A multi-intelligent logistics vehicle cooperative obstacle avoidance method and device in a warehouse scenario

Abstract

The application discloses a kind of warehouse scene under the coordination of multiple intelligent logistics vehicles and device, and relates to the technical field of intelligent warehousing, which comprises the following steps: obtaining warehouse scene map, operation task information and the current state and target state of each intelligent logistics vehicle;Determine the basic path of each intelligent logistics vehicle;Based on the curvature of basic path, vehicle body relative attitude, obstacle clearance, regional speed limit and tracking error margin distribution speed profile, generate time-stamped global trajectory based on the allocated speed profile;The time-stamped global trajectory of each intelligent logistics vehicle is mapped into a three-dimensional space-time occupancy model, and the global trajectory of each intelligent logistics vehicle is allocated based on the more three-dimensional space-time occupancy model by multiple iterations, until each intelligent logistics vehicle is reserved space-time occupancy to generate conflict-free coordinated obstacle avoidance trajectory of each intelligent logistics vehicle.

Description

Technical Field

[0001] This invention relates to the fields of intelligent warehousing, autonomous driving, multi-vehicle collaborative control and path planning, and in particular to a method and device for collaborative obstacle avoidance of multiple intelligent logistics vehicles in a warehousing scenario. Background Technology

[0002] In smart warehousing, unmanned transportation, and in-plant logistics scenarios, multiple smart logistics vehicles typically need to operate concurrently in shared lanes, intersections, temporary storage areas, and loading / unloading areas. These scenarios are characterized by limited road space, frequent traffic convergence, dense static obstacles, and tightly coupled task sequences. Without an effective collaborative obstacle avoidance mechanism, problems such as oncoming traffic congestion, intersection conflicts, deadlocks, and decreased traffic efficiency can easily occur.

[0003] In existing multi-vehicle path planning schemes, a common approach is to first plan a two-dimensional geometric path for each vehicle, and then resolve conflicts through simple delays, stopping to yield, or local obstacle avoidance. This method is applicable to small mobile robot scenarios, but in scenarios with larger vehicle sizes, limited turning radii, the presence of tractor-trailer combinations, or strict control of passage timing in warehouses, problems such as geometrically reachable but practically unexecutable paths, the introduction of new timing conflicts through local repairs, and multi-vehicle oscillations during yielding can easily arise.

[0004] For articulated autonomous trucks, unmanned tractors, and large logistics trailers, vehicle motion satisfies nonholonomic constraints, and there is a relative motion relationship between the tractor and the trailer. If only a two-dimensional point mass model or a single rigid body model is used in the planning stage, it is usually difficult to accurately characterize the dual-body occupancy of the tractor and trailer, the articulation angle constraint, and the risk of folding blades. This can lead to the planning path being difficult to execute directly, or to the generation of large tailsweep errors and collision risks during the tracking process.

[0005] Dynamic windowing, a common local obstacle avoidance method, can sample candidate control commands in the velocity space and evaluate short-term predicted trajectories, offering advantages in real-time performance. However, traditional dynamic windowing methods are typically geared towards single-vehicle local obstacle avoidance, focusing more on static or short-term dynamic obstacles, and lacking a unified utilization of the time reservation relationships between the global trajectories of multiple vehicles. In scenarios such as warehouse intersections, narrow passing points, or continuous intersecting traffic, relying solely on traditional dynamic windowing methods can easily lead to local optima, repeated yielding, and deadlocks.

[0006] While existing multi-vehicle coordination methods include approaches based on time windows, reservation tables, priorities, or conflict search, many solutions fail to organically combine global feasible path planning for articulated vehicles, time parameterization under speed constraints, 3D spatiotemporal occupancy modeling, and reservation-aware local repair. This results in a disconnect between global path feasibility and local temporal coordination, thereby affecting the stable implementation of the technical results.

[0007] Therefore, it is necessary to propose a collaborative obstacle avoidance scheme for multiple intelligent logistics vehicles in warehousing scenarios. This scheme can generate basic paths that satisfy kinematic constraints and multi-rigid-body collision constraints for each intelligent logistics vehicle, transform the trajectories of multiple vehicles into a unified spatiotemporal occupancy relationship, and implement constrained local repair based on the intensity of the conflict. This results in an executable, conflict-free collaborative obstacle avoidance trajectory that is suitable for use in real vehicle control systems. Summary of the Invention

[0008] The purpose of this invention is to provide a collaborative obstacle avoidance method for multiple intelligent logistics vehicles in a warehousing scenario, in order to solve the problems in the prior art of insufficient consideration of the feasibility of articulated vehicle paths, lack of unified spatiotemporal coordination of multiple vehicle paths, and poor connection between local obstacle avoidance and global planning.

[0009] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0010] This invention provides a method for collaborative obstacle avoidance by multiple intelligent logistics vehicles in a warehousing scenario, applied to a collaborative planning system, wherein the method includes:

[0011] Obtain a map of the warehouse scene, operational task information, and the current and target status of each intelligent logistics vehicle;

[0012] For each intelligent logistics vehicle, a global path search is performed based on a preset algorithm within a state space that includes the vehicle's pose state and at least one vehicle body relative pose state to obtain the basic path corresponding to the intelligent logistics vehicle that satisfies the vehicle's kinematic constraints.

[0013] Based on the curvature of the base path, the relative attitude of the vehicle body, the obstacle clearance, the area speed limit, and the tracking error margin, a speed profile is allocated, and a global trajectory with timestamps is generated based on the allocated speed profile.

[0014] The timestamped global trajectory of each intelligent logistics vehicle is mapped into a three-dimensional spatiotemporal occupancy model. The vehicle body occupancy area and safety margin of each intelligent logistics vehicle in each time index or time interval are recorded in the three-dimensional spatiotemporal occupancy model, and path conflicts between intelligent logistics vehicles are detected.

[0015] Based on the number of path conflicts and the global trajectory length of each intelligent logistics vehicle, the first intelligent logistics vehicle that retains its time and space occupancy and the second intelligent logistics vehicle that does not retain its time and space occupancy are determined.

[0016] For each second intelligent logistics vehicle, the conflict trajectory segment between the second intelligent logistics vehicle and the first intelligent logistics vehicle is determined, and the local repair corridor is constructed by the reservation perception dynamic window method in the conflict trajectory segment and the neighborhood to obtain the locally repaired trajectory.

[0017] The repaired trajectories of each local area are input into the three-dimensional spatiotemporal occupancy model to update the three-dimensional spatiotemporal occupancy model;

[0018] Based on the updated 3D spatiotemporal occupancy model, global trajectories are assigned to each intelligent logistics vehicle until all intelligent logistics vehicles have their spatiotemporal occupancy preserved, generating conflict-free collaborative obstacle avoidance trajectories for each intelligent logistics vehicle.

[0019] This invention also provides a multi-intelligent logistics vehicle collaborative obstacle avoidance device for warehousing scenarios, applied to a collaborative planning system, wherein the device includes:

[0020] The acquisition module is used to acquire the warehouse scene map, operation task information, and the current and target status of each intelligent logistics vehicle;

[0021] The basic path determination module is used to perform a global path search based on a preset algorithm for each intelligent logistics vehicle, within a state space that includes the vehicle's pose state and at least one relative attitude state of the vehicle body, to obtain the basic path corresponding to the intelligent logistics vehicle that satisfies the vehicle's kinematic constraints.

[0022] The global trajectory determination module is used to allocate speed profiles based on the curvature of the base path, the relative attitude of the vehicle body, the obstacle clearance, the area speed limit and the tracking error margin, and generate a global trajectory with timestamps based on the allocated speed profiles.

[0023] The model building module is used to map the timestamped global trajectory of each intelligent logistics vehicle into a three-dimensional spatiotemporal occupancy model. In the three-dimensional spatiotemporal occupancy model, the vehicle body occupancy area and safety margin of each intelligent logistics vehicle in each time index or time interval are recorded, and path conflicts between intelligent logistics vehicles are detected.

[0024] The filtering module is used to determine the first intelligent logistics vehicle to retain its time and space occupancy and the second intelligent logistics vehicle to not retain its time and space occupancy, based on the number of path conflicts and the global trajectory length of each intelligent logistics vehicle.

[0025] The trajectory repair module is used to determine the conflict trajectory segment between the second intelligent logistics vehicle and the first intelligent logistics vehicle for each second intelligent logistics vehicle, and to use the reservation perception dynamic window method to perform local repair and build a local repair corridor in the conflict trajectory segment and its neighborhood to obtain the locally repaired trajectory.

[0026] The iteration module is used to input the locally repaired trajectories into the three-dimensional spatiotemporal occupancy model to update the three-dimensional spatiotemporal occupancy model; based on the updated three-dimensional spatiotemporal occupancy model, global trajectories are assigned to each intelligent logistics vehicle until the spatiotemporal occupancy of each intelligent logistics vehicle is preserved, generating conflict-free cooperative obstacle avoidance trajectories for each intelligent logistics vehicle.

[0027] This invention provides an electronic device, which includes a processor, a memory, and a program or instructions stored in the memory and executable on the processor. When the program or instructions are executed by the processor, they implement the steps of the multi-intelligent logistics vehicle collaborative obstacle avoidance method in any of the above-described warehousing scenarios.

[0028] This invention provides a readable storage medium storing a program or instructions, which, when executed by a processor, implement the steps of the multi-intelligent logistics vehicle collaborative obstacle avoidance method in any of the above-described warehousing scenarios.

[0029] The multi-intelligent logistics vehicle collaborative obstacle avoidance scheme provided in this embodiment of the invention has the following advantages: First, in the global planning stage, vehicle kinematic constraints, multi-rigid-body collision detection, and articulation state constraints are uniformly incorporated into the global search process, which can improve the executability of the basic path from the source. Second, by allocating velocity profiles to the basic path, multi-vehicle coordination is established on a global trajectory with timestamps, thereby improving the accuracy of time-series coordination. Third, by uniformly detecting path conflicts through a three-dimensional spatiotemporal occupancy model and combining it with the reservation-aware dynamic window method to implement local repair, multi-vehicle conflicts, deadlocks, and oscillations can be reduced. Attached Figure Description

[0030] Figure 1 This is a flowchart illustrating the steps of a collaborative obstacle avoidance method for multiple intelligent logistics vehicles in a warehousing scenario, as described in an embodiment of this application.

[0031] Figure 2 This is a schematic diagram of the collaborative obstacle avoidance system architecture for multiple intelligent logistics vehicles in a warehousing scenario, as described in this application embodiment.

[0032] Figure 3 This is a schematic diagram of a warehouse scene map according to an embodiment of this application;

[0033] Figure 4 This is a schematic diagram of the kinematic model of the articulated intelligent logistics vehicle according to an embodiment of this application;

[0034] Figure 5 This is a flowchart of the hybrid A* global path planning process according to an embodiment of this application.

[0035] Figure 6 This is a schematic diagram of velocity profile allocation according to an embodiment of this application;

[0036] Figure 7 This is a three-dimensional spatiotemporal occupancy model and reservation representation for an embodiment of this application;

[0037] Figure 8 This is a schematic diagram illustrating the determination of positional and intersectional conflicts in an embodiment of this application;

[0038] Figure 9This is a flowchart illustrating the vehicle priority calculation process according to an embodiment of this application.

[0039] Figure 10 This is a flowchart illustrating the partial repair process of the appointment-aware dynamic window in an embodiment of this application.

[0040] Figure 11 This is a flowchart illustrating the overall process of multi-intelligent logistics vehicle collaborative obstacle avoidance in an embodiment of this application.

[0041] Figure 12 This is a structural block diagram illustrating a collaborative obstacle avoidance device for multiple intelligent logistics vehicles in a warehousing scenario, according to an embodiment of this application. Detailed Implementation

[0042] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0043] The following description, in conjunction with the accompanying drawings, details the multi-intelligent logistics vehicle collaborative obstacle avoidance scheme in a warehousing scenario provided by the embodiments of this application through specific implementations and application scenarios.

[0044] As attached Figure 1 As shown in the figure, the multi-intelligent logistics vehicle cooperative obstacle avoidance method in a warehousing scenario according to this application embodiment includes the following steps:

[0045] Step 101: Obtain the warehouse scene map, operation task information, and the current and target status of each intelligent logistics vehicle.

[0046] The multi-intelligent logistics vehicle cooperative obstacle avoidance method provided in this application embodiment ultimately provides a conflict-free cooperative obstacle avoidance trajectory for the on-board controller to execute. It processes technical data related to the operation of warehouse vehicles, such as maps, poses, speeds, steering, and articulation states, and can be directly applied to the real vehicle control process.

[0047] The warehouse scenario map includes at least two of the following: static obstacle areas, passable aisles, intersection areas, loading and unloading areas, and temporary storage areas, so that the collaborative planning system can perform collaborative obstacle avoidance in scenarios such as vehicles meeting in narrow aisles, intersections, or loading and unloading ports.

[0048] When the intelligent logistics vehicle is an articulated vehicle, the current state also includes the trailer heading or articulation angle. Specifically, the vehicle state can be represented by a state vector s=[x, y, θ, β] or s=[x, y, θ, γ]; and the vehicle control input can be represented by a control variable u=[v, δ].

[0049] Where x and y represent the reference point position of the tractor, θ represents the heading angle of the tractor, β represents the heading angle of the trailer, γ represents the articulation angle, and γ = θ - β; v represents the longitudinal velocity, and δ represents the steering amount.

[0050] Step 102: For each intelligent logistics vehicle, in the state space containing the vehicle pose state and at least one vehicle body relative attitude state, perform a global path search based on a preset algorithm to obtain the basic path corresponding to the intelligent logistics vehicle that satisfies the vehicle kinematic constraints.

[0051] In an optional embodiment, when performing step 102, for each intelligent logistics vehicle, in the state space containing the vehicle pose state and at least one vehicle body relative pose state, during the global path search based on the hybrid A* algorithm, collision detection is performed on the multi-rigid-body occupancy of the intelligent logistics vehicle, and each cost parameter is substituted into the global cost function to obtain the basic path corresponding to the intelligent logistics vehicle that satisfies the vehicle kinematic constraints.

[0052] The global cost function includes at least a portion of the following: path length cost, steering amplitude cost, steering change cost, reversing cost, articulation angle cost, folding knife approach cost, obstacle approach cost, and heuristic cost.

[0053] In this optional embodiment, a hybrid A* algorithm is used to perform collision detection on the multi-rigid-body occupancy of vehicles during the path search process, and at least a portion of the path length, steering amplitude, steering change, reversing, articulation angle, folding knife risk, obstacle proximity, and heuristic estimation are included in the global cost function, so that the obtained basic path can take into account traffic efficiency, executability, and safety.

[0054] In one optional embodiment, the collision detection method for the multi-rigid-body occupancy of the intelligent logistics vehicle can be as follows: when the intelligent logistics vehicle is an articulated vehicle, the tractor and trailer of the intelligent logistics vehicle are represented as envelopes of preset shapes; and collision detection is performed on multiple intermediate states of each motion primitive to detect static obstacle collisions, boundary crossings, and trailer rear-sweep collisions. The preset-shaped envelopes include: rectangular envelopes, polygonal envelopes, and approximate envelopes composed of multiple circular envelopes.

[0055] When the intelligent logistics vehicle is an articulated vehicle, the state space includes the tractor position, tractor heading angle, trailer heading angle, and / or articulation angle. Performing a global search based on this state space is beneficial for simultaneously reflecting turning constraints, trailer following relationships, and folding risks during the planning phase.

[0056] Step 103: Assign a speed profile based on the curvature of the base path, the relative attitude of the vehicle body, the obstacle clearance, the area speed limit, and the tracking error margin, and generate a global trajectory with timestamps based on the assigned speed profile.

[0057] Step 103 may include the following sub-steps during its execution:

[0058] Sub-step 1: Determine the upper limit of curvature constraint velocity based on the local curvature of the base path;

[0059] Sub-step 2: Determine the articulation state based on the relative attitude of the vehicle body, and determine the upper limit of the attitude constraint speed based on the articulation state;

[0060] Sub-step 3: Determine the upper limit of safe speed based on the obstacle clearance, and determine the target speed of each trajectory point or trajectory segment in combination with the area speed limit and the tracking error margin.

[0061] Among them, the target velocity of each trajectory point or trajectory segment is the minimum value among the upper limit of curvature constraint velocity, upper limit of attitude constraint velocity, upper limit of safe velocity, upper limit of regional speed limit, and upper limit of velocity corresponding to tracking error margin; the velocity profile is determined by the target velocity of each trajectory point or trajectory segment.

[0062] In this optional embodiment, the allocation of the velocity profile is determined based on the curvature of the base path, the relative attitude of the vehicle body, obstacle clearance, area speed limit, and tracking error margin. By performing time parameterization during the planning phase, the original geometric path can be converted into a timestamped global trajectory, enabling subsequent multi-vehicle collision detection to be based on a temporal sequence consistent with real vehicle operation.

[0063] Step 104: Map the timestamped global trajectory of each intelligent logistics vehicle into a three-dimensional spatiotemporal occupancy model. Record the vehicle body occupancy area and safety margin of each intelligent logistics vehicle in each time index or time interval in the three-dimensional spatiotemporal occupancy model, and detect path conflicts between intelligent logistics vehicles.

[0064] Path conflicts include at least location conflicts and intersection conflicts. Location conflicts are used to characterize the situation where different intelligent logistics vehicles occupy the same spatial unit or spatial area at the same time. Intersection conflicts are used to characterize the situation where different intelligent logistics vehicles perform reverse exchange passages on the same road segment in adjacent time periods.

[0065] Location conflict refers to two intelligent logistics vehicles occupying the same grid cell or the same spatial area at the same time; cross conflict refers to the first intelligent logistics vehicle being located in the first spatial cell at the first moment and in the second spatial cell at the second moment, while the second intelligent logistics vehicle is located in the second spatial cell at the first moment and in the first spatial cell at the second moment.

[0066] In one optional embodiment, the three-dimensional spatiotemporal occupancy model is a three-dimensional spatiotemporal grid model and / or a spatiotemporal reservation table with time intervals; when the intelligent logistics vehicle is an articulated vehicle, the tractor-occupied area and the trailer-occupied area are recorded in each time index or time interval, and a safety margin is extended outside the occupancy area, which can more accurately characterize the actual sweeping range.

[0067] In one optional embodiment, the three-dimensional spatiotemporal occupancy model includes: the vehicle body occupancy area and safety margin of each intelligent logistics vehicle in each time index or time interval; when generating the three-dimensional spatiotemporal occupancy model, the vehicle body occupancy area in the three-dimensional spatiotemporal occupancy model can be adaptively expanded according to at least one of vehicle positioning error, trajectory tracking error, communication delay and trailer tail sweep effect.

[0068] Step 105: Based on the number of path conflicts and the global trajectory length corresponding to each intelligent logistics vehicle, determine the first intelligent logistics vehicle that retains its time and space occupancy and the second intelligent logistics vehicle that does not retain its time and space occupancy.

[0069] When calculating the number of path conflicts and the global trajectory length for each intelligent logistics vehicle, the vehicle priority can be determined by comparing the number of path conflicts first and then the global trajectory length. The time and space occupancy of the intelligent logistics vehicles with higher priority is retained in order of vehicle priority. The intelligent logistics vehicles whose time and space occupancy is retained are the first intelligent logistics vehicles, and the intelligent logistics vehicles whose time and space occupancy is not retained are the second intelligent logistics vehicles.

[0070] More specifically, determining vehicle priorities includes: first sorting by the number of path conflicts from largest to smallest; when the number of path conflicts is the same, then sorting by the global trajectory length from longest to shortest; when the global trajectory length is still the same, then sorting by vehicle number or task generation time for a stable sort.

[0071] Vehicle priority is determined by first comparing the number of path conflicts and then comparing the global trajectory length. This priority rule helps to prioritize trajectories that have a greater impact on overall traffic order, thereby reducing the overhead of conflict propagation and repeated replanning.

[0072] Step 106: For each second intelligent logistics vehicle, determine the conflict trajectory segment between the second intelligent logistics vehicle and the first intelligent logistics vehicle, and use the reservation perception dynamic window method to locally repair and build a local repair corridor in the conflict trajectory segment and its neighborhood to obtain the locally repaired trajectory.

[0073] For the second intelligent logistics vehicle with low priority and not occupied by the reserved spacetime, it indicates that there are conflicting trajectory segments in its global trajectory. Therefore, in this step, a local repair corridor is constructed for the conflicting trajectory segments to avoid the conflict.

[0074] In one optional embodiment, the local repair corridor is constructed by using a reservation-aware dynamic window method in the conflict trajectory segment and its neighborhood. The locally repaired trajectory can be obtained in the following way:

[0075] The candidate control commands are simulated forward to eliminate candidate trajectories that collide with static obstacles or reserved space-time occupancy, resulting in the remaining candidate trajectories. From the remaining candidate trajectories, the conflict-free candidate trajectory is selected as the trajectory after local repair. When there are no feasible candidate trajectories in the current prediction time domain, a safety waiting action is inserted for the second intelligent logistics vehicle, and the local repair is re-executed in the subsequent safe time interval.

[0076] The safe waiting action can be any one or more of the following: waiting in place, maintaining low speed, or minimum maneuvering waiting within the current safe area; after executing the safe waiting action, the trajectory local repair can be re-executed based on the updated spatiotemporal reservation state.

[0077] For the local repair of low-priority intelligent logistics vehicles (i.e., the second intelligent logistics vehicle), a reservation-aware dynamic window method is adopted. This method determines the set of candidate control commands based on the current vehicle state and constraints, performs forward simulation of the candidate control commands, and simultaneously executes static obstacle collision detection and spatiotemporal reservation collision detection. This allows for explicit consideration of reserved spatiotemporal occupancy during the local repair phase. When the dynamic window method fails to find a feasible candidate trajectory within the current prediction time domain, a safety waiting action is inserted for the low-priority intelligent logistics vehicle, and local repair is repeated within the subsequent available safe time interval. By introducing a safety waiting action, planning failure can be avoided directly due to the lack of a solution within the local window at narrow passing points, intersections, or temporarily congested areas.

[0078] An optional method for constructing a locally repaired corridor by employing a reservation-aware dynamic window method in the conflict trajectory segment and its neighborhood to obtain the locally repaired trajectory may include the following sub-steps:

[0079] Sub-step 1: Determine the speed sampling window and the steering sampling window based on the current longitudinal speed, maximum acceleration, maximum deceleration, current steering amount, maximum steering change rate, and discrete time step of the second intelligent logistics vehicle.

[0080] Sub-step 2: Sample candidate control commands within the speed sampling window and steering sampling window.

[0081] Sub-step 3: Determine the local repair corridor based on the conflict trajectory segment, the passable areas on both sides of the conflict trajectory segment, and the safe passage envelope of the target vehicle, and obtain the locally repaired trajectory corresponding to the second intelligent logistics vehicle.

[0082] In actual implementation, when scoring candidate trajectories that do not collide with static obstacles or reserved space-time occupancy, at least one of the following should be considered: progress along the reference trajectory, heading consistency, obstacle clearance, relative reference corridor deviation, articulation angle penalty, and control smoothness. The candidate trajectory with the best score is selected as the local repair trajectory.

[0083] Step 107: Input the locally repaired trajectories into the three-dimensional spatiotemporal occupancy model to update the three-dimensional spatiotemporal occupancy model.

[0084] The updated model will replace the original local trajectory with the locally repaired trajectory generated by the second intelligent logistics vehicle.

[0085] Step 108: Assign global trajectories to each intelligent logistics vehicle based on the updated 3D spatiotemporal occupancy model until each intelligent logistics vehicle has its spatiotemporal occupancy preserved, generating conflict-free cooperative obstacle avoidance trajectories for each intelligent logistics vehicle.

[0086] After updating the 3D spatiotemporal occupancy model, the process returns to step 104, and steps 104-108 are executed repeatedly to determine the second intelligent logistics vehicle, modify its local trajectory, update the 3D spatiotemporal occupancy model, and reserve spatiotemporal occupancy for each intelligent logistics vehicle. This continues until the conflict-free cooperative obstacle avoidance trajectories of all intelligent logistics vehicles are obtained, and these trajectories are then sent to the corresponding onboard controllers for execution.

[0087] The multi-intelligent logistics vehicle cooperative obstacle avoidance method provided in this application embodiment operates in a rolling time domain manner; after sending the conflict-free cooperative obstacle avoidance trajectory in the current time domain to the vehicle controller, it continuously receives real-time status information fed back by each intelligent logistics vehicle, re-verifies the spatiotemporal occupancy of subsequent trajectories, and re-executes at least a portion of steps 104 to 108 when a new conflict is detected.

[0088] Real-time status information includes at least the current position, current speed, current heading, current steering angle, and, when the intelligent logistics vehicle is an articulated vehicle, the current trailer heading or current articulation angle.

[0089] The method provided in this application operates in a rolling time domain manner. After sending the conflict-free cooperative obstacle avoidance trajectory in the current time domain to the vehicle controller, it continuously receives real-time status information from each intelligent logistics vehicle, re-verifies subsequent spatiotemporal occupancy, and re-executes conflict detection and local repair when necessary.

[0090] The multi-intelligent logistics vehicle collaborative obstacle avoidance method provided in this application embodiment has the following advantages: First, in the global planning stage, vehicle kinematic constraints, multi-rigid-body collision detection, and articulation state constraints are uniformly incorporated into the global search process, which can improve the executability of the basic path from the source. Second, by allocating speed profiles to the basic path, multi-vehicle coordination is established on a global trajectory with timestamps, thereby improving the accuracy of time-series coordination. Third, by uniformly detecting path conflicts through a three-dimensional spatiotemporal occupancy model and combining it with the reservation-aware dynamic window method to implement local repair, multi-vehicle conflicts, deadlocks, and oscillations can be reduced.

[0091] The following specific examples illustrate the collaborative obstacle avoidance method for multiple intelligent logistics vehicles in a warehousing scenario provided in the embodiments of this application.

[0092] The obstacle avoidance method for multiple intelligent logistics vehicles in a warehousing scenario provided in this application embodiment is applied to a collaborative planning system.

[0093] like Figure 2 As shown, the collaborative planning system may include a collaborative planning server 100, a map database 200, a reservation database 300, a communication network 400, and multiple intelligent logistics vehicles 500. Each intelligent logistics vehicle 500 may be an articulated autonomous truck, an unmanned tractor, an automated guided vehicle, an unmanned forklift, or other ground vehicles capable of automatically operating according to a planned trajectory. The collaborative planning server 100 may centrally execute global path planning, velocity profile allocation, 3D spatiotemporal occupancy modeling, conflict detection, priority sorting, and local repair, and may also collaborate with the onboard controller to execute some of these steps.

[0094] like Figure 3 As shown in the schematic diagram of the warehouse scene map, the warehouse scene map can be an occupancy grid map, a topological road network map, a lane segment map, a polygon obstacle map, or any combination of two or more of the above. The map includes at least a static obstacle area 601, a passable lane 602, an intersection area 603, a loading and unloading area 604, a temporary storage area 605, and a task start and end point. Preferably, to balance planning efficiency and geometric accuracy, the occupancy grid can be used for global passability determination, and the polygon boundaries can be used for vehicle envelope collision detection.

[0095] In one embodiment, the spatial resolution r of the warehouse scene map can be set to 0.05 meters to 0.50 meters, and the time interval Δt can be set to 0.05 seconds to 0.50 seconds. For intersection areas requiring high-precision vehicle passing control, a local higher-resolution raster or polygon sub-map can be used; for long-distance lanes, a relatively coarse resolution can be used to reduce the computational burden.

[0096] When the intelligent logistics vehicle is an articulated vehicle, the vehicle state can be represented as s=[x, y, θ, β] or s=[x, y, θ, γ]. Here, x and y represent the position coordinates of the tractor's reference point, θ represents the tractor's heading angle, β represents the trailer's heading angle, and γ represents the articulation angle, where γ=θ-β. The control variable can be represented as u=[v, δ], where v represents the longitudinal velocity and δ represents the steering amount. For non-articulated vehicles, the β or γ related states can be omitted, and the multi-rigid-body occupancy is reduced to single-vehicle-body occupancy.

[0097] like Figure 4The schematic diagram of the kinematic model of the articulated intelligent logistics vehicle shows that, in the low-speed warehouse operation scenario, the state propagation can be described by a discrete kinematic model: x(k+1)=x(k)+v(k)cosθ(k)Δt;

[0098] y(k+1)=y(k)+v(k)sinθ(k)Δt;

[0099] θ(k+1)=θ(k)+v(k)tanδ(k)Δt / L1;

[0100] β(k+1)=β(k)+v(k)sin(θ(k)-β(k))Δt / L2,

[0101] Where L1 is the wheelbase of the tractor and L2 is the equivalent length of the trailer. When γ is used as the state variable, β(k+1) can also be converted into an equivalent update form of γ(k+1).

[0102] like Figure 5 The flowchart of the hybrid A* global path planning is shown in the figure. The hybrid A* algorithm is used in the global path planning stage. First, spatial discrete cells are established based on map resolution, and the heading angle and articulation angle are discretized. Then, motion primitives are constructed based on forward, reverse and different steering amounts, and candidate nodes are propagated from the current node. After collision detection, cost calculation and heuristic estimation of candidate nodes, candidate nodes that meet the constraints are added to the open list. The above process is repeated until the target area is reached or the termination condition is met.

[0103] In one embodiment, the global cost function of the hybrid A* algorithm can be expressed as:

[0104] Jg=w1·Lg+w2·Cs+w3·CΔs+w4·Cr+w5·Cγ+w6·Cjk+w7·Ccl+w8·H.

[0105] Where Lg represents path length cost, Cs represents steering amplitude cost, CΔs represents steering change cost, Cr represents reversing cost, Cγ represents articulation angle cost, Cjk represents folding knife approach cost, Ccl represents obstacle approach cost, H represents heuristic cost, and w1 to w8 are the corresponding weights. By weighting each cost item, traffic efficiency, smoothness, executability, and safety can be balanced.

[0106] In one embodiment, collision detection does not only detect whether the center trajectory of the tractor collides with an obstacle, but performs collision detection separately on the tractor envelope 701 and the trailer envelope 702. Specifically, the tractor and trailer can be represented as rectangular envelopes, polygonal envelopes, or approximate envelopes composed of multiple circular envelopes, and multiple intermediate states on each motion primitive are sampled to detect static obstacle collisions, boundary crossings, and trailer tail-sweep collisions.

[0107] After obtaining the basic path that satisfies the geometric and kinematic constraints, as follows: Figure 6 As shown in the speed profile allocation diagram, this embodiment further performs speed profile allocation on the basic path. The purpose of speed profile allocation is to transform the basic path from a geometric path into a timestamped global trajectory, enabling multi-vehicle conflict detection and coordination to be based on an executable temporal trajectory.

[0108] In one embodiment, for the i-th trajectory point or trajectory segment, the target velocity vi can be determined, satisfying: vi = min{vκ(i), vγ(i), vcl(i), vzone(i), vtrack(i)}. Here, vκ(i) represents the upper speed limit obtained from local curvature constraints, vγ(i) represents the upper speed limit obtained from articulated state constraints, vcl(i) represents the safe upper speed limit obtained from obstacle clearance, vzone(i) represents the zone speed limit, and vtrack(i) represents the upper speed limit determined by the tracking error margin.

[0109] For example, in one embodiment, vκ(i) = √(ay,max / (|κ(i)|+ε)) can be determined based on the permissible lateral acceleration threshold ay,max and the local curvature κ(i); and vγ(i) = v0 / (1+λ|γ(i)|) can be determined based on the absolute value of the hinge angle |γ(i)|. Here, ε is a small constant to prevent the denominator from being zero, and v0 and λ are preset parameters. In this way, the target velocity can be automatically reduced in high curvature regions, narrow passages, or large hinge angle regions.

[0110] After the velocity profile is determined, the arrival time can be accumulated segment by segment based on the path length between adjacent trajectory points and the corresponding target velocity, thereby assigning a timestamp to each trajectory point; alternatively, the trajectory can be interpolated and sampled at a uniform discrete time step to obtain a state sequence on a uniform time scale. The resulting timestamped global trajectory contains both spatial and temporal information.

[0111] like Figure 7 As shown in the schematic diagram of the three-dimensional spatiotemporal occupancy model and reservation representation, the timestamped global trajectory of each intelligent logistics vehicle is further mapped to a three-dimensional spatiotemporal occupancy model 801. The three-dimensional spatiotemporal occupancy model 801 can use a combination of spatial grid cells and time indices as basic storage units, or it can use spatial polygonal regions and corresponding time intervals as basic storage units. For articulated vehicles, the occupancy of the tractor and trailer is recorded simultaneously within each time index or time interval.

[0112] In one embodiment, to improve the reservation model's tolerance to real-world operational errors, the vehicle's occupied area can be adaptively expanded. The expansion amount ρ(t) can be related to the vehicle positioning error eloc, trajectory tracking error etrack, communication delay τcomm, and trailer tail scan compensation amount etail, and can be expressed as follows:

[0113] ρ(t)=ρ0+k1·eloc+k2·etrack+v(t)·τcomm+k3·etail,

[0114] ρ0, k1, k2, and k3 are preset parameters.

[0115] Figure 8 This is a schematic diagram illustrating the determination of positional and intersection conflicts in an embodiment of this application. Figure 8 As shown, conflict detection includes at least positional conflict detection and intersection conflict detection. If the first vehicle and the second vehicle occupy the same grid cell or the same spatial area at the same time index t, it is determined to be a positional conflict; if the first vehicle moves from spatial cell a to spatial cell b at time tk, and the second vehicle moves from spatial cell b to spatial cell a within the same time period, it is determined to be an intersection conflict. In some embodiments, it can further detect same-direction rear-end collisions, intersection merging conflicts, or area overload conflicts.

[0116] Figure 9 This is a flowchart illustrating the vehicle priority calculation process according to an embodiment of this application. Figure 9 As shown, the system can count the number of conflicts Ni and the global trajectory length Di for each intelligent logistics vehicle, and sort them according to the priority rule Pi=(Ni,Di). First, Ni is compared, and the one with larger Ni has higher priority. When Ni is the same, Di is compared, and the one with longer Di has higher priority. When Di is still the same, it can be sorted stably by vehicle number or task generation time to ensure that the sorting result is deterministic.

[0117] After determining the priorities, the spatiotemporal occupancy of the highest priority intelligent logistics vehicle is first written into the reservation database 300. For subsequent lower priority intelligent logistics vehicles, it is checked whether their existing global trajectory conflicts with the reserved spatiotemporal occupancy. If a conflict exists, it is not necessary to completely replan the entire trajectory; instead, a local repair corridor can be constructed only for the conflicting trajectory segment and its neighborhood.

[0118] A local repair corridor can be defined by the conflict trajectory segment, the passable areas on both sides of the conflict trajectory segment, the safe passage envelope of the target vehicle, and adjacent available time intervals. By limiting the scope of local repair, the computational cost of local repair can be reduced while maintaining the overall intent of the original global path.

[0119] Figure 10This is a flowchart illustrating the partial repair process of the appointment-aware dynamic window according to an embodiment of this application. Figure 10 As shown, the local repair phase employs the scheduled perception dynamic window method. First, the speed sampling window and steering sampling window are determined based on the current vehicle state and vehicle constraints. For example, the longitudinal speed window [max(vmin, vcur-bmaxΔt), min(vmax, vcur+amaxΔt)] can be determined based on the current longitudinal speed vcur, maximum acceleration amax, maximum deceleration bmax, and discrete time step Δt; the steering window [max(δmin, δcur-ωδΔt), min(δmax, δcur+ωδΔt)] can be determined based on the current steering amount δcur, maximum steering change rate ωδ, and discrete time step Δt.

[0120] Candidate control commands are sampled within the speed sampling window and the steering sampling window, and each candidate control command is forward simulated according to the vehicle kinematics model within the prediction time domain Th to obtain the candidate predicted trajectory. Preferably, the prediction time domain Th can be set to 0.5 seconds to 5 seconds, and the sampling resolution can be set according to the vehicle size, operating speed, and computing resources.

[0121] For each candidate predicted trajectory, static obstacle collision detection is performed first, followed by collision detection with reserved spacetime occupancy. If any candidate predicted trajectory collides with a static obstacle or reserved spacetime occupancy at any sampling time, the candidate predicted trajectory is discarded. The resulting set of candidate trajectories constitutes a feasible local control set without immediate conflicts.

[0122] When evaluating candidate trajectories that pass collision detection, the following evaluation function can be used:

[0123] Jdwa=α·Jprog+β·Jhead+χ·Jclear-η·Jcorr-μ·Jγ-ν·Jsmooth

[0124] Where Jprog represents the progress along the reference trajectory, Jhead represents heading consistency, Jclear represents obstacle clearance, Jcorr represents the deviation from the reference corridor, Jγ represents the articulation angle penalty, Jsmooth represents the control smoothness, and α, β, χ, η, μ, and ν are the corresponding weights. The candidate trajectory with the best score is selected as the local repair trajectory.

[0125] In a preferred embodiment, if no feasible candidate trajectory exists within the current prediction time domain, the planning is not directly deemed a failure. Instead, a safety waiting action is inserted for the current low-priority intelligent logistics vehicle. This safety waiting action can be in-place waiting, maintaining low speed, or minimal maneuvering within the current safe area. After waiting for one or more time steps, the local repair is re-executed based on the updated reservation status.

[0126] After a low-priority vehicle completes partial repair, the new trajectory segment obtained from the repair replaces the original conflicting trajectory segment, and its new spatiotemporal occupancy is written into the reservation database 300. Subsequently, the conflict detection, partial repair, and reservation update are repeated for the next low-priority intelligent logistics vehicle until all vehicles have obtained conflict-free cooperative obstacle avoidance trajectories.

[0127] Figure 11 This is a flowchart illustrating the overall process of multi-intelligent logistics vehicle cooperative obstacle avoidance in an embodiment of this application. Figure 11 As shown, the method of the present invention can be executed according to the process of S101 to S107. It should be noted that in different embodiments, some steps of S101 to S107 can be combined, split, or executed in an equivalent order, as long as they do not deviate from the core concept of the present invention of achieving multi-vehicle cooperative obstacle avoidance by combining globally feasible paths, time-parameterized trajectories, spatiotemporal occupancy reservations, and reservation-aware local repair.

[0128] In one embodiment, the method of the present invention operates in a rolling time domain manner. After the collaborative planning server 100 sends the conflict-free collaborative obstacle avoidance trajectory in the current time domain to each vehicle controller, it can continuously receive the current position, current speed, current heading, current steering amount, and current trailer heading or current articulation angle when the intelligent logistics vehicle is an articulated vehicle, according to a preset update frequency, and re-verify or partially replan the subsequent trajectory to adapt to tracking errors, changes in temporary obstacles, or communication delays.

[0129] The data processed in this invention includes warehouse map data, vehicle pose data, speed data, steering data, articulation state data, and reservation / occupancy data. The output is a sequence of trajectory points, speed commands, and / or steering commands for the onboard controller to execute. This processing directly affects the motion control of intelligent logistics vehicles in a real warehouse environment, aiming to reduce collision risks and improve multi-vehicle traffic efficiency. Therefore, it is a technical solution proposed for a specific technical problem, rather than an abstract algorithm rule detached from the application scenario.

[0130] To further illustrate the feasibility of this invention, a warehouse simulation embodiment is given below. The warehouse scenario is a warehouse plan containing multiple narrow alleyways and a central intersection area. The map size can be 40 meters × 30 meters, and the grid resolution can be 0.20 meters. Four articulated autonomous trucks are set up to drive from different starting points to different destinations, and each vehicle needs to complete its task by passing through the central intersection area.

[0131] In this embodiment, the tractor wheelbase L1 can be set to 3.6 meters, the equivalent trailer length L2 to 8.0 meters, the vehicle width to 2.5 meters, the maximum longitudinal speed vmax to 2.5 m / s, the maximum longitudinal acceleration amax to 0.8 m / s², the maximum deceleration bmax to 1.0 m / s², and the maximum rate of change of steering ωδ to 0.35 radians / s. The state discrete resolution, motion primitive length, and cost weight of the hybrid A* stage can be set according to operational requirements.

[0132] If only two-dimensional geometric paths are planned for the four vehicles without introducing spatiotemporal reservations, the vehicles are prone to in-place conflicts and head-on exchange conflicts in the central intersection area. Using the method of this invention, each vehicle first obtains a basic path that satisfies articulation constraints and multi-rigid-body collision constraints, and then forms a timestamped global trajectory through velocity profile allocation; subsequently, a unified conflict detection and priority ranking are performed using a three-dimensional spatiotemporal occupancy model, and local repair of the reservation-aware dynamic window is performed for low-priority vehicles.

[0133] In a specific simulation result, the initial independent planning resulted in 27 spatiotemporal conflicts. After adopting the baseline scheme based solely on priority delay, the number of conflicts could be reduced, but residual conflicts still existed. After adopting the method of this invention, all conflicts could be eliminated, the minimum safe distance was increased, and low-priority vehicles could delay entering the intersection area by waiting for safety actions when necessary, thereby forming a conflict-free passage scheme.

[0134] The simulation results above demonstrate that the method of this invention can effectively resolve spatiotemporal conflicts between multiple vehicles while maintaining the overall feasibility of the basic path. It is particularly suitable for high-conflict scenarios such as warehouse intersections, loading / unloading dock merging areas, and narrow lane encounters. It should be understood that the simulation parameters and results are only used to illustrate the feasibility of this invention and do not constitute a limitation on the scope of protection of this invention.

[0135] In one system implementation, the collaborative planning server 100 may include a processor, a memory, and a communication interface. The memory stores a computer program for executing the method of the present invention. When the processor invokes the computer program, it can perform map reading, global path planning, velocity profile allocation, spatiotemporal occupancy modeling, conflict detection, priority sorting, local repair, and trajectory distribution.

[0136] In one implementation, the vehicle controller can receive a conflict-free cooperative obstacle avoidance trajectory issued by the cooperative planning server 100, and control the vehicle operation in conjunction with vehicle perception and actuators; at the same time, the vehicle controller can feed back the current position, current speed, current heading, current steering amount and articulation status to the cooperative planning server 100 to support rolling time-domain updates.

[0137] This invention is not limited to the embodiments described above. For non-articulated vehicles, the trailer heading angle or articulation angle related states can be removed from the state space, and the multi-rigid-body occupancy can be degenerated into single-vehicle-body occupancy; for vehicles with multiple trailers, adaptation can also be achieved by expanding the state vector and collision detection envelope.

[0138] For map representation, spatiotemporal occupancy model form, conflict determination granularity, priority function, dynamic window evaluation function, speed profile model and safety waiting strategy, those skilled in the art can make equivalent substitutions or parameter adjustments according to specific warehousing tasks, vehicle forms and computing resources. As long as they do not deviate from the technical concept of combining global path feasibility and multi-vehicle spatiotemporal coordination of this invention, they should all fall within the protection scope of this invention.

[0139] Figure 12 The structural block diagram of the multi-intelligent logistics vehicle collaborative obstacle avoidance device in the warehousing scenario of this application embodiment is shown.

[0140] This application embodiment presents a multi-intelligent logistics vehicle collaborative obstacle avoidance device for warehousing scenarios, applied to a collaborative planning system. The device includes the following functional modules:

[0141] The acquisition module 1201 is used to acquire the warehouse scene map, operation task information, and the current and target status of each intelligent logistics vehicle;

[0142] The basic path determination module 1202 is used to perform a global path search based on a preset algorithm for each intelligent logistics vehicle in a state space that includes the vehicle's pose state and at least one vehicle body relative pose state, to obtain the basic path corresponding to the intelligent logistics vehicle that satisfies the vehicle's kinematic constraints.

[0143] The global trajectory determination module 1203 is used to allocate a speed profile based on the curvature of the base path, the relative attitude of the vehicle body, the obstacle clearance, the area speed limit and the tracking error margin, and generate a global trajectory with timestamps based on the allocated speed profile.

[0144] The model building module 1204 is used to map the timestamped global trajectory of each intelligent logistics vehicle into a three-dimensional spatiotemporal occupancy model. In the three-dimensional spatiotemporal occupancy model, the vehicle body occupancy area and safety margin of each intelligent logistics vehicle in each time index or time interval are recorded, and path conflicts between intelligent logistics vehicles are detected.

[0145] The filtering module 1205 is used to determine the first intelligent logistics vehicle that retains its time and space occupancy and the second intelligent logistics vehicle that does not retain its time and space occupancy based on the number of path conflicts and the global trajectory length of each intelligent logistics vehicle.

[0146] The trajectory repair module 1206 is used to determine the conflict trajectory segment between the second intelligent logistics vehicle and the first intelligent logistics vehicle for each second intelligent logistics vehicle, and to use the reservation perception dynamic window method to perform local repair and build a local repair corridor in the conflict trajectory segment and its neighborhood to obtain the locally repaired trajectory.

[0147] The iteration module 1207 is used to input the locally repaired trajectories into the three-dimensional spatiotemporal occupancy model to update the three-dimensional spatiotemporal occupancy model; and to allocate global trajectories to each intelligent logistics vehicle based on the updated three-dimensional spatiotemporal occupancy model until each intelligent logistics vehicle is retained in spatiotemporal occupancy to generate conflict-free cooperative obstacle avoidance trajectories for each intelligent logistics vehicle.

[0148] Optionally, the basic path determination module is specifically used to: for each intelligent logistics vehicle, in a state space containing the vehicle pose state and at least one vehicle body relative attitude state, during the global path search based on the hybrid A* algorithm, perform collision detection on the multi-rigid-body occupancy of the intelligent logistics vehicle, and substitute each cost parameter into the global cost function to obtain the basic path corresponding to the intelligent logistics vehicle that satisfies the vehicle kinematic constraints; wherein, the global cost function includes at least one of the following: path length cost, steering amplitude cost, steering change cost, reversing cost, articulation angle cost, folding knife approach cost, obstacle approach cost, and heuristic cost.

[0149] Optionally, the trajectory repair module includes:

[0150] The first submodule is used to determine the conflict trajectory segments between the second intelligent logistics vehicle and the first intelligent logistics vehicle for each second intelligent logistics vehicle.

[0151] The second submodule is used to perform forward simulation of candidate control commands, eliminate candidate trajectories that collide with static obstacles or reserved spacetime occupancy, and obtain the remaining candidate trajectories.

[0152] The third submodule is used to select a conflict-free candidate trajectory from the remaining candidate trajectories as the locally repaired trajectory.

[0153] The fourth submodule is used to insert a safety waiting action for the second intelligent logistics vehicle when there is no feasible candidate trajectory in the current prediction time domain, and to re-execute the local repair in the subsequent safe time interval.

[0154] Optionally, when the intelligent logistics vehicle is an articulated vehicle, the current state also includes the trailer heading or articulation angle, and the vehicle state is represented by the state vector s=[x, y, θ, β] or s=[x, y, θ, γ].

[0155] Where x and y represent the reference point position of the tractor, θ represents the heading angle of the tractor, β represents the heading angle of the trailer, and γ represents the articulation angle, and γ = θ - β;

[0156] The control input of the vehicle is represented by the control quantity u=[v,δ], where v represents the longitudinal speed and δ represents the steering amount.

[0157] Optionally, when the basic path determination module performs collision detection on the multi-rigid-body occupancy of the intelligent logistics vehicle, it is specifically used for:

[0158] When the intelligent logistics vehicle is an articulated vehicle, the tractor and trailer of the intelligent logistics vehicle are respectively represented as envelopes of preset shapes, wherein the envelopes of preset shapes include: rectangular envelopes, polygonal envelopes, and approximate envelopes composed of multiple circular envelopes.

[0159] It also samples and performs collision detection on multiple intermediate states of each motion primitive to detect static obstacle collisions, boundary crossings, and trailer tail-sweep collisions.

[0160] Optionally, the global trajectory determination module is specifically used for:

[0161] The upper limit of the curvature constraint velocity is determined based on the local curvature of the basic path.

[0162] The articulation state is determined based on the relative attitude of the vehicle body, and the upper limit of the attitude constraint speed is determined based on the articulation state.

[0163] The upper limit of safe speed is determined based on the obstacle clearance, and the target speed of each trajectory point or trajectory segment is determined by combining the area speed limit and the tracking error margin; wherein, the target speed of each trajectory point or trajectory segment is the minimum value among the upper limit of curvature constraint speed, upper limit of attitude constraint speed, upper limit of safe speed, upper limit of area speed limit and upper limit of speed corresponding to tracking error margin; the speed profile is determined by the target speed of each trajectory point or trajectory segment.

[0164] Optionally, the three-dimensional spatiotemporal occupancy model is a three-dimensional spatiotemporal grid model and / or a spatiotemporal reservation table with time intervals;

[0165] When the intelligent logistics vehicle is an articulated vehicle, the tractor-occupied area and the trailer-occupied area are recorded in each time index or time interval, and a safety margin is extended outside the occupied area.

[0166] Optionally, the trajectory repair module uses a reservation-aware dynamic window method to perform local repair and construct a local repair corridor in the conflict trajectory segment and its neighborhood. When obtaining the locally repaired trajectory, it specifically performs the following:

[0167] Based on the current longitudinal speed, maximum acceleration, maximum deceleration, current steering amount, maximum steering change rate, and discrete time step of the second intelligent logistics vehicle, determine the speed sampling window and steering sampling window;

[0168] Candidate control commands are sampled within the speed sampling window and the steering sampling window;

[0169] The local repair corridor is determined based on the conflict trajectory segment, the passable areas on both sides of the conflict trajectory segment, and the safe passage envelope of the target vehicle, thus obtaining the locally repaired trajectory corresponding to the second intelligent logistics vehicle.

[0170] Optionally, the three-dimensional spatiotemporal occupancy model includes: the vehicle body occupancy area and safety margin of each intelligent logistics vehicle in each time index or time interval;

[0171] The vehicle-occupied area in the three-dimensional spatiotemporal occupancy model can be adaptively expanded based on at least one of the following: vehicle positioning error, trajectory tracking error, communication delay, and trailer tail sweep effect.

[0172] The embodiments provided in this application Figure 12 The collaborative obstacle avoidance device for multiple intelligent logistics vehicles in the warehousing scenario shown can achieve... Figure 1 The various processes implemented in the method implementation examples will not be described again here to avoid repetition.

[0173] In one embodiment, the method of the present invention can also be implemented in the form of an electronic device, a computer-readable storage medium, or a computer program product. The electronic device includes a processor, a memory, and program instructions stored in the memory; when executed by the processor, the program instructions cause the electronic device to perform the multi-intelligent logistics vehicle cooperative obstacle avoidance method in a warehousing scenario described in this specification. The computer-readable storage medium or computer program product stores program instructions for implementing the method described in this specification. Since the same technical effects can be achieved, further details are omitted here to avoid repetition.

[0174] It should be noted that the electronic device in this application embodiment includes the server described above.

[0175] The processor is the processor in the electronic device described in the above embodiments. The readable storage medium includes computer-readable storage media, such as computer read-only memory (ROM), random access memory (RAM), magnetic disk, or optical disk.

[0176] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0177] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A collaborative obstacle avoidance method for multiple intelligent logistics vehicles in a warehousing scenario, applied to a collaborative planning system, characterized in that, The method includes: Obtain a map of the warehouse scene, operational task information, and the current and target status of each intelligent logistics vehicle; For each intelligent logistics vehicle, a global path search is performed based on a preset algorithm within a state space that includes the vehicle's pose state and at least one vehicle body relative pose state to obtain the basic path corresponding to the intelligent logistics vehicle that satisfies the vehicle's kinematic constraints. Based on the curvature of the base path, the relative attitude of the vehicle body, the obstacle clearance, the area speed limit, and the tracking error margin, a speed profile is allocated, and a global trajectory with timestamps is generated based on the allocated speed profile. The timestamped global trajectory of each intelligent logistics vehicle is mapped into a three-dimensional spatiotemporal occupancy model. The vehicle body occupancy area and safety margin of each intelligent logistics vehicle in each time index or time interval are recorded in the three-dimensional spatiotemporal occupancy model, and path conflicts between intelligent logistics vehicles are detected. Based on the number of path conflicts and the global trajectory length of each intelligent logistics vehicle, the first intelligent logistics vehicle that retains its time and space occupancy and the second intelligent logistics vehicle that does not retain its time and space occupancy are determined. For each second intelligent logistics vehicle, the conflict trajectory segment between the second intelligent logistics vehicle and the first intelligent logistics vehicle is determined, and the local repair corridor is constructed by the reservation perception dynamic window method in the conflict trajectory segment and the neighborhood to obtain the locally repaired trajectory. The repaired trajectories of each local area are input into the three-dimensional spatiotemporal occupancy model to update the three-dimensional spatiotemporal occupancy model; Based on the updated 3D spatiotemporal occupancy model, global trajectories are assigned to each intelligent logistics vehicle until all intelligent logistics vehicles have their spatiotemporal occupancy preserved, generating conflict-free collaborative obstacle avoidance trajectories for each intelligent logistics vehicle.

2. The method according to claim 1, characterized in that, The step of performing a global path search based on a preset algorithm for each intelligent logistics vehicle, within a state space that includes the vehicle's pose state and at least one relative vehicle body posture state, to obtain the basic path corresponding to the intelligent logistics vehicle that satisfies the vehicle's kinematic constraints, includes: For each intelligent logistics vehicle, in a state space containing the vehicle's pose state and at least one vehicle body relative attitude state, during the global path search based on the hybrid A* algorithm, collision detection is performed on the multi-rigid-body occupancy of the intelligent logistics vehicle, and each cost parameter is substituted into the global cost function to obtain the basic path corresponding to the intelligent logistics vehicle that satisfies the vehicle's kinematic constraints; wherein, the global cost function includes at least one of the following: path length cost, steering amplitude cost, steering change cost, reversing cost, articulation angle cost, folding knife approach cost, obstacle approach cost, and heuristic cost.

3. The method according to claim 1, characterized in that, The steps for establishing a local repair corridor using the reservation-aware dynamic window method in the conflict trajectory segment and its neighborhood to obtain the locally repaired trajectory include: The candidate control commands are simulated in the forward direction to eliminate candidate trajectories that collide with static obstacles or reserved spacetime occupancy, and the remaining candidate trajectories are obtained. Select a conflict-free candidate trajectory from the remaining candidate trajectories as the locally repaired trajectory; When no feasible candidate trajectory exists in the current prediction time domain, a safety waiting action is inserted for the second intelligent logistics vehicle, and the local repair is re-executed in the subsequent safe time interval.

4. The method according to claim 1, characterized in that: When the intelligent logistics vehicle is an articulated vehicle, the current state also includes the trailer heading or articulation angle, and the vehicle state is represented by the state vector s=[x, y, θ, β] or s=[x, y, θ, γ]. Where x and y represent the reference point position of the tractor, θ represents the heading angle of the tractor, β represents the heading angle of the trailer, and γ represents the articulation angle, and γ = θ - β; The control input of the vehicle is represented by the control quantity u=[v,δ], where v represents the longitudinal speed and δ represents the steering amount.

5. The method according to claim 2, characterized in that, The steps for collision detection of the multi-rigid-body occupancy of the intelligent logistics vehicle include: When the intelligent logistics vehicle is an articulated vehicle, the tractor and trailer of the intelligent logistics vehicle are respectively represented as envelopes of preset shapes, wherein the envelopes of preset shapes include: rectangular envelopes, polygonal envelopes, and approximate envelopes composed of multiple circular envelopes. It also samples and performs collision detection on multiple intermediate states of each motion primitive to detect static obstacle collisions, boundary crossings, and trailer tail-sweep collisions.

6. The method according to claim 1, characterized in that, The steps based on the curvature of the base path, vehicle relative attitude, obstacle clearance, area speed limit, and tracking error margin allocation speed profile include: The upper limit of the curvature constraint velocity is determined based on the local curvature of the basic path. The articulation state is determined based on the relative attitude of the vehicle body, and the upper limit of the attitude constraint speed is determined based on the articulation state. The upper limit of safe speed is determined based on the obstacle clearance, and the target speed of each trajectory point or trajectory segment is determined by combining the area speed limit and the tracking error margin; wherein, the target speed of each trajectory point or trajectory segment is the minimum value among the upper limit of curvature constraint speed, upper limit of attitude constraint speed, upper limit of safe speed, upper limit of area speed limit and upper limit of speed corresponding to tracking error margin; the speed profile is determined by the target speed of each trajectory point or trajectory segment.

7. The method according to claim 1, characterized in that, The three-dimensional spatiotemporal occupancy model is a three-dimensional spatiotemporal grid model and / or a spatiotemporal reservation table with time intervals; When the intelligent logistics vehicle is an articulated vehicle, the tractor-occupied area and the trailer-occupied area are recorded in each time index or time interval, and a safety margin is extended outside the occupied area.

8. The method according to claim 1, characterized in that, The step of using the reservation-aware dynamic window method to locally repair and construct a local repair corridor in the conflict trajectory segment and its neighborhood to obtain the locally repaired trajectory includes: Based on the current longitudinal speed, maximum acceleration, maximum deceleration, current steering amount, maximum steering change rate, and discrete time step of the second intelligent logistics vehicle, determine the speed sampling window and steering sampling window; Candidate control commands are sampled within the speed sampling window and the steering sampling window; The local repair corridor is determined based on the conflict trajectory segment, the passable areas on both sides of the conflict trajectory segment, and the safe passage envelope of the target vehicle, thus obtaining the locally repaired trajectory corresponding to the second intelligent logistics vehicle.

9. The method according to claim 1, characterized in that, The three-dimensional spatiotemporal occupancy model includes: the vehicle body occupancy area and safety margin of each intelligent logistics vehicle in each time index or time interval; The method further includes: adaptively expanding the vehicle body occupancy area in the three-dimensional spatiotemporal occupancy model based on at least one of vehicle positioning error, trajectory tracking error, communication delay, and trailer tail sweep effect.

10. A collaborative obstacle avoidance device for multiple intelligent logistics vehicles in a warehousing scenario, applied to a collaborative planning system, characterized in that, The device includes: The acquisition module is used to acquire the warehouse scene map, operation task information, and the current and target status of each intelligent logistics vehicle; The basic path determination module is used to perform a global path search based on a preset algorithm for each intelligent logistics vehicle, within a state space that includes the vehicle's pose state and at least one relative attitude state of the vehicle body, to obtain the basic path corresponding to the intelligent logistics vehicle that satisfies the vehicle's kinematic constraints. The global trajectory determination module is used to allocate speed profiles based on the curvature of the base path, the relative attitude of the vehicle body, the obstacle clearance, the area speed limit and the tracking error margin, and generate a global trajectory with timestamps based on the allocated speed profiles. The model building module is used to map the timestamped global trajectory of each intelligent logistics vehicle into a three-dimensional spatiotemporal occupancy model. In the three-dimensional spatiotemporal occupancy model, the vehicle body occupancy area and safety margin of each intelligent logistics vehicle in each time index or time interval are recorded, and path conflicts between intelligent logistics vehicles are detected. The filtering module is used to determine the first intelligent logistics vehicle to retain its time and space occupancy and the second intelligent logistics vehicle to not retain its time and space occupancy, based on the number of path conflicts and the global trajectory length of each intelligent logistics vehicle. The trajectory repair module is used to determine the conflict trajectory segment between the second intelligent logistics vehicle and the first intelligent logistics vehicle for each second intelligent logistics vehicle, and to use the reservation perception dynamic window method to perform local repair and build a local repair corridor in the conflict trajectory segment and its neighborhood to obtain the locally repaired trajectory. The iteration module is used to input the locally repaired trajectories into the three-dimensional spatiotemporal occupancy model to update the three-dimensional spatiotemporal occupancy model; based on the updated three-dimensional spatiotemporal occupancy model, global trajectories are assigned to each intelligent logistics vehicle until the spatiotemporal occupancy of each intelligent logistics vehicle is preserved, generating conflict-free cooperative obstacle avoidance trajectories for each intelligent logistics vehicle.

11. The apparatus according to claim 10, characterized in that, The basic path determination module is specifically used for: for each intelligent logistics vehicle, in a state space containing the vehicle pose state and at least one vehicle body relative attitude state, during the global path search based on the hybrid A* algorithm, to perform collision detection on the multi-rigid-body occupancy of the intelligent logistics vehicle, and to substitute each cost parameter into the global cost function to obtain the basic path corresponding to the intelligent logistics vehicle that satisfies the vehicle kinematic constraints; wherein, the global cost function includes at least one of the following: path length cost, steering amplitude cost, steering change cost, reversing cost, articulation angle cost, folding knife approach cost, obstacle approach cost, and heuristic cost.

12. The apparatus according to claim 10, characterized in that, The trajectory repair module includes: The first submodule is used to determine the conflict trajectory segments between the second intelligent logistics vehicle and the first intelligent logistics vehicle for each second intelligent logistics vehicle. The second submodule is used to perform forward simulation of candidate control commands, eliminate candidate trajectories that collide with static obstacles or reserved spacetime occupancy, and obtain the remaining candidate trajectories. The third submodule is used to select a conflict-free candidate trajectory from the remaining candidate trajectories as the locally repaired trajectory. The fourth submodule is used to insert a safety waiting action for the second intelligent logistics vehicle when there is no feasible candidate trajectory in the current prediction time domain, and to re-execute the local repair in the subsequent safe time interval.

13. The apparatus according to claim 10, characterized in that: When the intelligent logistics vehicle is an articulated vehicle, the current state also includes the trailer heading or articulation angle, and the vehicle state is represented by the state vector s=[x, y, θ, β] or s=[x, y, θ, γ]. Where x and y represent the reference point position of the tractor, θ represents the heading angle of the tractor, β represents the heading angle of the trailer, and γ represents the articulation angle, and γ = θ - β; The control input of the vehicle is represented by the control quantity u=[v,δ], where v represents the longitudinal speed and δ represents the steering amount.

14. The apparatus according to claim 11, characterized in that, When the basic path determination module performs collision detection on the multi-rigid-body occupancy of the intelligent logistics vehicle, it is specifically used for: When the intelligent logistics vehicle is an articulated vehicle, the tractor and trailer of the intelligent logistics vehicle are respectively represented as envelopes of preset shapes, wherein the envelopes of preset shapes include: rectangular envelopes, polygonal envelopes, and approximate envelopes composed of multiple circular envelopes. It also samples and performs collision detection on multiple intermediate states of each motion primitive to detect static obstacle collisions, boundary crossings, and trailer tail-sweep collisions.

15. The apparatus according to claim 10, characterized in that, The global trajectory determination module is specifically used for: The upper limit of the curvature constraint velocity is determined based on the local curvature of the basic path. The articulation state is determined based on the relative attitude of the vehicle body, and the upper limit of the attitude constraint speed is determined based on the articulation state. The upper limit of safe speed is determined based on the obstacle clearance, and the target speed of each trajectory point or trajectory segment is determined by combining the area speed limit and the tracking error margin; wherein, the target speed of each trajectory point or trajectory segment is the minimum value among the upper limit of curvature constraint speed, upper limit of attitude constraint speed, upper limit of safe speed, upper limit of area speed limit and upper limit of speed corresponding to tracking error margin; the speed profile is determined by the target speed of each trajectory point or trajectory segment.

16. The apparatus according to claim 10, characterized in that, The three-dimensional spatiotemporal occupancy model is a three-dimensional spatiotemporal grid model and / or a spatiotemporal reservation table with time intervals; When the intelligent logistics vehicle is an articulated vehicle, the tractor-occupied area and the trailer-occupied area are recorded in each time index or time interval, and a safety margin is extended outside the occupied area.

17. The apparatus according to claim 10, characterized in that, The trajectory repair module uses a reservation-aware dynamic window method to perform local repair and construct a local repair corridor in the conflict trajectory segment and its neighborhood. Specifically, when obtaining the locally repaired trajectory, it is used for: Based on the current longitudinal speed, maximum acceleration, maximum deceleration, current steering amount, maximum steering change rate, and discrete time step of the second intelligent logistics vehicle, determine the speed sampling window and steering sampling window; Candidate control commands are sampled within the speed sampling window and the steering sampling window; The local repair corridor is determined based on the conflict trajectory segment, the passable areas on both sides of the conflict trajectory segment, and the safe passage envelope of the target vehicle, thus obtaining the locally repaired trajectory corresponding to the second intelligent logistics vehicle.

18. The apparatus according to claim 10, characterized in that, The three-dimensional spatiotemporal occupancy model includes: the vehicle body occupancy area and safety margin of each intelligent logistics vehicle in each time index or time interval; The vehicle-occupied area in the three-dimensional spatiotemporal occupancy model can be adaptively expanded based on at least one of the following: vehicle positioning error, trajectory tracking error, communication delay, and trailer tail sweep effect.

Images