AGV (Automatic Guided Vehicle) system and method based on multi-dimensional conflict pre-judgment and dynamic task scheduling

Abstract

The invention discloses an AGV system based on multi-dimensional conflict pre-judgment and dynamic task scheduling, and the system comprises a task scheduling module which is used for dynamically generating a transportation task according to the logic cargo quantity of a cargo carrying point, and distributing the task to a plurality of AGVs; the path planning module is used for planning a path for each AGV by adopting an improved A * path-finding algorithm, and the improved A * path-finding algorithm is integrated with physical behavior simulation of the AGV and is used for carrying out path search based on dynamic environment information; and the conflict management module is used for performing multi-dimensional conflict pre-judgment before the AGV executes movement, and the multi-dimensional conflict pre-judgment at least comprises target point occupation pre-judgment and next coordinate conflict pre-judgment. According to the technical scheme of the invention, by integrating dynamic environment perception, physical behavior simulation, multi-dimensional conflict pre-judgment and intelligent task scheduling, the AGV cluster can effectively avoid collision deadlock, generate a feasible smooth path and realize load balancing, so that the traffic efficiency, the safety and the resource utilization efficiency of the system in a high-density environment are comprehensively improved.

Description

Technical Field

[0001] This invention relates to the field of intelligent warehousing and logistics automation technology, and more specifically, to an AGV system and method based on multi-dimensional conflict prediction and dynamic task scheduling. Background Technology

[0002] With the deepening of Industry 4.0 and intelligent manufacturing, Automated Guided Vehicle (AGV) systems have become a core component of modern intelligent warehousing and factory logistics automation. Traditional AGV systems rely heavily on classic static algorithms such as A* and Dijkstra for path planning. While these methods are applicable in single AGVs or simple static environments, in complex dynamic scenarios with high-density deployment of multiple AGVs, their path planning often fails to consider other AGVs as real-time obstacles, easily leading to path intersections, mutual obstruction, and even system deadlock, severely restricting the system's efficiency and safety.

[0003] Furthermore, existing technologies do not adequately consider the physical characteristics of AGVs themselves. In actual operation, the starting, acceleration, deceleration, and turning behaviors of AGVs are all subject to dynamic constraints such as inertia and minimum turning radius. However, traditional algorithms often plan ideal geometric paths, ignoring these physical limitations, leading to a disconnect between planning and execution, which affects the smoothness, stability, and energy efficiency of AGV movement. At the same time, at the task scheduling level, simple strategies such as first-come, first-served or shortest path are commonly used, lacking awareness of global material requirements, real-time load of each node, and traffic conditions. This easily leads to problems such as uneven task allocation, high AGV empty-running rates, and insufficient utilization of system resources.

[0004] Currently, while some studies attempt to resolve conflicts or optimize paths by improving single algorithms (such as grid-based collision avoidance or ant colony optimization), these solutions mostly focus on local improvements and fail to construct a collaborative control framework at the system level that integrates intelligent task allocation, physical constraint integration, proactive conflict prediction, and dynamic scheduling. Therefore, when facing complex operating environments with high density and high dynamics, existing systems still have significant shortcomings in terms of flexibility, adaptability, and overall efficiency, urgently requiring a comprehensive solution. Summary of the Invention

[0005] In view of the above-mentioned technical problems in related technologies, this invention proposes an AGV system and method based on multi-dimensional conflict prediction and dynamic task scheduling, which can overcome the above-mentioned shortcomings of the prior art.

[0006] To achieve the above-mentioned technical objectives, the technical solution of the present invention is implemented as follows: An AGV system based on multi-dimensional conflict prediction and dynamic task scheduling; The AGV system based on multi-dimensional conflict prediction and dynamic task scheduling includes a task scheduling module, which is used to dynamically generate transportation tasks according to the logical quantity of goods at the loading point and assign the tasks to multiple AGVs. The path planning module uses an improved A* pathfinding algorithm to plan paths for each AGV. The improved A* pathfinding algorithm integrates the physical behavior simulation of the AGV and performs path search based on dynamic environmental information. The dynamic environment information is incorporated into the path cost calculation through a dynamic cost map. The total cost calculation formula for the dynamic cost map is as follows: C_total(x,y)=C_static(x,y)+η*D(x,y), Where C_total(x,y) is the total cost of position (x,y), C_static(x,y) is the static map cost of position (x,y), D(x,y) is the dynamic obstacle density function, and η is the dynamic cost weight coefficient with η greater than 1, so as to prioritize avoiding dynamic obstacles. The conflict management module is used to perform multi-dimensional conflict prediction before the AGV moves. The multi-dimensional conflict prediction includes at least target point occupation prediction and next coordinate conflict prediction.

[0007] Furthermore, the multi-dimensional conflict prediction also includes a deceleration and yielding strategy and a traffic jam adaptive mechanism; the deceleration and yielding strategy includes forcing other AGVs to wait outside the deceleration area when an AGV enters the deceleration area; the traffic jam adaptive mechanism includes triggering path replanning or forced waiting operation when the number of AGVs that have not moved exceeds a preset threshold.

[0008] Furthermore, the physical behavior simulation includes simulating the acceleration, deceleration and turning processes of the AGV, and optimizing path feasibility by introducing at least one of turning time penalty, acceleration / deceleration time cost and energy consumption cost in path planning; the conflict management module predicts the trajectory conflict of the AGV in future time steps by constructing a spatiotemporal conflict probability map, and adjusts the path planning based on the conflict probability.

[0009] Furthermore, the task scheduling module is configured to prioritize the loading point with the most logical cargo based on the logical cargo quantity at the loading point as the task starting point. The target loading point P_target is determined by the formula: P_target=argmax_{p∈P}(L_logic(p)), where P is the set of all loading points and L_logic(p) is the logical cargo inventory at loading point p. The module also assigns single-destination or multi-destination unloading tasks to the AGV. The system also includes a cargo management module for real-time management of cargo quantities at loading and unloading points and for dynamically triggering replenishment operations from the distribution center to the loading points based on cargo status.

[0010] Furthermore, the system also includes a visualization module for displaying the AGV's location, status, cargo quantity, and system operation data in real time; the task scheduling module also receives real-time traffic density information from the conflict management module and dynamically adjusts the task allocation strategy based on this information to avoid assigning tasks to congested areas.

[0011] According to another aspect of the present invention, a working method for an AGV system based on multi-dimensional conflict prediction and dynamic task scheduling is provided; The working method of this AGV system based on multi-dimensional conflict prediction and dynamic task scheduling includes the following steps: The system initialization steps include setting the work area map, the initial position of the AGV, and the initial state of the goods. The task generation and allocation steps involve real-time statistics of the logical cargo quantity at each loading point, dynamic generation of transportation tasks based on the logical cargo quantity, and allocation of tasks to idle AGVs. The path planning and conflict prediction steps employ an improved A* pathfinding algorithm to plan paths for each AGV, and perform multi-dimensional conflict prediction before the AGV moves. The multi-dimensional conflict prediction includes at least target point occupation prediction and next-step coordinate conflict prediction. The AGV movement and status update steps involve controlling the movement of the AGV and updating its position and status in real time based on the path planning results and conflict prediction results.

[0012] Furthermore, the multi-dimensional conflict prediction also includes a deceleration and yielding strategy for entering the station and a traffic jam adaptive step; The deceleration and yielding strategy includes forcing other AGVs to wait outside the deceleration zone when an AGV enters the deceleration zone. The traffic jam adaptive steps include triggering a path replanning or forced waiting operation when the AGV's inactive count exceeds a preset threshold.

[0013] Furthermore, in the path planning and conflict prediction steps, the improved A* pathfinding algorithm integrates at least one of turning penalty, acceleration / deceleration time cost, and energy consumption cost when calculating path cost; and predicts the trajectory conflict of the AGV in future time steps by constructing a spatiotemporal conflict probability map, and adjusts the path planning based on the conflict probability.

[0014] Furthermore, in the task generation and allocation step, the loading point with the most logical cargo is selected as the task starting point based on the logical cargo quantity at the loading point. The target loading point P_target is determined by the formula: P_target=argmax_{p∈P}(L_logic(p)), where P is the set of all loading points and L_logic(p) is the logical cargo quantity at loading point p. Single-destination or multi-destination unloading tasks are assigned to the AGV. The method also includes a cargo management step, which monitors the cargo quantity at loading and unloading points in real time and dynamically triggers replenishment operations from the distribution center to the loading points.

[0015] Furthermore, the system includes a visualization step, which displays the AGV's location, status, cargo quantity, and key system operating indicators in real time. In the task generation and allocation step, the task allocation strategy is dynamically adjusted based on real-time traffic density information to avoid assigning tasks to congested areas.

[0016] The beneficial effects of this invention are as follows: By constructing a collaborative control system that deeply integrates dynamic environment perception, AGV physical behavior simulation, multi-dimensional conflict prediction, and intelligent task scheduling, the AGV cluster can effectively avoid collisions and deadlocks, plan physically feasible and efficient smooth paths, and achieve dynamic load balancing of the system. This significantly improves the overall passage efficiency, operational safety, resource utilization, and adaptive capabilities of multi-AGV systems in complex and high-density environments. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is an overall architecture diagram of an AGV system based on multi-dimensional conflict prediction and dynamic task scheduling according to an embodiment of the present invention; Figure 2 This is a logical block diagram of the working method of an AGV system based on multi-dimensional conflict prediction and dynamic task scheduling according to an embodiment of the present invention. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] like Figure 1 As shown in the figure, an AGV system based on multi-dimensional conflict prediction and dynamic task scheduling according to an embodiment of the present invention includes multiple AGVs and a control system, wherein the control system includes: The task scheduling module is used to dynamically generate transportation tasks based on the logical quantity of goods at the loading point and assign the tasks to multiple AGVs. The path planning module uses an improved A* pathfinding algorithm to plan paths for each AGV. The improved A* pathfinding algorithm integrates the physical behavior simulation of the AGV and performs path search based on dynamic environmental information. The dynamic environment information is incorporated into the path cost calculation through a dynamic cost map. The total cost calculation formula for the dynamic cost map is as follows: C_total(x,y)=C_static(x,y)+η*D(x,y), Where C_total(x,y) is the total cost of position (x,y), C_static(x,y) is the static map cost of position (x,y), D(x,y) is the dynamic obstacle density function, and η is the dynamic cost weight coefficient with η greater than 1, so as to prioritize avoiding dynamic obstacles. The conflict management module is used to perform multi-dimensional conflict prediction before the AGV moves. The multi-dimensional conflict prediction includes at least target point occupation prediction and next coordinate conflict prediction.

[0021] According to an embodiment of the present invention, an AGV system based on multidimensional conflict prediction and dynamic task scheduling is provided. In a specific embodiment, the multidimensional conflict prediction further includes a deceleration and entry yielding strategy and a traffic jam adaptive mechanism. The deceleration and entry yielding strategy includes forcing other AGVs to wait outside the deceleration area when an AGV enters the deceleration area. The traffic jam adaptive mechanism includes triggering path replanning or forced waiting operation when the AGV's inactive count exceeds a preset threshold.

[0022] According to an embodiment of the present invention, an AGV system based on multi-dimensional conflict prediction and dynamic task scheduling is provided. In a specific embodiment, the physical behavior simulation includes simulating the acceleration, deceleration and turning processes of the AGV, and optimizing path feasibility by introducing at least one of turning time penalty, acceleration / deceleration time cost and energy consumption cost in path planning; the conflict management module predicts the trajectory conflict of the AGV in future time steps by constructing a spatiotemporal conflict probability map, and adjusts the path planning based on the conflict probability.

[0023] According to an embodiment of the present invention, an AGV system based on multi-dimensional conflict prediction and dynamic task scheduling is provided. In a specific embodiment, the task scheduling module is further configured to prioritize the loading point with the most logical cargo based on the logical cargo quantity at the loading point as the task starting point. The target loading point P_target is determined by the formula: P_target=argmax_{p∈P}(L_logic(p)), where P is the set of all loading points and L_logic(p) is the logical cargo quantity at loading point p. The system also assigns single-destination or multi-destination unloading tasks to the AGV. The system further includes a cargo management module for real-time management of the cargo quantity at loading and unloading points and for dynamically triggering replenishment operations from the distribution center to the loading point based on the cargo status.

[0024] According to an embodiment of the present invention, an AGV system based on multi-dimensional conflict prediction and dynamic task scheduling is provided. In a specific embodiment, the system further includes a visualization module for displaying the AGV's location, status, cargo quantity, and system operation data in real time. The task scheduling module also receives real-time traffic density information from the conflict management module and dynamically adjusts the task allocation strategy based on this information to avoid assigning tasks to congested areas.

[0025] Secondly, such as Figure 2 As shown in the figure, the working method of an AGV system based on multi-dimensional conflict prediction and dynamic task scheduling according to an embodiment of the present invention includes the following steps: The system initialization steps include setting the work area map, the initial position of the AGV, and the initial state of the goods. The task generation and allocation steps involve real-time statistics of the logical cargo quantity at each loading point, dynamic generation of transportation tasks based on the logical cargo quantity, and allocation of tasks to idle AGVs. The path planning and conflict prediction steps employ an improved A* pathfinding algorithm to plan paths for each AGV, and perform multi-dimensional conflict prediction before the AGV moves. The multi-dimensional conflict prediction includes at least target point occupation prediction and next-step coordinate conflict prediction. The AGV movement and status update steps involve controlling the movement of the AGV and updating its position and status in real time based on the path planning results and conflict prediction results.

[0026] According to an embodiment of the present invention, a working method of an AGV system based on multidimensional conflict prediction and dynamic task scheduling is provided. In a specific embodiment, the multidimensional conflict prediction further includes a deceleration and station entry yielding strategy and a traffic jam adaptive step. The deceleration and yielding strategy includes forcing other AGVs to wait outside the deceleration zone when an AGV enters the deceleration zone. The traffic jam adaptive steps include triggering a path replanning or forced waiting operation when the AGV's inactive count exceeds a preset threshold.

[0027] According to an embodiment of the present invention, a working method for an AGV system based on multidimensional conflict prediction and dynamic task scheduling is described. In a specific embodiment, in the path planning and conflict prediction steps, the improved A* pathfinding algorithm integrates at least one of turning penalty, acceleration / deceleration time cost, and energy consumption cost when calculating path cost; and predicts the trajectory conflict of the AGV in future time steps by constructing a spatiotemporal conflict probability map, and adjusts the path planning based on the conflict probability.

[0028] According to an embodiment of the present invention, a working method for an AGV system based on multi-dimensional conflict prediction and dynamic task scheduling is described. In a specific embodiment, the task generation and allocation step further selects the loading point with the most logical cargo as the task starting point based on the logical cargo quantity at the loading point. The target loading point P_target is determined by the formula: P_target=argmax_{p∈P}(L_logic(p)), where P is the set of all loading points and L_logic(p) is the logical cargo quantity at loading point p. Single-destination or multi-destination unloading tasks are assigned to the AGV. The method also includes a cargo management step, which monitors the cargo quantity at loading and unloading points in real time and dynamically triggers replenishment operations from the distribution center to the loading points.

[0029] According to an embodiment of the present invention, a working method for an AGV system based on multi-dimensional conflict prediction and dynamic task scheduling includes, in a specific embodiment, a visualization step, which displays the AGV's location, status, cargo quantity, and key system operating indicators in real time; in the task generation and allocation step, the task allocation strategy is also dynamically adjusted based on real-time traffic density information to avoid assigning tasks to congested areas.

[0030] To facilitate understanding of the above technical solutions of the present invention, the following detailed description of the above technical solutions of the present invention is provided through specific embodiments and examples.

[0031] In practical application, the AGV system based on multi-dimensional conflict prediction and dynamic task scheduling according to the present invention includes multiple AGVs and a centralized control system. The system adopts a hierarchical architecture design, mainly including a task scheduling module, a path planning module, a conflict management module, a cargo management module, and a visualization module. These modules work collaboratively to achieve intelligent scheduling and coordinated control of the AGV cluster.

[0032] The task scheduling module is responsible for dynamically generating transportation tasks based on the logical cargo quantity at each loading point. In practice, this module continuously monitors the logical cargo inventory at each loading point (i.e., goods committed but not physically retrieved by the AGV). When the AGV is idle, it prioritizes the loading point with the most logical cargo as the task start point. Simultaneously, this module supports assigning single-destination or multi-destination unloading tasks to the AGV. For example, when the cargo quantity at a loading point is greater than or equal to 2 items, there is a 90% probability of assigning a dual-destination task and a 10% probability of assigning a single-destination task, thus achieving intelligent task batch processing. This module also receives real-time traffic density information (represented by a congestion index of 0-100%) from the conflict management module. When it detects that the traffic density in a certain area exceeds the 70% threshold, it dynamically adjusts the task allocation strategy to avoid assigning tasks to congested areas.

[0033] The path planning module employs a deeply improved A* pathfinding algorithm as its core. This algorithm integrates simulation of the AGV's physical behavior and performs path search based on dynamic environmental information. In its implementation, the algorithm optimizes path feasibility by introducing turning time penalties, acceleration / deceleration time costs, and energy costs when calculating path costs. The comprehensive G-value calculation formula is: G_new(current,next)=Dist(current,next)+α*Time_penalty+β*Energy_penalty + γ*Turn_penalty, where α, β, and γ are configurable weighting coefficients used to balance time, energy consumption, and turning costs. The H-value calculation considers the path's "reachability" and "smoothness" estimations, avoiding the selection of paths with short geometric distances but low actual execution efficiency. The generated discrete path points are converted into continuous and smooth motion trajectories using a trajectory smoother (such as B-spline curve fitting) to ensure compliance with the AGV's minimum turning radius and maximum acceleration limits.

[0034] The conflict management module performs multi-dimensional conflict prediction before the AGV executes a movement. Specifically, this includes: (1) Target point occupancy prediction: Determine whether the target location is occupied by other AGVs or the unloading point is closed. If it is occupied for a long time, guide the AGV into the waiting queue or queue at the temporary parking point; (2). Next step coordinate conflict prediction: After all AGVs submit the next coordinates, a central arbitration is conducted, taking into account the AGV volume and safety margin, and resolving conflicts according to the preset priority (e.g., cargo AGVs take priority over empty AGVs); (3) Deceleration and yielding strategy: When an AGV enters the deceleration zone, other AGVs are forced to wait outside the zone; (4) Traffic jam adaptive mechanism: When the noMoveCount of the AGV exceeds the threshold of 25 time steps, or the average speed in the local area is lower than the safety threshold, path replanning is triggered. This module also predicts trajectory conflicts in the next N time steps by constructing a spatiotemporal conflict probability map (3D X, Y, Time grid), and actively adjusts the path planning when the conflict probability exceeds the threshold.

[0035] The cargo management module manages the quantity and status of goods at loading and unloading points in real time. In practice, the distribution center initially has 5,000 items, with each loading point having a maximum capacity of 30 items. The system checks the cargo status every 60 seconds. If the distribution center has goods and the loading points are not full, it prioritizes replenishing the loading points with the fewest items to ensure a balanced distribution of goods. The replenishment strategy adopts a distribution from the center outwards to both ends to optimize replenishment efficiency.

[0036] The visualization module displays the AGV's location, status (distinguished by color changes: light blue indicates normal driving, pink indicates carrying a task, purple indicates waiting with cargo, orange indicates waiting to unload, and red indicates a conflict), cargo quantity (displayed in the format of "logical cargo quantity|physical cargo quantity"), and key system operating indicators (such as the number of transported goods, the number of busy AGVs, and the system running time).

[0037] System Workflow The system's operating method includes the following steps: System initialization: Set a 10x10 work area map, initial AGV positions (5 AGVs, initial speed 0, maximum speed 2m / s), loading points (5), unloading points (10), and initial cargo status of the distribution center.

[0038] Task generation and allocation: Real-time statistics of the logical cargo quantity at each loading point are used to dynamically generate transportation tasks. When the AGV is idle, the task scheduling module selects the task starting point based on the load balancing principle (P_target = argmax_{p ∈ P} (L_logic(p))) and intelligently allocates single-destination or multi-destination unloading tasks according to the cargo quantity.

[0039] Path planning and conflict prediction: Each AGV uses an improved A* algorithm to plan its path. Before moving, the conflict management module performs multi-dimensional conflict prediction: checking the target point's occupancy; detecting next-step coordinate conflicts and forcing some AGVs to wait if necessary; managing the deceleration entry area; monitoring traffic congestion and triggering replanning. The algorithm integrates physical behavior simulation; when an AGV starts, turns, or decelerates, it triggers lowSpeedFlags and enters a 3-second acceleration / deceleration waiting countdown to simulate the actual physical process.

[0040] AGV Movement and Status Update: Control AGV movement based on planning results. When an AGV arrives at a loading or unloading point, it enters a 3-second countdown to simulate the loading and unloading process. After completion, update the quantity of goods (logical goods decrease at the loading point, physical goods increase at the unloading point) and the AGV status.

[0041] System monitoring and loop execution: The visualization module updates the system status in real time. The system continuously loops through steps 2 to 5 until the task is completed or the system stops.

[0042] Core Algorithm Implementation The system scheduling logic is executed iteratively through the main loop. The pseudocode framework is as follows: WHILE system running DO Mark the current position of all AGVs as dynamic obstacles. FOR each AGV DO IF target point is not occupied THEN Execute the A* pathfinding algorithm (integrating physical behavior constraints). IF Spatiotemporal conflict detected THEN AGV waits in place, stationary_counter++ ELSE Record the expected movement location END IF END IF END FOR Implement deadlock resolution mechanism (trigger replanning when stationary_counter > 25). Execute material allocation and replenishment linkage Update AGV status and visualization END WHILE The core formula in the core algorithm is as follows: The formula for the dynamic cost map is: C_total(x,y)=C_static(x,y)+η*D(x,y) Where C_total(x,y): total cost of location (x,y), C_static(x,y): static map cost of location (x,y) (e.g., fixed obstacles), D(x,y): dynamic obstacle density function, η: dynamic cost weight coefficient (usually η>>1, to prioritize avoiding dynamic obstacles).

[0043] The formula for kinematic integration is: F_new=F_original+P_turn*δ(Δθ) Where F_new: the new evaluation function value, and F_original: the original A* evaluation value (F=G+H). P_turn: Penalty coefficient per unit turning angle, δ(Δθ): Direction change function, Δθ=|θ_current-θ_previous|.

[0044] The formula for collaborative conflict resolution is: Conflict(i,j,t)={1,if nextPos_i(t)==nextPos_j(t) {0,otherwise Where Conflict(i,j,t): the conflict state of AGV i and j at time t, and nextPos_i(t): the next position of AGV i at time t.

[0045] The formula for the safe distance approach strategy is: Action(s,g)={Approach,if D(s,g)>d_safe {Align / Wait,if D(s,g)≤d_safe Where D(s,g): the Euclidean distance from the current position s to the target position g, and d_safe: the system safety distance parameter.

[0046] The formula for load balancing task allocation is: P_target=argmax_{p∈P}(L_logic(p)) Wherein, P_target: the selected target loading point, P: the set of all loading points, and L_logic(p): the logical cargo inventory of loading point p.

[0047] The formula for system throughput efficiency is: η_throughput=N_completed / (T_simulation×N_AGV) Where η_throughput: average number of tasks completed per AGV per unit time, N_completed: total number of tasks completed during the simulation, T_simulation: total simulation time, and N_AGV: total number of AGVs.

[0048] The present invention will be described in detail below with reference to the embodiments, but the scope of protection of the present invention is not limited to the following embodiments.

[0049] Example 1: Simulation of AGV Physical Behavior and Conflict Prediction This embodiment demonstrates the basic operating mechanism of the AGV system. The system operates on a 10×10 simulated map, deploying 5 AGVs. Each AGV has an initial speed of 0, a maximum speed of 2 m / s, and an acceleration / deceleration time set to 3 seconds.

[0050] Physical behavior simulation implementation: When the AGV starts from a standstill, is about to reach the target point (1 unit away from the target point), or its direction of movement is different from the previous step, the AGV's lowSpeedFlags are set to true, and the lowSpeedWait countdown begins. During this countdown, the AGV remains stationary in its current position, simulating the actual acceleration or deceleration process.

[0051] Path planning implementation: The path planning module employs an improved A-pathfinding algorithm to plan paths for each AGV. This algorithm performs path search based on dynamic environmental information, which is incorporated into path cost calculation through a dynamic cost map. The total cost calculation formula for the dynamic cost map is: C_total(x,y)=C_static(x,y)+ηD(x,y), where η is greater than 1 to prioritize avoiding dynamic obstacles.

[0052] Conflict prediction implementation: The conflict management module performs multi-dimensional conflict prediction before the AGV moves. Target point occupancy prediction: If the target position of AGV1 (such as loading point A) has been occupied by AGV2, then AGV1 will stop and wait at a safe distance; Next step coordinate conflict prediction: If both AGV3 and AGV4 are planned to move to coordinates (X,Y) in the next step, the system will force AGV3 to maintain its current position and allow AGV4 to move; Deceleration and yielding: If AGV5 is decelerating into unloading point B, its previous position (X',Y') is considered the "deceleration and yielding area". At this time, even if the next step planned by AGV6 is (X',Y'), the system will force AGV6 to wait. Traffic jam adaptive: If AGV1 fails to move for 25 consecutive time steps (noMoveCount exceeds 25), the system will trigger it to replan its route or force it to select a temporary parking point.

[0053] Example 2: Dynamic Task Scheduling and Cargo Management This embodiment demonstrates the system's task scheduling and cargo management functions. The system has 5 loading points and 10 unloading points. The distribution center initially has 5,000 pieces of cargo, and each loading point has a maximum load of 30 pieces.

[0054] Cargo Management Implementation: The cargo management module manages the cargo quantity at loading and unloading points in real time. The system checks the status of distribution centers and loading points every 60 seconds. If the distribution center has cargo and the loading point is not full, the system replenishes cargo from the distribution center to the loading point, prioritizing the loading point in the middle and spreading outwards to both ends.

[0055] Task scheduling implementation: The task scheduling module dynamically generates transportation tasks based on the logical quantity of goods at each loading point. When the AGV completes its task or becomes idle, the task scheduling module iterates through all loading points, prioritizing the loading point with the most logical goods as the task start point. The target loading point P_target is determined by the formula: P_target = argmax_{p∈P}(L_logic(p)). After the AGV retrieves goods from the loading point, the system assigns it two different unloading point tasks based on the current quantity of goods at the loading point (if the quantity is greater than 2 pieces); otherwise, it assigns a single-destination task.

[0056] Status Management: After the AGV arrives at the loading or unloading point, it enters a 3-second waiting countdown, during which the AGV status color changes to purple (loading) or orange (unloading).

[0057] Example 3: System Visualization and Data Monitoring This embodiment demonstrates the system's visual monitoring function.

[0058] Status display implementation: The visualization module displays the AGV's position and status in real time. The AGV's color changes dynamically according to its status: light blue indicates normal operation, pink indicates carrying a task, purple indicates waiting at the loading point, orange indicates waiting at the unloading point, and red indicates a conflict or blockage.

[0059] Data shows that the loading point displays the current cargo quantity in the form of "logical cargo quantity | physical cargo quantity", while the unloading point displays the current accumulated cargo quantity.

[0060] Real-time monitoring: The interface displays key performance indicators such as the number of goods handled, the number of busy AGVs, the total number of goods, the number of replenishment times at the distribution center, and the system uptime in real time.

[0061] Through the implementation of the above embodiments, the present invention achieves efficient and conflict-free operation of the AGV system in a high-density environment, significantly improving the overall efficiency and safety of the system.

[0062] In summary, by utilizing the technical solutions described above in this invention, a collaborative control system that deeply integrates dynamic environment perception, AGV physical behavior simulation, multi-dimensional conflict prediction, and intelligent task scheduling is constructed. This enables AGV clusters to effectively avoid collisions and deadlocks, plan physically feasible and efficient smooth paths, and achieve dynamic load balancing of the system. Consequently, the overall traffic efficiency, operational safety, resource utilization, and adaptive capabilities of multi-AGV systems in complex and high-density environments are significantly improved.

[0063] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. An AGV system based on multi-dimensional conflict prediction and dynamic task scheduling, characterized in that, It includes multiple AGVs and a control system, the control system including: The task scheduling module is used to dynamically generate transportation tasks based on the logical quantity of goods at the loading point and assign the tasks to multiple AGVs. The path planning module uses an improved A* pathfinding algorithm to plan paths for each AGV. The improved A* pathfinding algorithm integrates the physical behavior simulation of the AGV and performs path search based on dynamic environmental information. The dynamic environment information is incorporated into the path cost calculation through a dynamic cost map. The total cost calculation formula for the dynamic cost map is as follows: C_total(x,y)=C_static(x,y)+η*D(x,y), Where C_total(x,y) is the total cost of position (x,y), C_static(x,y) is the static map cost of position (x,y), D(x,y) is the dynamic obstacle density function, and η is the dynamic cost weight coefficient with η greater than 1, so as to prioritize avoiding dynamic obstacles. The conflict management module is used to perform multi-dimensional conflict prediction before the AGV moves. The multi-dimensional conflict prediction includes at least target point occupation prediction and next coordinate conflict prediction.

2. The AGV system based on multi-dimensional conflict prediction and dynamic task scheduling according to claim 1, characterized in that, The multi-dimensional conflict prediction also includes a deceleration and yielding strategy and a traffic jam adaptive mechanism; the deceleration and yielding strategy includes forcing other AGVs to wait outside the deceleration area when an AGV enters the deceleration area; the traffic jam adaptive mechanism includes triggering path replanning or forced waiting operation when the AGV's inactive count exceeds a preset threshold.

3. The AGV system based on multi-dimensional conflict prediction and dynamic task scheduling according to claim 1, characterized in that, The physical behavior simulation includes simulating the acceleration, deceleration and turning processes of the AGV, and optimizing path feasibility by introducing at least one of turning time penalty, acceleration and deceleration time cost and energy consumption cost in path planning; the conflict management module predicts the trajectory conflict of the AGV in future time steps by constructing a spatiotemporal conflict probability map, and adjusts the path planning based on the conflict probability.

4. The AGV system based on multi-dimensional conflict prediction and dynamic task scheduling according to claim 1, characterized in that, The task scheduling module is also configured to prioritize the loading point with the most logical cargo based on the number of logical cargo items at the loading point as the task starting point. The target loading point P_target is determined by the formula: P_target=argmax_{p∈P}(L_logic(p)), where P is the set of all loading points and L_logic(p) is the logical cargo inventory at loading point p. The module also assigns single-destination or multi-destination unloading tasks to the AGV. The system also includes a cargo management module, which manages the cargo quantity at loading and unloading points in real time and dynamically triggers replenishment operations from the distribution center to the loading point based on the cargo status.

5. The AGV system based on multi-dimensional conflict prediction and dynamic task scheduling according to claim 1, characterized in that, The system also includes a visualization module for displaying the AGV's location, status, cargo quantity, and system operation data in real time; the task scheduling module also receives real-time traffic density information from the conflict management module and dynamically adjusts the task allocation strategy based on this information to avoid assigning tasks to congested areas.

6. A working method for an AGV system based on multi-dimensional conflict prediction and dynamic task scheduling, characterized in that, Includes the following steps: The system initialization steps include setting the work area map, the initial position of the AGV, and the initial state of the goods. The task generation and allocation steps involve real-time statistics of the logical cargo quantity at each loading point, dynamic generation of transportation tasks based on the logical cargo quantity, and allocation of tasks to idle AGVs. The path planning and conflict prediction steps employ an improved A* pathfinding algorithm to plan paths for each AGV, and perform multi-dimensional conflict prediction before the AGV moves. The multi-dimensional conflict prediction includes at least target point occupation prediction and next-step coordinate conflict prediction. The AGV movement and status update steps involve controlling the movement of the AGV and updating its position and status in real time based on the path planning results and conflict prediction results.

7. The working method of an AGV system based on multi-dimensional conflict prediction and dynamic task scheduling according to claim 6, characterized in that, The multi-dimensional conflict prediction also includes a deceleration and yielding strategy for entering the station and a traffic jam adaptive step. The deceleration and yielding strategy includes forcing other AGVs to wait outside the deceleration zone when an AGV enters the deceleration zone. The traffic jam adaptive steps include triggering a path replanning or forced waiting operation when the AGV's inactive count exceeds a preset threshold.

8. The working method of an AGV system based on multi-dimensional conflict prediction and dynamic task scheduling according to claim 6, characterized in that, In the path planning and conflict prediction steps, the improved A* pathfinding algorithm integrates at least one of turning penalty, acceleration / deceleration time cost, and energy consumption cost when calculating path cost; and predicts the trajectory conflict of the AGV in future time steps by constructing a spatiotemporal conflict probability map, and adjusts the path planning based on the conflict probability.

9. The working method of an AGV system based on multi-dimensional conflict prediction and dynamic task scheduling according to claim 6, characterized in that, In the task generation and allocation steps, the loading point with the most logical cargo is selected as the task starting point based on the logical cargo quantity at the loading point. The target loading point P_target is determined by the formula: P_target=argmax_{p∈P}(L_logic(p)), where P is the set of all loading points and L_logic(p) is the logical cargo inventory at loading point p. Single-destination or multi-destination unloading tasks are assigned to the AGV. The method also includes a cargo management step, which monitors the cargo quantity at loading and unloading points in real time and dynamically triggers replenishment operations from the distribution center to the loading points.

10. The working method of an AGV system based on multi-dimensional conflict prediction and dynamic task scheduling according to claim 6, characterized in that, The system includes a visualization step, which displays the AGV's location, status, cargo quantity, and key system operating indicators in real time. In the task generation and allocation step, the task allocation strategy is dynamically adjusted based on real-time traffic density information to avoid assigning tasks to congested areas.

Images