claim 1, characterized in that, The generation of the pose accommodation path set and the fitting failure reason set specifically includes: Based on the node type of each resource node in the loader operation resource map, the passage boundary, passage direction and resource connection edge connected to the current resource node are extracted. The occupancy status of each resource node is determined according to the resource status of the operation area, the loader status frame set, the area where dynamic obstacles are located and the prohibited resource nodes, forming a basic set of resource accommodation. Map the resource containment base set to the containment boundary of the material handling operation end pose anchor point set, the material unloading operation end pose anchor point set, and the articulated bucket coupling occupancy zone set, respectively, to generate the resource end pose containment window set. The set of articulated coupling occupancy zones is split sequentially according to the associated loader agent, operation phase chain, and resource nodes to form the set of end pose fitting requests; Perform node-by-node pose fitting matching between the pose fitting request set and the resource pose accommodating window set to generate pose fitting matching results; For end pose chiming requests that fail node-by-node chiming matching in the end pose chiming matching results, failure reasons are attributed and a set of chiming failure reasons is generated. The end pose fitting requests that are matched node by node in the end pose fitting results are concatenated according to the operation phase chain order to generate an end pose accommodating path set.

7. The loader operation path optimization method based on a multi-agent system according to claim 1, characterized in that, The generation of the conflict transfer result specifically includes: Resource usage is compared between the end pose accommodation path set and the set of embedding failure reasons to identify end pose occupancy conflicts between multiple loaders and generate an end pose occupancy conflict set. Take the conflicting loader agents in the pose occupancy conflict set as graph nodes, establish conflict relationship edges between two loader agents with pose occupancy conflicts in the same resource node, and construct the pose interruption cost conflict graph. For each conflicting edge in the attitude interruption cost conflict graph, calculate the total attitude interruption cost of each conflicting loader agent and generate an attitude interruption cost table. According to the end pose interruption cost table, the loader agents on the same conflict relationship edge are judged to determine the transfer object, and a set of transfer objects is generated, including the loader agent to which the transfer is made and the loader agent to which the transfer is made. Generate the transfer action result based on the set of transfer objects and the pose interruption cost conflict graph; The process involves encapsulating the transfer of the loader agent, the retention of the loader agent, the conflicting resource nodes, the type of pose occupancy conflict, the total cost of pose interruption, the transfer action result, the adjusted plan entry time, and the adjusted plan exit time to generate the conflict transfer result.

8. The loader operation path optimization method based on a multi-agent system according to claim 7, characterized in that, The transfer object is determined among the conflicting loader agents corresponding to the same conflict relationship edge. The loader agent with the smaller total cost of end posture interruption and the existence of a connectable preceding resource node or a connectable following resource node is selected as the transfer loader agent, and the other loader agent is selected as the holding loader agent. When the conflicting loader agent is in the working phase corresponding to the shovel attitude locking phase, the unloading alignment locking phase, or the reverse driving gear, the loader agent is excluded from the transfer loader agent candidate.

9. The method for optimizing the operation path of a loader based on a multi-agent system according to claim 1, characterized in that, The generation of the multi-loader collaborative operation path results specifically includes: Extract the loader agent that has passed the resource node from the end pose accommodation path set and conflict transfer results, and generate the previous pass record set; Based on the previous record set, construct an end-pose inheritance passage window set for the resource nodes that have been passed; Extract loader agents from the end pose accommodating path set that plan to pass through the same window resource node and whose planned entry time is later than the window start time, and generate a subsequent set of chimerism candidates. The subsequent set of chimerism candidates is matched with the set of end pose inheritance passage windows to generate end pose inheritance matching results. Based on the end pose inheritance matching results, the end pose accommodation path of the subsequent loader agent is adjusted for window reuse, and the end pose inheritance passage window reuse results are generated. The end pose containment path set, conflict transfer results, and end pose inheritance passage window reuse results are merged to generate multi-loader collaborative operation path results.

Images