Obstacle-avoiding method, device, equipment, medium and program for unmanned forklift

Abstract

Embodiments of the present application provide a barrier-avoiding method, device, equipment, medium and program for unmanned forklift. The present application relates to the technical field of autonomous barrier-avoiding of unmanned forklift. The method comprises: in response to an obstacle-avoiding instruction, obtaining a plurality of sampling points of the unmanned forklift within the path boundary of a target path; determining a plurality of obstacle-avoiding paths based on the plurality of sampling points; performing collision test on each obstacle-avoiding path and calculating the cost function value of each obstacle-avoiding path passing the collision test; determining a target obstacle-avoiding path according to the cost function values of the plurality of obstacle-avoiding paths passing the collision test; and performing barrier-avoiding control on the unmanned forklift based on the target obstacle-avoiding path. The method is used to achieve the technical effect of precise autonomous barrier-avoiding of the unmanned forklift.

Description

Technical Field

[0001] This application relates to the field of autonomous obstacle avoidance by unmanned forklifts, and more particularly to a method, apparatus, equipment, medium, and program for obstacle avoidance by unmanned forklifts. Background Technology

[0002] In complex and dynamic scenarios such as warehousing and logistics, and manufacturing plants, unmanned forklifts need to perform handling tasks according to preset paths in environments without fixed lane markings. In actual operation, the environment contains dynamic obstacles (such as temporarily stacked goods, moving workers, and temporarily set up barriers), which may suddenly appear on the preset path, causing traditional path planning methods to fail.

[0003] In existing technologies, autonomous obstacle avoidance for unmanned forklifts often employs a combination of global path planning and local obstacle avoidance. Global path planning generates the optimal path based on an algorithm, while local obstacle avoidance uses a graph search algorithm for real-time path adjustment.

[0004] However, in existing technologies, the shortest path is often used for real-time path adjustment during obstacle avoidance. But in actual operation, unmanned forklifts are also subject to the kinematic constraints of the forklift itself when avoiding obstacles. At the same time, path planning varies depending on the diverse needs of the task scenario. Therefore, existing technologies have the technical problem of low accuracy in autonomous obstacle avoidance by unmanned forklifts. Summary of the Invention

[0005] This application provides methods, devices, equipment, media, and programs for unmanned forklifts to bypass obstacles, so as to achieve the technical effect of enabling unmanned forklifts to accurately and autonomously bypass obstacles.

[0006] In a first aspect, embodiments of this application provide a method for an unmanned forklift to bypass obstacles, including:

[0007] In response to obstacle avoidance commands, acquire multiple sampling points of the unmanned forklift within the path boundary of the target path;

[0008] Multiple obstacle avoidance paths are determined based on multiple sampling points;

[0009] Perform a collision test on each obstacle avoidance path and calculate the cost function value for each obstacle avoidance path that passes the collision test;

[0010] The target obstacle avoidance path is determined based on the cost function values ​​of multiple obstacle avoidance paths that have passed the collision test.

[0011] Based on the target obstacle avoidance path, the unmanned forklift is controlled to avoid obstacles.

[0012] In one possible implementation, acquiring multiple sampling points of the unmanned forklift within the path boundary of the target path includes:

[0013] Based on the vehicle body angle constraint and speed constraint of the unmanned forklift, multiple sampling points of the unmanned forklift within the path boundary of the target path are obtained; among them, the vehicle body angle constraint is used to limit the range of sampling points obtained, and the speed constraint is used to limit the number of sampling points obtained.

[0014] In one possible implementation, multiple obstacle avoidance paths are determined based on multiple sampling points, including:

[0015] The system acquires the starting position information and the ending position information of the unmanned forklift. The starting position information is used to characterize the position of the unmanned forklift when it generates an obstacle avoidance command after detecting an obstacle. The ending position information is determined based on the position of the obstacle and a preset distance, and is the position reached by the unmanned forklift after obstacle avoidance.

[0016] Based on the fifth-order polynomial, multiple sampling points, starting position information, and ending position information, multiple obstacle avoidance paths are obtained.

[0017] In one possible implementation, a collision test is performed on each obstacle avoidance path, including:

[0018] Obtain the loading information of the unmanned forklift; the loading information includes both empty and loaded states.

[0019] Based on the loading information, determine the collision expansion distance of the unmanned forklift;

[0020] Collision tests are performed on each obstacle avoidance path based on the collision expansion distance to obtain the collision test results for each path; the collision test results include passing and failing the test.

[0021] Calculate the cost function value of the obstacle avoidance path that passes the test for each collision test result.

[0022] In one possible implementation, the cost function value is determined based on the rate of change of acceleration of the unmanned forklift, the time it takes for the unmanned forklift to reach the destination position after bypassing the obstacle, the degree of deviation of the unmanned forklift from the target path when bypassing the obstacle, the curvature of the obstacle avoidance path, and the steering angle of the steering wheel when the unmanned forklift bypasses the obstacle.

[0023] In one possible implementation, before acquiring multiple sampling points of the unmanned forklift within the path boundary of the target path in response to an obstacle avoidance command, the process includes:

[0024] Obtain the on-site layout diagram of the unmanned forklift, and determine the critical path points, path lines, restricted areas, and restricted virtual walls of the unmanned forklift during the working process based on the on-site layout diagram;

[0025] Obtain the target task order for the unmanned forklift and determine the start and end points of the unmanned forklift;

[0026] Based on critical path points, path lines, restricted areas, restricted virtual walls, starting points, and ending points, determine multiple operating trajectories for the unmanned forklift to complete the target task order;

[0027] Based on the unmanned forklift's operating location information and multiple operating trajectories, the target path of the unmanned forklift is determined;

[0028] The target path region is extended outward based on a preset length to obtain the path boundary of the target path.

[0029] In one possible implementation, the path boundary of the target path includes a first path boundary region and a second path boundary region, and the method further includes:

[0030] Obtain multiple sampling points of the unmanned forklift within the first path boundary area of ​​the target path;

[0031] Determine whether multiple sampling points can generate multiple obstacle avoidance paths. If so, perform a collision test on each obstacle avoidance path.

[0032] If not, then obtain multiple sampling points of the unmanned forklift in the first path boundary region and the second path boundary region of the target path, and generate multiple obstacle avoidance paths based on the multiple sampling points in the first path boundary region and the second path boundary region.

[0033] Secondly, embodiments of this application provide an obstacle avoidance device for an unmanned forklift, comprising:

[0034] The first processing module is used to respond to obstacle avoidance commands and acquire multiple sampling points of the unmanned forklift within the path boundary of the target path;

[0035] The determination module is used to determine multiple obstacle avoidance paths based on multiple sampling points;

[0036] The testing module is used to perform collision tests on each obstacle avoidance path and calculate the cost function value of each obstacle avoidance path that passes the collision test;

[0037] The second processing module is used to determine the target obstacle avoidance path based on the cost function values ​​of multiple obstacle avoidance paths that have passed the collision test.

[0038] The second processing module is also used to control the unmanned forklift to avoid obstacles based on the target obstacle avoidance path.

[0039] In one possible implementation, the first processing module is further configured to:

[0040] Based on the vehicle body angle constraint and speed constraint of the unmanned forklift, multiple sampling points of the unmanned forklift within the path boundary of the target path are obtained; among them, the vehicle body angle constraint is used to limit the range of sampling points obtained, and the speed constraint is used to limit the number of sampling points obtained.

[0041] In one possible implementation, the determining module is further configured to:

[0042] The system acquires the starting position information and the ending position information of the unmanned forklift. The starting position information is used to characterize the position of the unmanned forklift when it generates an obstacle avoidance command after detecting an obstacle. The ending position information is determined based on the position of the obstacle and a preset distance, and is the position reached by the unmanned forklift after obstacle avoidance.

[0043] Based on the fifth-order polynomial, multiple sampling points, starting position information, and ending position information, multiple obstacle avoidance paths are obtained.

[0044] In one possible implementation, the test module is also used for:

[0045] Obtain the loading information of the unmanned forklift; the loading information includes both empty and loaded states.

[0046] Based on the loading information, determine the collision expansion distance of the unmanned forklift;

[0047] Collision tests are performed on each obstacle avoidance path based on the collision expansion distance to obtain the collision test results for each path; the collision test results include passing and failing the test.

[0048] Calculate the cost function value of the obstacle avoidance path that passes the test for each collision test result.

[0049] In one possible implementation, the cost function value is determined based on the rate of change of acceleration of the unmanned forklift, the time it takes for the unmanned forklift to reach the destination position after bypassing the obstacle, the degree of deviation of the unmanned forklift from the target path when bypassing the obstacle, the curvature of the obstacle avoidance path, and the steering angle of the steering wheel when the unmanned forklift bypasses the obstacle.

[0050] In one possible implementation, the first processing module is further configured to:

[0051] Obtain the on-site layout diagram of the unmanned forklift, and determine the critical path points, path lines, restricted areas, and restricted virtual walls of the unmanned forklift during the working process based on the on-site layout diagram;

[0052] Obtain the target task order for the unmanned forklift and determine the start and end points of the unmanned forklift;

[0053] Based on critical path points, path lines, restricted areas, restricted virtual walls, starting points, and ending points, determine multiple operating trajectories for the unmanned forklift to complete the target task order;

[0054] Based on the unmanned forklift's operating location information and multiple operating trajectories, the target path of the unmanned forklift is determined;

[0055] The target path region is extended outward based on a preset length to obtain the path boundary of the target path.

[0056] In one possible implementation, the first processing module is further configured to:

[0057] Obtain multiple sampling points of the unmanned forklift within the first path boundary area of ​​the target path;

[0058] Determine whether multiple sampling points can generate multiple obstacle avoidance paths. If so, perform a collision test on each obstacle avoidance path.

[0059] If not, then obtain multiple sampling points of the unmanned forklift in the first path boundary region and the second path boundary region of the target path, and generate multiple obstacle avoidance paths based on the multiple sampling points in the first path boundary region and the second path boundary region.

[0060] Thirdly, embodiments of this application provide an electronic device, including: a memory and a processor;

[0061] The memory stores computer-executed instructions;

[0062] The processor executes computer execution instructions stored in the memory, causing the processor to perform the first aspect and / or various possible implementations of the first aspect as described above.

[0063] Fourthly, embodiments of this application provide a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, are used to implement the first aspect and / or various possible implementations of the first aspect.

[0064] Fifthly, embodiments of this application provide a computer program product, including a computer program that, when executed by a processor, implements the first aspect and / or various possible implementations of the first aspect.

[0065] The obstacle avoidance method, apparatus, device, medium, and program for unmanned forklifts provided in this application, upon receiving an obstacle avoidance command, first acquires a series of planned and selected sampling points within the path boundary of the target path that the unmanned forklift needs to follow; based on these sampling points, multiple selectable obstacle avoidance paths are generated. Subsequently, a rigorous collision test is performed on each obstacle avoidance path, selecting only those paths that can safely pass through the obstacle area, and calculating the corresponding cost function value for each obstacle avoidance path that passes the test. This value comprehensively reflects the path's performance in terms of smoothness, efficiency, and deviation. Finally, by comparing the cost function values ​​of all obstacle avoidance paths that pass the collision test, an optimal target obstacle avoidance path is determined, and precise obstacle avoidance control of the unmanned forklift is implemented based on this path. This application achieves safe, efficient, and smooth obstacle avoidance trajectory planning in complex dynamic environments by systematically sampling and generating paths within preset path boundaries, and combining collision testing and multi-objective cost function optimization. This not only significantly improves the reliability and operational efficiency of unmanned forklifts in dense warehousing environments, but also ensures the predictability of obstacle avoidance behavior and the overall economic efficiency of operation by introducing a quantitative cost evaluation and optimal selection mechanism, providing core algorithmic support for the intelligent navigation and autonomous decision-making of unmanned forklifts. Attached Figure Description

[0066] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0067] Figure 1 Flowchart of the obstacle avoidance method for the unmanned forklift provided in this application Figure 1 ;

[0068] Figure 2 Flowchart of the obstacle avoidance method for the unmanned forklift provided in this application Figure 2 ;

[0069] Figure 3 A schematic diagram of the first path boundary region and the second path boundary region for the obstacle avoidance method of the unmanned forklift provided in this application;

[0070] Figure 4 A sampling diagram illustrating the obstacle avoidance method for the unmanned forklift provided in this application;

[0071] Figure 5 A schematic diagram of the unmanned forklift with no load and with load, illustrating the obstacle avoidance method of the unmanned forklift provided in this application;

[0072] Figure 6 A schematic diagram of the obstacle avoidance device for the unmanned forklift provided in this application;

[0073] Figure 7A hardware schematic diagram of the obstacle avoidance device for the unmanned forklift provided in this application.

[0074] The accompanying drawings illustrate specific embodiments of this application, which will be described in more detail below. These drawings and descriptions are not intended to limit the scope of the concept in any way, but rather to illustrate the concept of this application to those skilled in the art through reference to particular embodiments. Detailed Implementation

[0075] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of methods and approaches consistent with some aspects of this application as detailed in the appended claims.

[0076] In complex real-world operations such as warehousing and logistics, and manufacturing plants, the working environment is highly dynamic and uncertain. Dynamic obstacles such as temporarily stacked goods, moving workers, temporary barriers, and even other mobile devices can intrude into pre-defined paths at any time, causing traditional planning methods relying on fixed paths to fail. Existing technologies typically employ a hierarchical path planning architecture. This involves a global planning module generating an optimal or suboptimal global reference path based on static environmental map information (such as shelf layout and aisle topology), and then a local obstacle avoidance module adjusting the global path in real time to cope with dynamic obstacles. Local obstacle avoidance often relies on real-time search algorithms such as Dynamic Window (DWA), artificial potential field, or sampled fast random tree (RRT) and its variants. Upon detecting an obstacle, a collision-free local path is replanned online.

[0077] However, existing technologies have several inherent limitations in practical implementation, resulting in insufficient accuracy and practicality of autonomous obstacle avoidance by unmanned forklifts. Firstly, the objectives for path optimization are too simplistic, with most methods using the shortest path or shortest time as the core or even sole optimization goal. While this simplification improves computational efficiency, it ignores the diverse constraints and requirements of actual operations. For example, when transporting precision instruments or fragile items, the smoothness and stability of the path are extremely important; sharp turns or frequent acceleration and deceleration can lead to tipping over or damage. Furthermore, in high-paced production lines, the predictability of the path and minimizing interference with other equipment may be more important than the absolute path length. Existing methods lack the ability to flexibly adapt to these diverse and sometimes conflicting task requirements.

[0078] Secondly, existing technologies often fail to fully and explicitly consider the strict satisfaction of the kinematic and dynamic constraints of the unmanned forklift itself throughout the obstacle avoidance process. Although some local planners (such as the dynamic window method) consider instantaneous motion constraints when sampling the velocity space, they often fail to adequately handle or verify complex constraints such as the minimum turning radius of Ackerman steering vehicles, the impact of forklift center of mass changes on stability under load, and the limits of angular velocity and angular acceleration of the steering wheel system during the path geometry generation stage. This may result in a planned "obstacle avoidance path" that is geometrically collision-free but infeasible at the control level, requiring the controller to make significant adjustments that could lead to instability or overshoot, thereby reducing the accuracy and smoothness of obstacle avoidance execution.

[0079] Furthermore, existing local obstacle avoidance algorithms often exhibit stiffness in their coordination with the global path. During obstacle avoidance replanning, the algorithm may generate temporary paths that deviate too far from the global path or even run in the opposite direction in order to avoid obstacles immediately. This causes the unmanned forklift to exhibit unnatural "circling" or "backward" behavior, which not only increases energy consumption and time but may also trigger new conflicts in confined spaces. At the same time, existing methods lack a systematic strategy for how to smoothly and safely rejoin the original global path after obstacle avoidance, often resulting in abrupt rejoining or the generation of new, unnecessary deviations.

[0080] In summary, while pursuing local real-time performance, existing obstacle avoidance technologies for unmanned forklifts often sacrifice comprehensive consideration of multi-dimensional task requirements, complex vehicle constraints, global path coordination, and detailed environmental modeling. This results in deficiencies in the planned obstacle avoidance paths regarding "executability," "task adaptability," "overall optimality," and "safety," leading to technical problems such as low accuracy, low efficiency, and sometimes even unreliability in obstacle avoidance behavior. This limits the large-scale and efficient application of unmanned forklifts in more complex and demanding scenarios.

[0081] This invention aims to address the low accuracy issues in existing obstacle avoidance technologies for unmanned forklifts, which suffer from neglecting multi-dimensional task requirements, complex vehicle constraints, and detailed environmental modeling. To solve these problems, the invention first addresses the issue at the global task planning level, pre-constructing a detailed environmental model and feasible path space before responding to obstacle avoidance commands. By integrating the on-site layout diagram, not only are critical path points and paths extracted, but also physical and logical constraints such as restricted areas and virtual walls are clearly marked. Combining the start and end points of the task order, multiple candidate trajectories conforming to the global operational logic are planned, and a target path is ultimately determined. To ensure the feasibility of obstacle avoidance operations, this method innovatively extends the target path region outwards by a preset length, forming a safe path boundary region. In complex situations, this region can be further divided into first and second path boundary regions, providing a structured search space for local obstacle avoidance.

[0082] Upon receiving real-time obstacle avoidance commands, the system enters a local fine-planning phase. Its core innovation lies in the deep integration of vehicle kinematic constraints into the sampling point generation process. Unlike existing technologies that simply sample in free space, this application dynamically determines sampling points within a preset path boundary based on the real-time vehicle angle and speed constraints of the unmanned forklift. The angle constraints effectively limit the directional distribution range of the sampling points, ensuring they are located within a kinematically reachable sector area of ​​the vehicle in its current pose. The speed constraints further restrict the longitudinal density and distance of the sampling points, ensuring that the generated candidate path point set covers both the necessary obstacle avoidance possibilities and strictly satisfies the vehicle's transient motion capabilities, fundamentally avoiding the generation of unexecutable paths. Based on these kinematically consistent sampling points, combined with the vehicle's current starting position information and the ending position information determined according to the obstacle position and preset safety distance, multiple smooth candidate obstacle avoidance paths are generated using a fifth-order polynomial curve. The application of the fifth-order polynomial ensures the continuity of the path at the position, speed, and even acceleration levels, laying the foundation for vehicle stability.

[0083] For each generated candidate obstacle avoidance path, this application implements a rigorous and refined collision test and multi-objective optimal selection mechanism. The collision test does not use a fixed vehicle body expansion model; instead, it first acquires the real-time loading information of the unmanned forklift (empty or loaded), and dynamically determines the corresponding collision expansion distance based on different loading states. When loaded, the expansion distance must consider the cargo size, thereby achieving adaptive precision in collision detection and avoiding overly conservative planning while ensuring safety. Only paths that pass the collision test can proceed to the cost function evaluation stage. This cost function is a comprehensive optimization index that not only considers traditional factors such as the degree of deviation from the target path (encouraging a rapid return to the main path) and the time to reach the destination (pursuing efficiency), but also innovatively introduces a penalty for the rate of change of acceleration to suppress rapid acceleration and deceleration, as well as considerations of path curvature and steering wheel angle, thus directly integrating practical engineering constraints such as ride comfort, mechanical wear, and control stability into the optimization objectives.

[0084] Finally, by comparing the cost function values ​​of all feasible paths, the optimal obstacle avoidance path is determined, and the unmanned forklift is precisely controlled to avoid obstacles accordingly. This complete technology chain, from global environment modeling, kinematic constraint embedding sampling, adaptive collision detection to multi-objective optimization decision-making, systematically improves the accuracy, safety, and task adaptability of unmanned forklifts in obstacle avoidance behavior in complex dynamic scenarios.

[0085] The technical solution of this application and how the technical solution of this application solves the above-mentioned technical problems are described in detail below with specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments. The embodiments of this application will now be described with reference to the accompanying drawings.

[0086] Figure 1 Flowchart of the obstacle avoidance method for the unmanned forklift provided in this application Figure 1 ,like Figure 1 As shown, the method includes:

[0087] S101. In response to the obstacle avoidance command, acquire multiple sampling points of the unmanned forklift within the path boundary of the target path.

[0088] In this embodiment, when the unmanned forklift detects an obstacle intruding into its preset driving path using onboard sensors (such as LiDAR and vision systems), it automatically generates an obstacle avoidance command. At this point, based on the path boundary pre-generated according to global path and environmental constraints, a series of candidate spatial location points are collected as sampling points within this boundary. These sampling points are not randomly distributed but are generated through regularization or optimization methods, comprehensively considering the vehicle's current pose, kinematic constraints, and environmental geometric features. This step compresses the path search space from the global environment into a structured, safe, and feasible area, significantly reducing the complexity and computational burden of subsequent planning, and fundamentally ensuring that all subsequent alternative paths do not violate basic path rules and safety margins, laying a spatial foundation for efficient and safe obstacle avoidance decisions.

[0089] S102. Determine multiple obstacle avoidance paths based on multiple sampling points.

[0090] In this embodiment, the core task is to transform discrete sampling points into continuous, executable motion trajectories. Based on the vehicle's current starting state and the preset obstacle avoidance stage endpoint, a path generation algorithm connects the relevant sampling points to form multiple smooth candidate obstacle avoidance paths. Each path describes a possible driving route from the starting point to the endpoint. This step realizes the transformation from a "set of possible points in the space" to a "concrete feasible trajectory scheme." By generating multiple paths, a rich set of choices is provided for subsequent evaluation and optimization, ensuring that feasible detour schemes can be found under different obstacle shapes and spatial constraints, avoiding the predicament that single path planning may fall into local unsolvable or ineffective situations.

[0091] S103. Perform a collision test on each obstacle avoidance path and calculate the cost function value of each obstacle avoidance path that passes the collision test.

[0092] In this embodiment, each candidate obstacle avoidance path generated in step S102 undergoes rigorous collision testing. The testing typically involves spatiotemporally extrapolating the vehicle's profile (including possible loading profiles) along the path and performing geometric intersection detection with static and dynamic obstacles in a high-precision environmental map, thereby filtering out unsafe paths that could lead to collisions. Subsequently, for all paths that pass the safety test, their corresponding cost function values ​​are calculated. This cost function is a multi-objective comprehensive evaluation index, typically integrating multiple dimensions such as path length, smoothness (e.g., curvature change), deviation from the original path, estimated execution time, energy consumption, and friendliness to vehicle actuators (e.g., steering angle change rate). This step, based on collision testing, ensures that any considered solution meets the most basic collision-free safety requirements; while the cost function calculation provides quantifiable performance indicators for these safe solutions, offering an objective and comprehensive basis for the next step of optimal decision-making.

[0093] S104. Determine the target obstacle avoidance path based on the cost function values ​​of multiple obstacle avoidance paths that have passed the collision test.

[0094] In this embodiment, all obstacle avoidance paths that have passed the collision test and whose cost function values ​​have been calculated are compared. Typically, they are sorted according to their cost function values, and the path with the lowest cost is selected as the final "target obstacle avoidance path." In more complex implementations, multi-objective optimization or rule-based selection strategies may be introduced. This step automatically selects the one with the best overall performance (or best suited to specific task preferences, such as the smoothest, fastest, or most energy-efficient) from multiple feasible solutions. This automatic decision-making process based on optimization objectives makes the obstacle avoidance behavior of the unmanned forklift no longer a simple "avoidance," but rather a pursuit of high-quality, high-efficiency passage under safe conditions, demonstrating the superiority of intelligent decision-making.

[0095] S105. Based on the target obstacle avoidance path, control the unmanned forklift to bypass obstacles.

[0096] In this embodiment, the determined target obstacle avoidance path is sent to the vehicle's motion controller. Based on the path information and the vehicle's kinematics and dynamics model, the controller calculates and outputs low-level control commands in real time, such as steering wheel angle and drive wheel speed, guiding the unmanned forklift to accurately and smoothly follow the path, thereby safely and effectively bypassing obstacles. This step ensures that the target obstacle avoidance path can be accurately implemented by the vehicle. Simultaneously, during the tracking process, continuous environmental perception and status feedback are typically performed, forming a real-time closed loop of "perception-planning-control" to cope with dynamic changes in the environment and ensure the successful completion of the obstacle avoidance task.

[0097] The obstacle avoidance method for unmanned forklifts provided in this application, upon responding to an obstacle avoidance command, firstly systematically selects multiple representative sampling points within the path boundary defined by the target path to which the unmanned forklift needs to travel; based on the spatial distribution of these sampling points, multiple possible obstacle avoidance paths are constructed. Next, rigorous collision tests are performed on each generated obstacle avoidance path to select feasible paths that can safely avoid obstacles, and a cost function value reflecting the comprehensive performance of each feasible path is calculated. Finally, by comparing the cost function values ​​of all feasible obstacle avoidance paths, the path with the lowest cost is selected as the final target obstacle avoidance path, and obstacle avoidance control of the unmanned forklift's movement is performed based on this path. This method constrains the path search space within the boundary of the target path and combines collision detection with multi-dimensional cost function optimization to achieve efficient generation and safe selection of obstacle avoidance trajectories. It ensures that the obstacle avoidance process conforms to both global path guidance and local obstacle avoidance requirements, providing unmanned forklifts with reliable, smooth, and energy-based dynamic obstacle avoidance capabilities in structured or semi-structured environments. This effectively improves their operational flexibility and efficiency in warehousing, logistics, and other scenarios, and is a key technological guarantee for realizing autonomous intelligent material handling.

[0098] Figure 2 Flowchart of the obstacle avoidance method for the unmanned forklift provided in this application Figure 2 ,like Figure 2 As shown, in this embodiment... Figure 1 Based on the embodiments, the obstacle avoidance method of the unmanned forklift is described in detail. Specifically, steps S201 to S202 acquire multiple sampling points of the unmanned forklift within the path boundary of the target path, and steps S203 to S206 control the unmanned forklift to avoid obstacles based on the target obstacle avoidance path. The method includes:

[0099] S201. Obtain the on-site layout diagram of the unmanned forklift. Based on the on-site layout diagram, determine the critical path points, path lines, restricted areas, and restricted virtual walls of the unmanned forklift during its operation. Obtain the target task order of the unmanned forklift and determine its starting point and ending point. Based on the critical path points, path lines, restricted areas, restricted virtual walls, starting point, and ending point, determine multiple running trajectories for the unmanned forklift to complete the target task order. Based on the unmanned forklift's running position information and multiple running trajectories, determine the target path of the unmanned forklift. Extend the area of ​​the target path outward based on a preset length to obtain the path boundary of the target path.

[0100] In this embodiment, this step is the infrastructure construction and environmental modeling stage of the unmanned forklift intelligent obstacle avoidance method. Its design principle is to provide a structured and semantic global action framework and safety boundary for subsequent local real-time obstacle avoidance decisions. In actual warehousing or production scenarios, the operation of unmanned forklifts does not take place in a disordered space, but is constrained by fixed shelf layouts, aisle structures, equipment areas, and management rules (such as certain areas where passage is prohibited). Step S201 first transforms the physical space into a machine-understandable environmental model by importing or constructing a work site layout diagram. In this model, critical path points (such as aisle intersections and shelf access points) and path lines (such as the center line of the main aisle) define the network skeleton through which forklifts can pass; while restricted areas (such as fixed equipment fences and charging areas) and restricted virtual walls (such as logical boundaries temporarily defined for safety management) clarify the inviolable spatial constraints. This step essentially abstracts the complex site environment into a constrained topological network, providing a basic routing map for path planning.

[0101] After obtaining the environmental model, and combining it with specific material handling orders (such as picking up goods from point A and delivering them to point B), the starting and ending points of the tasks are determined. Based on this, path search algorithms (such as A*, Dijkstra's algorithm, etc.) are used to plan one or more feasible trajectories from the starting point to the ending point within the constructed environmental topology network. These trajectories must strictly avoid all restricted areas and virtual walls, and proceed along critical waypoints or path lines.

[0102] Subsequently, based on the current actual operating position information of the unmanned forklift and the fixed distance in the direction of movement, a portion of the path in one of the multiple pre-planned trajectories is dynamically selected as the target path. Here, the target path refers to a small portion of the selected trajectories. The fixed distance in the direction of movement is determined by a preset minimum distance and the real-time speed of the unmanned forklift.

[0103] To provide sufficient operational space and ensure safety margins for subsequent local obstacle avoidance, this application extends the area where the target path is located (usually on both sides of the path centerline) outward by a preset length (e.g., 0.5 meters) to form a path boundary. This boundary area defines the spatial range that the unmanned forklift should not exceed when tracking the path and performing obstacle avoidance operations, and is the fundamental constraint for subsequent sampling point generation and obstacle avoidance path search.

[0104] Figure 3 A schematic diagram of the first path boundary region and the second path boundary region of the obstacle avoidance method for the unmanned forklift provided in this application is shown below. Figure 3 As shown in this embodiment, the path boundary of the target path includes a first path boundary region and a second path boundary region.

[0105] This step transforms the continuous, infinite free-space search problem into an optimization problem within a finite, structured strip region by introducing path boundaries. This significantly reduces the search space, lowers the computational complexity of subsequent real-time planning, and improves the obstacle avoidance response speed. Simultaneously, the inclusion of restricted areas and virtual walls allows the obstacle avoidance process to flexibly integrate physical obstacles with logical rules, enhancing environmental adaptability under different operating modes and safety requirements.

[0106] S202. In response to the obstacle avoidance command, acquire multiple sampling points of the unmanned forklift within the first path boundary area of ​​the target path; determine whether the multiple sampling points can generate multiple obstacle avoidance paths; if not, acquire multiple sampling points of the unmanned forklift within the first path boundary area and the second path boundary area of ​​the target path.

[0107] In this embodiment, the aim is to implement a layered sampling strategy to ensure that obstacle avoidance planning is efficient and conservative in most cases, while being flexible and adaptable in extreme cases. When responding to an obstacle avoidance command, the system first attempts to find a solution within the most conservative and safest range—the "first path boundary region." This first region is typically a narrow, safe corridor closely following the original target path; it represents the safest obstacle avoidance space with minimal deviation from the original plan.

[0108] Within the first path boundary region, the first batch of sampling points are generated based on the vehicle's real-time body angle and speed constraints. Multiple sampling points within the target path boundary of the unmanned forklift are acquired based on these constraints. The body angle constraint limits the range of sampling points acquired, while the speed constraint limits the number of sampling points. The angle constraint ensures that the sampling points are distributed within a fan-shaped area that the vehicle can actually reach in its current pose, based on its kinematic characteristics such as its minimum turning radius, thus fundamentally avoiding the generation of geometrically inaccessible path points. The speed constraint dynamically adjusts the longitudinal distribution density and distance of the sampling points according to the current vehicle speed. The faster the speed, the further the sampling points extend forward, but the density may decrease to accommodate vehicle inertia and reserve a longer decision distance. At slower speeds, a more refined local search is performed. This design ensures that each sampling point in the initial search is kinematically reachable and conforms to the current dynamic state, laying a solid foundation for generating feasible trajectories.

[0109] However, obstacles in complex environments may completely or extensively block the first path boundary region, preventing the sampling points collected within that region from constructing an effective obstacle avoidance path (e.g., all sampling points still conflict with obstacles or cannot be connected into a continuous path). In this case, a backup mechanism is activated to extend the search space to the "second path boundary region." The second region is typically a larger feasible area defined based on preset rules (such as a wider global channel boundary, an emergency buffer zone, etc.). Resampling is then performed within the union of the first and second regions.

[0110] The ingenuity of this hierarchical sampling mechanism lies in its adaptability: in most cases of conventional obstacles, a safe and efficient detour path can be found within the first area, maintaining a high degree of predictability and path optimization; when encountering large or tricky obstacles, it can automatically and seamlessly expand the search range, exploring more possibilities without sacrificing safety, greatly improving the success rate of planning and the robustness of the system.

[0111] Figure 4 A sampling diagram of the obstacle avoidance method for the unmanned forklift provided in this application is shown below. Figure 4 As shown, multiple sampling points can be obtained within the path boundary of the target path. From a technical perspective, this step achieves a delicate balance between computational resources and planning effectiveness. First, by deeply embedding the vehicle's kinematic constraints (angle, speed) into the sampling process, it fundamentally improves the "executability" of the generated path, avoiding the pressure on the subsequent trajectory tracking controller and ensuring the smoothness and stability of obstacle avoidance actions. Second, the search strategy of "first region priority, second region backup" enables satisfactory local planning results to be obtained with minimal computational overhead most of the time, while possessing the flexibility to break through conventional constraints in complex situations. Finally, this step provides crucial adaptability to the entire obstacle avoidance method, enabling the unmanned forklift to easily handle obstacle scenarios ranging from simple temporary stacking of objects to complex passageway blockages, significantly improving its autonomous survival and task completion capabilities in dynamic and uncertain environments.

[0112] S203. Obtain the starting position information and ending position information of the unmanned forklift; based on the fifth-order polynomial, multiple sampling points, starting position information and ending position information, obtain multiple obstacle avoidance paths.

[0113] In this embodiment, the starting position information is used to characterize the position of the unmanned forklift when it generates an obstacle avoidance command after detecting an obstacle; the ending position information is determined based on the position of the obstacle and a preset distance, and is the position information reached by the unmanned forklift after obstacle avoidance.

[0114] After acquiring the initial position information (including coordinates, heading angle, and velocity) of the unmanned forklift at the moment it detects an obstacle, as well as the endpoint position information calculated based on the obstacle's position and a preset safety distance, it does not simply connect the sampling points obtained in step S202 with straight lines or arcs. Instead, it uses a fifth-order polynomial curve as the basic mathematical tool to connect the starting point, intermediate sampling points (if used), and the endpoint. Fifth-order polynomials are widely used to generate trajectories where position, velocity, and acceleration are all continuous and smooth because they can independently constrain the boundary conditions of position, velocity, and acceleration.

[0115] In practice, for the overall path from the starting point to the ending point, or for a segmented path (such as from the starting point to a sampling point, and then to the ending point), a fifth-degree polynomial is solved to satisfy the preset position, velocity, and even acceleration constraints at the starting point and the target point (sampling point or ending point). These constraints are directly derived from the vehicle's current motion state (starting point velocity, acceleration) and the requirements for the smoothness of obstacle avoidance (the ending point velocity usually needs to match the requirement of re-merging with the target path, and the acceleration needs to be gentle).

[0116] For example, suppose an unmanned forklift detects an obstacle at position A (coordinates xA, yA) with velocity vA and heading angle θA. The planned obstacle avoidance endpoint is B (coordinates xB, yB), and the expected velocity and heading angle at point B are vB and θB, respectively. One or more intermediate sampling points C, generated in step S202 and satisfying kinematic constraints, are selected between A and B. For path segments A->C and C->B, two fifth-order polynomial curves are generated respectively. The first curve must satisfy the position (xA, yA) and velocity (vAcosθA, vAsinθA) constraints at point A, and the position constraint at point C; the second curve must satisfy the position constraint at point C (connecting with the endpoint of the first curve), and the position (xB, yB) and velocity (vBcosθB, vBsinθB) constraints at point B. By adjusting the polynomial coefficients, multiple curves satisfying the same start and end constraints but with different shapes (i.e., passing through different points in space or with different curvature distributions) can be generated, thus forming multiple candidate obstacle avoidance paths. These paths are geometrically smooth and kinematically natural.

[0117] This step, by using a fifth-order polynomial, fundamentally ensures the continuity of position, velocity, and acceleration in the generated obstacle avoidance path. This is crucial for unmanned forklifts operating with heavy loads, effectively preventing the risk of cargo swaying and slipping due to sudden acceleration changes, thus improving handling safety and stability. Secondly, this method directly encodes the vehicle's motion state (starting speed / heading) and task objective (ending speed / heading) as boundary conditions for the trajectory. This ensures that the generated path is not only spatially collision-free but also seamlessly integrated with the vehicle's current dynamic state, making it easy to track. This significantly reduces the tracking difficulty for the underlying controller and improves trajectory execution accuracy. Finally, multiple polynomial trajectories of different shapes are generated based on the same set of sampling points and start / end constraints. This provides a rich and high-quality set of candidate solutions for subsequent cost function optimization. From multiple paths that satisfy basic kinematic and geometric requirements, the optimal solution with the best overall performance can be selected, achieving a fine balance and optimization of obstacle avoidance behavior across multiple dimensions such as safety, smoothness, and efficiency.

[0118] S204. Obtain the loading information of the unmanned forklift; determine the collision expansion distance of the unmanned forklift based on the loading information; perform a collision test on each obstacle avoidance path based on the collision expansion distance, and obtain the collision test results of each obstacle avoidance path.

[0119] In this embodiment, the loading information includes empty and loaded states. Figure 5 This diagram illustrates the unmanned forklift's obstacle avoidance method provided in this application, showing its empty and loaded states. The unmanned forklift operates in two basic states during task execution: "empty" and "loaded." Furthermore, the size, shape, and stacking height of different goods significantly alter the vehicle's effective profile. Traditional methods often employ fixed or maximized expansion distances for collision detection, leading to overly conservative planning (wasting passable space) when empty or transporting small items, or insufficient assessment (posing collision risks) when transporting oversized goods.

[0120] Step S204 dynamically determines a matching collision expansion distance by acquiring loading information in real time (which can be obtained through scheduling system instructions, vehicle-mounted weight sensors, or visual recognition). This distance is not a simple uniform expansion, but rather an addition of a safety margin to the vehicle's basic outline (body, fork carriage) based on key parameters such as fork lifting height and the outward extension dimensions of the goods on the pallet, forming a precise collision envelope for the current work cycle.

[0121] Specifically, when the forklift is unloaded, its collision expansion distance mainly considers the fork thickness, sensor installation position, and control error, and may only add a safety margin of 0.1-0.2 meters outside the basic profile. When the forklift is carrying a standard pallet of goods, the amount of the goods protruding relative to the forks in length, width, and height is calculated based on a preset goods size library or real-time measurement data, and a larger expansion distance is added in the corresponding directions (especially in front and on both sides). For example, if the width of the goods exceeds the forks by 0.3 meters, the collision detection boundary of the vehicle profile in the width direction needs to be increased by an additional 0.3 meters, and may also be expanded accordingly in the vertical direction based on the height of the goods to consider the risk of top collision. Subsequently, in the collision test phase, this dynamically generated collision envelope is "scanned" along each candidate obstacle avoidance path generated in step S203. Through geometric calculation or discretization detection, it is determined whether the envelope interferes with any obstacles in the environment (from the local map perceived in real time) along the entire path. Only paths with no collisions throughout the entire envelope are marked as "passed the test".

[0122] This step, through adaptive expansion based on loading status, significantly improves collision detection accuracy, effectively avoiding potential scratches or collisions caused by inaccurate models, especially ensuring operational safety when handling non-standard or oversized goods. Secondly, the dynamic model allows the system to plan paths closer to obstacles or more directly when space is ample (e.g., when empty), improving space utilization and operational efficiency; while in confined spaces (e.g., when carrying a load through narrow passages), it automatically adopts more conservative detection boundaries to ensure absolute safety. Finally, this adaptive mechanism enhances the unmanned forklift's environmental adaptability when switching between different tasks. Without manual reconfiguration of safety parameters, it can autonomously adjust safety strategies based on the actual handling situation each time, demonstrating highly intelligent operational characteristics and providing key technical support for achieving robust autonomous handling in complex dynamic environments.

[0123] S205. Based on the rate of change of acceleration of the unmanned forklift, the time it takes for the unmanned forklift to reach the endpoint after bypassing the obstacle, the degree of deviation of the unmanned forklift from the target path when bypassing the obstacle, the curvature of the obstacle avoidance path, and the steering wheel angle of the unmanned forklift when bypassing the obstacle, the cost function value of the obstacle avoidance path that passes the test is obtained for each collision test result.

[0124] In this embodiment, after selecting all obstacle avoidance paths that pass each collision test through collision detection, these paths still exhibit significant differences in terms of smoothness, efficiency, stability, and adherence to the original plan. If only a single indicator (such as the shortest path) is used for selection, it may lead to suboptimal behaviors that do not meet the actual operational requirements, such as sharp turns, violent acceleration and deceleration, or unnecessary long-distance detours.

[0125] Therefore, step S205 defines a comprehensive cost function that weights and fuses five key performance indicators: acceleration change rate (Jerk), obstacle avoidance time, path deviation, path curvature, and steering wheel angle. The acceleration change rate directly relates to cargo stability and mechanical shock; a smaller value indicates smoother movement. Obstacle avoidance time represents task execution efficiency. Path deviation measures the deviation from the original target path, affecting overall task continuity and predictability. Path curvature and steering wheel angle together reflect the curvature characteristics of the path; excessive curvature or steering angle may exceed the vehicle's kinematic limits or lead to control difficulties. By calculating the quantitative performance of each safe path across these dimensions and weighting the sum, a comparable comprehensive cost value is ultimately assigned to each path.

[0126] For example, suppose there are two candidate obstacle avoidance paths, A and B, both of which have passed collision detection. Path A is a relatively direct detour, but its curvature at the turns is large, resulting in a high calculated Jerk value, and the required steering wheel angle is close to the limit. Path B is a gentler arc, with a low Jerk value and a moderate steering angle, but the path is slightly longer, resulting in a slightly longer expected obstacle avoidance time and a slightly larger lateral offset from the original path. A weight is assigned to each metric for each path (e.g., Jerk has a higher weight when handling fragile items; time has a higher weight when pursuing efficiency), and then the specific cost of each metric is calculated according to a pre-set mathematical model (such as the integral square of Jerk, a linear function of time, the square of the offset distance, etc.). Finally, a weighted sum is obtained to obtain the total cost function value. Path A may receive a higher penalty value due to its high Jerk and large steering angle, while path B may receive a lower total cost due to its good smoothness, although its time and offset costs are slightly higher.

[0127] This step transforms complex engineering experience and multi-objective decision-making into an automatically executable optimization algorithm. First, it avoids the one-sidedness of single-objective optimization, ensuring that the final selected path achieves an optimal balance in smoothness, efficiency, and controllability while maintaining safety, thereby improving the overall quality of operations and equipment lifespan. Second, by adjusting the weights of various indicators, it can flexibly adapt to different operational scenarios and task requirements. For example, increasing the weight of smoothness in precision instrument handling scenarios and increasing the weight of efficiency in high-efficiency logistics scenarios achieves scenario-adaptive optimization of obstacle avoidance behavior. Finally, this quantitative evaluation mechanism provides an objective and consistent decision-making basis for path selection, avoiding suboptimal decisions that may result from randomness or heuristic rules. This ensures that the obstacle avoidance behavior of the unmanned forklift is not only safe and reliable but also intelligent, efficient, and consistent with higher-order operational intentions, representing a key step in improving its autonomous operation capabilities and intelligence level.

[0128] S206. Determine the target obstacle avoidance path based on the cost function values ​​of multiple obstacle avoidance paths that have passed the collision test; and perform obstacle avoidance control on the unmanned forklift based on the target obstacle avoidance path.

[0129] In this embodiment, by quantitatively comparing the results of multi-objective optimization, the path with the best overall performance is automatically selected from multiple feasible solutions, completing the closed loop from planning to control. After the preceding steps have completed the comprehensive cost calculation for all safe paths, the cost value of all candidate paths is sorted, and the path with the lowest cost value is usually selected as the "target obstacle avoidance path". This selection mechanism is not a simple comparison of a single index, but a global trade-off based on the multi-dimensional cost function constructed in step S205. This function integrates multiple engineering optimization objectives such as smoothness (rate of change of acceleration), efficiency (obstacle avoidance time), path tracking accuracy (degree of deviation from the target path), geometric feasibility (path curvature), and control friendliness (steering wheel angle).

[0130] Therefore, the final selected obstacle avoidance path is a Pareto optimal solution that achieves the best balance across all performance indicators while ensuring absolute safety (passing collision detection). Once the target path is selected, it is converted into a series of executable control commands and sent to the forklift's underlying motion controller. The controller, based on the path's geometric parameters (such as curvature and heading changes) and motion constraints (such as velocity profiles), calculates and outputs the drive motor's speed and steering angle in real time, guiding the vehicle to accurately and smoothly follow the obstacle avoidance trajectory.

[0131] The obstacle avoidance process is not the end point. After the unmanned forklift successfully bypasses the obstacle based on the target obstacle avoidance path, its endpoint is usually set at a "merging point" ahead of the original target path. This point is located at a safe distance behind the obstacle, and its speed and heading angle are pre-constrained to match the expected state of the original target path at that point. Therefore, as the obstacle avoidance maneuver nears its end, the forklift's state (position, speed, heading) is ready to rejoin the original path. At this time, the trajectory switching logic is automatically triggered: the obstacle avoidance local controller returns control to the global path tracking controller. The global controller uses the current vehicle state as the new starting point and a point or continuous path segment further ahead on the original target path as the tracking target, generating a "regression trajectory" that smoothly transitions from the current state to the original path. The generation of this regression trajectory also follows similar optimization principles (such as using polynomial curves to ensure smoothness) to ensure that the switching process is seamless and natural, without causing sudden shaking or deviation of the vehicle. Subsequently, the vehicle will gradually adjust its lateral position and heading along this return trajectory, eventually stably and accurately reattaching itself to the original target path to continue performing the unfinished transport task.

[0132] This step, based on optimal path selection using a multi-dimensional cost function, ensures that obstacle avoidance is not only safe but also of high quality, balancing operational efficiency, equipment protection, and operational stability, thus improving overall system performance. Secondly, the smooth regression mechanism design avoids problems such as vehicle hesitation, stagnation, or abrupt entry after obstacle avoidance that may occur in traditional methods, ensuring the continuity and efficiency of task execution and reducing global task delays caused by local interference. Finally, this fully automated process from decision-making to execution to regression significantly reduces reliance on external intervention, demonstrating a high level of autonomous intelligence. This allows unmanned forklifts to calmly handle sudden obstacles in dynamic and complex industrial environments, much like experienced drivers, and quickly and gracefully return to normal operation, truly realizing the engineering implementation of unmanned and intelligent material handling.

[0133] Figure 6 This is a structural schematic diagram of the obstacle avoidance device for the unmanned forklift provided in this application, as shown below. Figure 6 As shown, the obstacle avoidance device 60 for the unmanned forklift provided in this embodiment includes:

[0134] The first processing module 601 is used to respond to the obstacle avoidance command and acquire multiple sampling points of the unmanned forklift within the path boundary of the target path;

[0135] The determination module 602 is used to determine multiple obstacle avoidance paths based on multiple sampling points;

[0136] The test module 603 is used to perform collision tests on each obstacle avoidance path and calculate the cost function value of each obstacle avoidance path that passes the collision test;

[0137] The second processing module 604 is used to determine the target obstacle avoidance path based on the cost function values ​​of multiple obstacle avoidance paths that have passed the collision test.

[0138] The second processing module 604 is also used to control the unmanned forklift to bypass obstacles based on the target obstacle avoidance path.

[0139] In one possible implementation, the first processing module 601 is further configured to:

[0140] Based on the vehicle body angle constraint and speed constraint of the unmanned forklift, multiple sampling points of the unmanned forklift within the path boundary of the target path are obtained; among them, the vehicle body angle constraint is used to limit the range of sampling points obtained, and the speed constraint is used to limit the number of sampling points obtained.

[0141] In one possible implementation, the determining module 602 is further configured to:

[0142] The system acquires the starting position information and the ending position information of the unmanned forklift. The starting position information is used to characterize the position of the unmanned forklift when it generates an obstacle avoidance command after detecting an obstacle. The ending position information is determined based on the position of the obstacle and a preset distance, and is the position reached by the unmanned forklift after obstacle avoidance.

[0143] Based on the fifth-order polynomial, multiple sampling points, starting position information, and ending position information, multiple obstacle avoidance paths are obtained.

[0144] In one possible implementation, the test module 603 is further configured to:

[0145] Obtain the loading information of the unmanned forklift; the loading information includes both empty and loaded states.

[0146] Based on the loading information, determine the collision expansion distance of the unmanned forklift;

[0147] Collision tests are performed on each obstacle avoidance path based on the collision expansion distance to obtain the collision test results for each path; the collision test results include passing and failing the test.

[0148] Calculate the cost function value of the obstacle avoidance path that passes the test for each collision test result.

[0149] In one possible implementation, the cost function value is determined based on the rate of change of acceleration of the unmanned forklift, the time it takes for the unmanned forklift to reach the destination position after bypassing the obstacle, the degree of deviation of the unmanned forklift from the target path when bypassing the obstacle, the curvature of the obstacle avoidance path, and the steering angle of the steering wheel when the unmanned forklift bypasses the obstacle.

[0150] In one possible implementation, the first processing module 601 is further configured to:

[0151] Obtain the on-site layout diagram of the unmanned forklift, and determine the critical path points, path lines, restricted areas, and restricted virtual walls of the unmanned forklift during the working process based on the on-site layout diagram;

[0152] Obtain the target task order for the unmanned forklift and determine the start and end points of the unmanned forklift;

[0153] Based on critical path points, path lines, restricted areas, restricted virtual walls, starting points, and ending points, determine multiple operating trajectories for the unmanned forklift to complete the target task order;

[0154] Based on the unmanned forklift's operating location information and multiple operating trajectories, the target path of the unmanned forklift is determined;

[0155] The target path region is extended outward based on a preset length to obtain the path boundary of the target path.

[0156] In one possible implementation, the first processing module 601 is further configured to:

[0157] Obtain multiple sampling points of the unmanned forklift within the first path boundary area of ​​the target path;

[0158] Determine whether multiple sampling points can generate multiple obstacle avoidance paths. If so, perform a collision test on each obstacle avoidance path.

[0159] If not, then obtain multiple sampling points of the unmanned forklift in the first path boundary region and the second path boundary region of the target path, and generate multiple obstacle avoidance paths based on the multiple sampling points in the first path boundary region and the second path boundary region.

[0160] The obstacle avoidance device for the unmanned forklift provided in this embodiment can execute the method provided in the above method embodiment. Its implementation principle and technical effect are similar, and will not be described in detail here.

[0161] Figure 7 This is a hardware schematic diagram of the obstacle avoidance device for the unmanned forklift provided in this application. Figure 7 As shown, the electronic device 70 provided in this embodiment includes at least one processor 701 and a memory 702. Optionally, the device 70 further includes a communication component 703. The processor 701, memory 702, and communication component 703 are connected via a bus 704.

[0162] In a specific implementation, at least one processor 701 executes computer execution instructions stored in memory 702, causing at least one processor 701 to perform the above-described method.

[0163] The specific implementation process of processor 701 can be found in the above method embodiments, and its implementation principle and technical effect are similar. It will not be repeated here.

[0164] In the above embodiments, it should be understood that the processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), etc. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the method disclosed in this invention can be directly implemented by a hardware processor, or implemented by a combination of hardware and software modules within the processor.

[0165] The memory may include random access memory (RAM) and may also include non-volatile memory (NVM), such as at least one disk storage device.

[0166] The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. Buses can be categorized as address buses, data buses, control buses, etc. For ease of illustration, the buses shown in the accompanying drawings are not limited to a single bus or a single type of bus.

[0167] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the above-described method.

[0168] This application also provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the above-described method.

[0169] The aforementioned readable storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. The readable storage medium can be any available medium accessible to a general-purpose or special-purpose computer.

[0170] An exemplary readable storage medium is coupled to a processor, enabling the processor to read information from and write information to the readable storage medium. Of course, the readable storage medium can also be a component of the processor. The processor and the readable storage medium can reside in an Application Specific Integrated Circuit (ASIC). Alternatively, the processor and the readable storage medium can exist as discrete components in the device.

[0171] The division of units is merely a logical functional division; in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be indirect coupling or communication connection through some interfaces, methods, or units, and may be electrical, mechanical, or other forms.

[0172] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0173] In addition, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0174] If a function is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0175] Those skilled in the art will understand that all or part of the steps of the above-described method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When executed, the program performs the steps of the above-described method embodiments; and the aforementioned storage medium includes various media capable of storing program code, such as ROM, RAM, magnetic disks, or optical disks.

[0176] Finally, it should be noted that other embodiments of the invention will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This invention is intended to cover any variations, uses, or adaptations of the invention that follow the general principles of the invention and include common knowledge or customary techniques in the art not disclosed herein, and is not limited to the precise structures described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of the invention is limited only by the appended claims.

Claims

1. A method for obstacle avoidance by an unmanned forklift, characterized in that, include: In response to obstacle avoidance commands, acquire multiple sampling points of the unmanned forklift within the path boundary of the target path; Multiple obstacle avoidance paths are determined based on multiple sampling points; For each of the obstacle avoidance paths, a collision test is performed, and the cost function value of each obstacle avoidance path that passes the collision test is calculated; The target obstacle avoidance path is determined based on the cost function values ​​of multiple obstacle avoidance paths that have passed the collision test. Based on the target obstacle avoidance path, the unmanned forklift is controlled to bypass obstacles.

2. The method according to claim 1, characterized in that, The acquisition of multiple sampling points of the unmanned forklift within the path boundary of the target path includes: Based on the vehicle body angle constraint and speed constraint of the unmanned forklift, multiple sampling points of the unmanned forklift within the path boundary of the target path are obtained; wherein, the vehicle body angle constraint is used to limit the range of sampling points obtained, and the speed constraint is used to limit the number of sampling points obtained.

3. The method according to claim 1, characterized in that, The determination of multiple obstacle avoidance paths based on multiple sampling points includes: The starting position information and ending position information of the unmanned forklift are obtained; wherein, the starting position information is used to characterize the position of the unmanned forklift when it generates an obstacle avoidance command after detecting an obstacle; the ending position information is determined based on the position of the obstacle and a preset distance, and is the position reached by the unmanned forklift after obstacle avoidance; Multiple obstacle avoidance paths are obtained based on the fifth-order polynomial, multiple sampling points, the starting position information, and the ending position information.

4. The method according to claim 1, characterized in that, The collision test for each of the obstacle avoidance paths includes: Obtain the loading information of the unmanned forklift; wherein, the loading information includes empty and loaded states; Based on the loading information, the collision expansion distance of the unmanned forklift is determined; Collision tests are performed on each obstacle avoidance path based on the collision expansion distance to obtain collision test results for each obstacle avoidance path; wherein, the collision test results include passing the test and failing the test; Calculate the cost function value of the obstacle avoidance path that passes the test for each collision test result.

5. The method according to claim 4, characterized in that, The cost function value is determined based on the rate of change of acceleration of the unmanned forklift, the time it takes for the unmanned forklift to reach the destination position after bypassing the obstacle, the degree of deviation of the unmanned forklift from the target path when bypassing the obstacle, the curvature of the obstacle avoidance path, and the steering angle of the steering wheel when the unmanned forklift bypasses the obstacle.

6. The method according to any one of claims 1-5, characterized in that, Before acquiring multiple sampling points of the unmanned forklift within the path boundary of the target path in response to the obstacle avoidance command, the process includes: Obtain the on-site layout diagram of the unmanned forklift, and determine the critical path points, path lines, restricted areas, and restricted virtual walls of the unmanned forklift during its operation based on the on-site layout diagram; Obtain the target task order of the unmanned forklift and determine the start and end points of the unmanned forklift; Based on the critical path points, the path lines, the restricted areas, the restricted virtual walls, the starting point, and the ending point, multiple operating trajectories of the unmanned forklift are determined to complete the target task order; The target path of the unmanned forklift is determined based on its operating location information and multiple operating trajectories. The target path region is extended outward based on a preset length to obtain the path boundary of the target path.

7. The method according to claim 6, characterized in that, The path boundary of the target path includes a first path boundary region and a second path boundary region. The method further includes: Obtain multiple sampling points of the unmanned forklift within the first path boundary area of ​​the target path; Determine whether multiple sampling points can generate multiple obstacle avoidance paths; if so, perform a collision test on each obstacle avoidance path. If not, then obtain multiple sampling points of the unmanned forklift in the first path boundary region and the second path boundary region of the target path, and generate multiple obstacle avoidance paths based on the multiple sampling points in the first path boundary region and the second path boundary region.

8. An obstacle avoidance device for an unmanned forklift, characterized in that, include: The first processing module is used to respond to obstacle avoidance commands and acquire multiple sampling points of the unmanned forklift within the path boundary of the target path; The determination module is used to determine multiple obstacle avoidance paths based on multiple sampling points; The testing module is used to perform collision tests on each of the obstacle avoidance paths and calculate the cost function value of each obstacle avoidance path that passes the collision test; The second processing module is used to determine the target obstacle avoidance path based on the cost function values ​​of multiple obstacle avoidance paths that have passed the collision test; The second processing module is also used to control the unmanned forklift to bypass obstacles based on the target obstacle avoidance path.

9. An electronic device, characterized in that, include: Memory, processor; The memory stores computer-executed instructions; The processor executes computer execution instructions stored in the memory, causing the processor to perform the method as described in any one of claims 1-6.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-executable instructions, which, when executed by a processor, are used to implement the method as described in any one of claims 1-7.

Images