A novel path planning method based on three-level communication and forward and reverse expansion search strategies

Abstract

This invention relates to a novel path planning method based on three-level communication and a forward and reverse expansion search strategy. First, the factory is divided into different areas and sensors are deployed. Then, the sensors collect various parameters of a vehicle and send the collected information to the main control platform. The main control platform analyzes the collected information to determine if there are obstacles threatening the forward path, and decides whether to adjust the vehicle's path. When the vehicle's path needs adjustment, the main control platform uses a forward and reverse expansion search algorithm to plan the vehicle's path, obtaining a planned path. The above steps are repeated to obtain multiple planned paths. Finally, a comprehensive evaluation function is constructed to evaluate the multiple planned paths, and the planned path with the highest evaluation is selected as the optimal path.

Description

Technical Field

[0001] This invention relates to a path planning method, specifically a novel path planning method based on three-level communication and forward and reverse expansion search strategies. Background Technology

[0002] With the advent of Industry 4.0 and artificial intelligence, automated guided vehicles (AGVs) have been widely used in warehousing, logistics, and manufacturing. As crucial equipment in logistics and warehousing, the path planning capability of AGVs determines the operational efficiency of warehouses; therefore, reasonable path planning is essential for AGVs. Currently widely used algorithms, such as A* algorithm, Dijkstra's algorithm, artificial potential field method, genetic algorithm, particle swarm optimization, and ant colony optimization, while possessing good parallelism and robustness, also have some drawbacks, such as low global search efficiency, slow convergence speed, and susceptibility to local optima. Traditional heuristic search algorithms perform a unidirectional search from the starting point to the target point, ultimately obtaining the minimum cost path. However, the planned paths also have some problems, such as the OpenList containing too many nodes unrelated to the generated path, the phenomenon of continuous back-and-forth searching in complex environments, and the existence of redundant points and unnecessary turning points in the planned paths. These factors cause traditional heuristic search algorithms to suffer from high computational cost, severe memory consumption, low efficiency, uneven paths, poor obstacle avoidance, excessive expansion of nodes, and limitations in the direction of path search during large-scale path planning. Summary of the Invention

[0003] The purpose of this invention is to overcome the shortcomings of existing technologies and provide a novel path planning method based on three-level communication and a forward and reverse expansion search strategy. This novel path planning method performs global static path planning based on a forward and reverse expansion search algorithm within a known environment model, reducing the number of nodes searched by the automated guided vehicle (AGV) in each two-dimensional plane, effectively avoiding the selection of a large number of invalid points, and achieving the goal of reducing computational overhead and memory usage. Simultaneously, it dynamically adjusts the weights of the heuristic function, improving search efficiency; furthermore, it improves the search scheme for special nodes and uses B-spline curves to smooth inflection points in path planning, further shortening the path length; it sets constraints on the speed and deflection angle of the AGV, and selects the optimal path and vehicle attributes through an evaluation function, adjusting the shortest obstacle avoidance path in real time.

[0004] The technical solution of the present invention to solve the above-mentioned technical problems is:

[0005] A novel path planning method based on three-level communication and forward / reverse expansion search strategy includes the following steps:

[0006] S1. Divide the factory into different areas and deploy sensors accordingly;

[0007] S2. Collect various parameters of the vehicle through sensors and send the collected information to the main control platform;

[0008] S3. The main control platform analyzes the collected information to determine whether there are obstacles or threats on the forward path, and then decides whether to adjust the vehicle's forward path. When it is necessary to adjust the vehicle's forward path, the main control platform uses a forward and reverse expansion search algorithm to plan the vehicle's forward path to obtain a planned path.

[0009] S4. Repeat steps S2 and S3 to obtain multiple planned paths;

[0010] S5. Evaluate the multiple planning paths obtained by constructing a comprehensive evaluation function, and select the planning path with the highest evaluation as the optimal path.

[0011] Preferably, in step S1, the parameters include, but are not limited to, the position and speed of the vehicle in different areas.

[0012] Preferably, in step S3, when the sensor reports an obstacle ahead, the vehicle's path is adjusted again at the current position to achieve local obstacle avoidance; at the same time, the sensor should transmit the vehicle's position information in real time and display it as an icon on the main control platform to facilitate monitoring of the vehicle's operation and timely adjustments in case of problems.

[0013] Preferably, in step S3, the main control platform uses a forward and reverse expansion search algorithm to plan the forward path of the vehicle, including the following steps:

[0014] S31. Reduce the number of search nodes in the traditional heuristic search algorithm. Reduce the number of nodes in the traditional heuristic search algorithm from eight nodes in the original nine-square grid centered on the node, excluding the current node, to five nodes.

[0015] S32. In the traditional heuristic search algorithm, during the expansion of nodes in both the forward and reverse directions, the minimum cost point in the opposite direction is regarded as the target point in the current search direction for expansion, and an optimal trajectory weight function λ is introduced. j * Establish the cost evaluation function shown below:

[0016]

[0017] In the formula: f1(n) is the cost function for expanding nodes from the starting point to the target point; f2(n) is the cost function for expanding nodes from the target point to the starting point; g1(n) is the actual cost value from the starting point to node n, g2(n) is the actual cost value from the target point to node n; h1(n) is the estimated cost value of the distance from node n to the target point, h2(n) is the estimated cost value of the distance from node n to the starting point; j is a constant, and (j=1,2);

[0018] in,

[0019]

[0020] In the formula, σ1 and σ2 are the angles between the current node and the line connecting the minimum cost of the opposite search in the forward and reverse search processes, respectively.

[0021] Step S33: After obtaining the cost evaluation function, treat redundant points and turning points as special nodes in the search process; connect all nodes sequentially to generate a temporary path. If the current node is collinear with the nodes before and after it, it is a collinear node and needs to be removed; after the collinear nodes are removed, let the nodes in the path be {P}. k Given |k = 1, 2, ..., n}, let P1 be the parent node. Starting from the parent node P1, connect the next node in sequence until... Intersecting with an obstacle; at this point, the special node is P. k-1 The formal path became Then update the parent node P1 to P k-1 Repeat the above steps until all nodes have been traversed.

[0022] Preferably, in step S33, a cubic uniform B-spline curve is used to optimize the path at the turning point. After smoothing the turning point with the cubic uniform B-spline curve, the curve is used instead of the broken line to obtain the shortest path. After the redundant nodes and turning points in the shortest path are removed, the shortest path only contains the starting point, special nodes and the ending point.

[0023] Preferably, in step S5, constructing the comprehensive evaluation function includes the following steps:

[0024] S51. Construct the velocity evaluation function Vel(v, ω);

[0025] S52. Construct the deflection evaluation function Deflect(v, ω) for the car;

[0026] S53. Construct the path change rationality evaluation function Change(v, ω);

[0027] S54. Construct a comprehensive evaluation function G(v, ω), where,

[0028] G(v,ω)=γ*Change(v,ω)+δ*Deflect(v,ω)+β*Vel(v,ω);

[0029] In the formula, γ, δ, and β are the weighting coefficients of each evaluation function.

[0030] Preferably, in step S51, the step of constructing the velocity evaluation function Vel(v, ω) is as follows:

[0031] S511. Apply speed constraints to the trolley:

[0032] V i ={(v, ω)|v∈[v min v max ], ω∈[ω min ω max ]};

[0033] In the formula, v min and v max ω represents the minimum and maximum linear velocities of the trolley, respectively; min and ω max These represent the minimum and maximum angular velocities of the trolley, respectively.

[0034] S512. Apply acceleration constraints to the trolley:

[0035]

[0036] In the formula, v c ω c These are the linear velocity and angular velocity at the current moment; These are the maximum linear and angular accelerations of the vehicle. These are the minimum linear and angular accelerations of the vehicle.

[0037] S513. In order to ensure that the car stops before colliding with the obstacle, the speed must also meet the following requirements:

[0038]

[0039] In the formula, dist(v, ω) is the shortest distance between the end of the car's trajectory and the obstacle on the corresponding (v, ω) path;

[0040] S514, The velocity evaluation function Vel(v, ω) is:

[0041]

[0042] Where V is the constraint range of the car's speed, when the car's speed attribute V is within the range of V... i +V j +V kThe smaller the absolute value of the direct difference, the better the evaluation of the current path planning; otherwise, a new path planning should be implemented.

[0043] Preferably, in step S52, the deflection evaluation function Deflect(v, ω) of the vehicle is constructed as follows:

[0044]

[0045] In the formula, This represents the deflection angle of the next path segment relative to the previous path segment.

[0046] Preferably, in step S53, the step of constructing the path change rationality evaluation function Change(v, ω) is as follows:

[0047] S531, The car is continuously scanned by sensors, at t a The dynamic threat trajectory is obtained at all times as θ a (t), and generate a local obstacle avoidance trajectory; at t b (t b >t a Dynamic threat trajectory θ at time ) b (t) and θ a (t) The reasonableness of path changes must be satisfied:

[0048] max(|θ a (t)-θ b (t)|)<ε;

[0049] In the formula, t∈[t a , t b ε is a constant, used as the criterion for deciding whether to replan the route;

[0050] S532, Path Change Reasonableness Evaluation Function Change(v, ω):

[0051]

[0052] The lower the maximum value of the dynamic threat trajectory change is compared to the set ε value, the higher the evaluation of the path; if the maximum value is equal to or greater than the ε value, the path segment is canceled and replanned.

[0053] Compared with the prior art, the present invention has the following advantages:

[0054] This invention presents a novel path planning method based on three-level communication and a forward and reverse expansion search strategy. In a known environment model, this method performs global static path planning using a forward and reverse expansion search algorithm, reducing the number of nodes searched by the vehicle in each two-dimensional plane, effectively avoiding the selection of a large number of invalid points, and achieving the goal of reducing computational overhead and memory usage. Simultaneously, it dynamically adjusts the weights of the heuristic function, improving search efficiency. Furthermore, it improves the search scheme for special nodes and uses B-spline curves to smooth inflection points in path planning, further shortening the path length. Finally, it sets constraints on the vehicle's speed and deflection angle, and selects the optimal path and vehicle attributes through a comprehensive evaluation function, adjusting the shortest obstacle avoidance path in real time. Attached Figure Description

[0055] Figure 1 For environmental maps.

[0056] Figure 2 This is a flowchart of the dispatching process for the small vehicle.

[0057] Figure 3 This is a diagram illustrating node search.

[0058] Figure 4 This is a schematic diagram of the deflection angle. Detailed Implementation

[0059] The present invention will be further described in detail below with reference to the embodiments and accompanying drawings, but the embodiments of the present invention are not limited thereto.

[0060] See Figures 1-3 The novel path planning method of this invention, based on three-level communication and a forward and reverse expansion search strategy, first divides the factory into different areas and sets up sensors; then, the sensors collect various parameters of the vehicle and send them to the main control platform; the main control platform analyzes the collected information to determine whether there are obstacles threatening the forward path, and thus decides whether to adjust the vehicle's forward path; when the vehicle's forward path needs to be adjusted, the main control platform uses a forward and reverse expansion search algorithm to plan the vehicle's forward path and obtain the optimal path; the above steps can be repeated to obtain multiple paths; finally, the multiple paths are evaluated, and the path with the highest evaluation is selected as the optimal path.

[0061] The following details the novel path planning method based on three-level communication and forward / reverse expansion search strategy of the present invention:

[0062] See Figure 1A solution space is constructed on a two-dimensional plane using a grid method, creating a 20x20 grid map environment defined as: {(x, y)|0≤x≤20, 0≤y≤20}, where (x, y) are the coordinates of nodes in the planning plane, which can also represent the horizontal and vertical coordinates of the car's position in space. The trajectory of the car in the two-dimensional plane consists of a series of node sequences {G, P1, P2, ..., P...}. n-1 The trajectory segment {L} consisting of H, n≤399} i The expression is constructed as follows: |i=1,2,...,i,i≤400}. Here, G represents the starting point, H represents the target point (which is also the starting point for the reverse search), and P1, P2, ..., P... n-1 The intermediate node is defined as an intermediate node; at the same time, obstacles in the scene are defined as black infeasible areas, and white grids are defined as feasible regions that can generate path nodes.

[0063] See Figure 2 When goods arrive at the factory, the main control platform receives the information, assigns tasks to the trolleys (such as automated guided vehicles), and plans their paths. The trolleys then begin to move. To facilitate obtaining the trolleys' positions within the factory, the factory is divided into different areas and sensors are installed. The sensors collect data on the trolleys' positions in different areas, their speed, and other attributes, and feed this information back to the main control platform. The main control platform uses this information to determine whether to adjust the trolleys' paths. If the sensors indicate a threat from an obstacle ahead, the platform adjusts the trolleys' paths again at the current location to achieve local obstacle avoidance. Simultaneously, the sensors can transmit the trolleys' position information in real time, which is displayed as icons on the main control platform for easy monitoring of the trolleys' operation and timely adjustments in case of problems.

[0064] Traditional heuristic search algorithms perform a one-way search from the starting point to the target point to obtain the minimum cost path. The cost evaluation function for its two-dimensional trajectory is:

[0065] f(n) = g(n) + h(n);

[0066] In the formula, f(n) is the global evaluation cost of the current node n, g(n) is the actual cost from the starting point to node n, and h(n) is the estimated cost of the distance from node n to the target point. This reflects the heuristic nature of the algorithm. In two-dimensional space, Euclidean distance is often used as a heuristic function to evaluate nodes.

[0067]

[0068] In the formula, n x n y g represents the coordinates of the current node in the grid. x g yHere are the coordinates of the target point. The forward and reverse expansion search strategy is based on the traditional heuristic search algorithm, expanding nodes simultaneously from the starting point and the target point towards each other. This reduces search complexity and computation time to some extent. Their cost functions are f1(n) and f2(n), as shown in the following equations:

[0069]

[0070] The algorithm ends and obtains the global trajectory when node n simultaneously satisfies the following two conditions.

[0071] (1) The forward path and the reverse path lead to the same node, i.e., f1(n) = f2(n);

[0072] (2) The value of g1(n)+g2(n) is the smallest.

[0073] When traditional search algorithms employ a forward and reverse expansion search mechanism, they ideally meet at the center of the map, resulting in faster path planning while retaining the high accuracy of traditional algorithms. However, in complex two-dimensional environments, using Euclidean distance as the heuristic function to expand nodes requires searching eight nodes at each step. During node expansion, the distance h(n) from the current node to the target point is always less than the actual distance cost, leading to wasted computation time and memory space. This makes it difficult for the forward and reverse expansion search paths to meet at the map center, and the search time can even be longer than traditional algorithms. Furthermore, factors such as map size and obstacle locations result in numerous redundant points and turning points in the calculated trajectory. To address these issues, this embodiment improves the number of nodes, the weight of the heuristic function, and the selection of special nodes in the forward and reverse expansion search algorithm, achieving secondary optimization of the search path.

[0074] See Figure 3 To avoid generating a large number of invalid nodes during the search process and reduce computational overhead, this embodiment reduces the number of search nodes in the traditional algorithm by expanding the number of nodes from the original nine-square grid centered on the node to eight nodes excluding the current node (e.g., ...). Figure 3 a) Reduced to five (e.g.) Figure 3 (b) Figure 3 (c)); Figure 3 In the diagram, the dark square represents the current node, and the light square represents the next batch of points to be added to the OpenList. The range between the two arrows indicates the deflection direction of the current node towards the target point. Each 60° region is defined, and the range is defined by clockwise rotation. Figure 3 (b) The range is [30°, -30°), and the two cases of deflection direction are as follows: Figure 3 (a) Figure 3 As shown in (b).

[0075] Reducing the number of search nodes can decrease the number of nodes in the forward and reverse expansion search algorithms, but this inevitably sacrifices some valuable path nodes. Therefore, during the forward and reverse node expansion processes, the minimum cost point in the opposite direction is treated as the target point for the current search direction, and the heuristic function is improved by introducing an optimal trajectory weight function λ. j * (j = 1, 2), establish the cost evaluation function as shown below:

[0076]

[0077] In the formula: f1(n) is the cost function for expanding nodes from the starting point to the target point; f2(n) is the cost function for expanding nodes from the target point to the starting point; g1(n) is the actual cost value from the starting point to node n, g2(n) is the actual cost value from the target point to node n; h1(n) is the estimated cost value of the distance from node n to the target point, and h2(n) is the estimated cost value of the distance from node n to the starting point;

[0078]

[0079] σ1 and σ2 are the angles between the current node in the forward and reverse searches and the line connecting the minimum cost value in the opposite search, respectively. The improvement lies in increasing the weight of the heuristic function in the early stage of node expansion. As the estimated cost value of the current node approaches 0, the weight coefficient approaches 1, and the estimated cost value approaches the actual value. Dynamically adjusting the weight of the heuristic function reduces the cost error between the estimated cost value and the actual value when the current node is far from the target point. Introducing angle information corrects the search direction, making nodes in the forward and reverse directions attract each other, further accelerating the algorithm's search speed in complex obstacle environments.

[0080] Furthermore, while the optimized forward and reverse expansion search algorithm can quickly find a relatively short path, redundant points and turning points still exist, increasing the travel distance and making the automated guided vehicle (AGV) prone to collisions during frequent turns. Therefore, redundant points and turning points are considered special nodes in the search process. A special node selection scheme is introduced into the initial path to optimize the path from the starting point to the target point in the following two situations: first, the lines connecting the two nodes are collinear; second, there are obstacles between the lines connecting the two nodes. The specific optimization steps are as follows:

[0081] (1) Connect all nodes in sequence to generate a temporary path. If the current node is collinear with the nodes before and after it, then discard it.

[0082] (2) After collinear nodes are removed, let the nodes in the path be {P}. kGiven |k = 1, 2, ..., n}, let P1 be the parent node. Starting from the parent node P1, connect the next node in sequence until... It intersects with an obstacle. At this point, the special node is P. k-1 The formal path became Update the parent node P1 to P k-1 Repeat the above steps until all nodes have been traversed.

[0083] Because the optimal path generated by the forward and reverse expansion search algorithms is not smooth enough, many turning points still exist in certain parts of the path. Therefore, a cubic uniform B-spline curve is used to optimize the path at the turning points, as shown by the formula:

[0084]

[0085] The mathematical expression for the B-spline curve when k=3 is:

[0086]

[0087]

[0088] In the formula, 0 ≤ x ≤ 1; i = 0, 1, ..., k-1;

[0089] A cubic B-spline curve is a cubic uniform B-spline curve when the interpolation between the node vectors is constant. The mathematical expression for the i-th segment of the cubic uniform B-spline curve is:

[0090]

[0091] From equations (1), (3), and (4), the mathematical expression for the basis function of a cubic uniform B-spline curve can be obtained as follows:

[0092]

[0093] Substituting equation (5) into equation (4), we can express it in matrix form as follows:

[0094]

[0095] After smoothing the inflection points using B-spline curves, a relatively better shortest path is obtained by replacing the polyline with a curve. After removing redundant and inflection points, the path contains only the starting point, special nodes, and the ending point. This shortens the travel distance and reduces turns, ensuring the smooth operation of the automated guided vehicle (AGV) in complex smart factories. Finally, this path is output as the optimal path in this algorithm.

[0096] After the forward and reverse expansion search algorithm plans a feasible path, a local path planning algorithm is used to plan the local driving strategy of the automated guided transport vehicle based on the path information and real-time obstacle information. Therefore, the velocity pairs (linear velocity and angular velocity) of the vehicle in two-dimensional space are selected, and appropriate constraints are imposed on them to simulate its trajectory at different speeds. A reasonable evaluation function is designed for this function, and the calculated optimal path is appropriately evaluated and filtered based on the evaluation function. The path with the highest evaluation is selected as the current optimal path, thus completing the overall path planning optimization.

[0097] 1. Speed ​​Constraint

[0098] Based on the constraints of the automated guided vehicle and the environment, the speed sampling can be limited to a certain range, as shown in the following formula:

[0099] V i ={(v,ω)|v∈[v min ,v max ],ω∈[ω min ,ω max ]};

[0100] In the formula, v min v max ω min ω max Let represent the minimum and maximum linear and angular velocities of the vehicle, respectively. Due to the influence of the motor torque, the vehicle also exhibits maximum acceleration and deceleration, the constraints of which are shown in the following formulas:

[0101]

[0102] In the formula, v c ω c The linear velocity and angular velocity at the current moment are... These are the maximum linear and angular accelerations of the car. These are the minimum linear and angular accelerations of the vehicle.

[0103] Considering the car's safety, it needs to stop before colliding with the obstacle; therefore, the speed must meet the following requirements:

[0104]

[0105] In the formula, dist(v, ω) is the shortest distance between the end of the car's trajectory and the obstacle on the corresponding (v, ω) path.

[0106] When the car's speed is limited to a certain range, the speed of travel along the planned path is evaluated:

[0107]

[0108] Where Vel(v, ω) is the speed evaluation function for whether the car's speed on the corresponding (v, ω) path is within the constraint range, and V is the constraint range for the car's speed. The more the car's speed attribute conforms to V = V... i +V j +V k The higher the evaluation, the better the planned path.

[0109] 2. Maximum deflection angle constraint

[0110] Let the trajectory of the i-th segment be represented by α in the two-dimensional plane. i =(x i -x i-1 y i -y i-1 ),like Figure 4 As shown, the constraint is expressed as:

[0111]

[0112] Let Deflect(v,ω) be the deflection evaluation function of the automated guided vehicle, representing the deflection angle between the current trajectory segment and the target point. Considering collision avoidance and its own properties, this deflection angle should not be too large. Therefore, the following deflection angle evaluation function is defined:

[0113]

[0114] 3. Reasonableness of route changes

[0115] To ensure the effectiveness and safety of the obtained optimal path, sensors in the environment need to continuously scan the moving target. Assuming at time t... a The dynamic threat trajectory is obtained at all times as θ a (t), and generate a local obstacle avoidance trajectory; t b (t b >t a Dynamic threat trajectory θ at time ) b (t) and θ a (t) If the path change rationality is satisfied, the obstacle avoidance path of the car remains unchanged; otherwise, the obstacle avoidance path is replanned. The expression for the path change rationality is:

[0116] max(|θ a (t)-θ b (t)|)<ε;

[0117] In the formula, t∈[t a , t b ε is a constant, used as the criterion for deciding whether to replan the route;

[0118] The reasonableness evaluation function for path change is Change(v, ω):

[0119]

[0120] The lower the maximum value of the dynamic threat trajectory change is compared to the set ε value, the higher the evaluation of the path; if the maximum value is equal to or greater than the ε value, the path segment is canceled and replanned.

[0121] 4. Comprehensive evaluation function

[0122] The evaluation criterion for the function is: the selected path enables the automated guided vehicle to avoid obstacles during its journey and to move quickly toward the target point with reasonable changes in deflection angle. The designed evaluation function is as follows:

[0123] G(v,ω)=γ*Change(v,ω)+δ*Deflect(v,ω)+β*Vel(v,ω);

[0124] In the formula, Change(v, ω) is the path change rationality evaluation function corresponding to the (v, ω) trajectory; Deflect(v, ω) is the deflection angle evaluation function of the car; Vel(v, ω) is the speed evaluation function of the current simulated speed; and γ, δ, and β are the weighting coefficients of each evaluation function.

[0125] If the evaluation function determines that the evaluation of a local path is too low, the trajectory segment can be deleted, improved, or a new trajectory segment can be selected. The local paths with high evaluation can be retained, and the optimal global path can be obtained through the improved optimal path algorithm.

[0126] In the above process: This invention achieves intelligent scheduling of vehicles based on a three-level communication strategy involving the main control platform, the vehicle, and sensors. The main control platform, through environmental and vehicle information fed back by sensors, can monitor the real-time status of each vehicle and adjust its path strategy in a timely manner. Each level performs its specific function, forming a new scheduling method that promotes intelligent factory operations and automated logistics. Furthermore, this invention reduces the computational memory and time of the forward and reverse expansion search algorithm based on a search node reduction strategy. It divides the angle between the current node and the target point, specifying two point selection schemes within different angle ranges, reducing the number of nodes searched per step from 8 to 5, thus improving algorithm efficiency. In addition, this invention introduces a strategy based on the optimal trajectory weight function to further enhance the algorithm's search speed. Since some valuable points are inevitably sacrificed during path planning, h(n) is weighted, and the weights of the heuristic function are dynamically adjusted. Introducing angle information helps correct the search direction, which is beneficial for the algorithm to quickly move towards the target point in complex environments.

[0127] Therefore, the present invention achieves the following advantages:

[0128] (1) By adopting the scheme of reducing search nodes, a large number of invalid nodes are avoided, the computational overhead is reduced, the algorithm running speed is improved, and an optimal path can be quickly planned for the automated guided transport vehicle in factories with large cargo throughput, thereby improving work efficiency.

[0129] (2) By adopting a weighted approach based on the traditional formula, the cost error between the estimated cost and the actual cost is reduced when the current node is far from the target point, so that the estimated cost is close to the actual cost. The two nodes attract each other, effectively improving the speed of forward and reverse search and planning an optimal path.

[0130] (3) By adopting a special node selection scheme and using the smoothing strategy of cubic B-spline curves, redundant points and turning points in the path are removed, the shortest path is obtained, and collisions caused by multiple turns are avoided, thus ensuring the safety and reliability of the automated guided transport vehicle during operation.

[0131] (4) By adopting a reasonable evaluation function, not only is the running speed of the automatic guided transport vehicle evaluated, but also the deflection angle between the current node and the target point and the rationality of the path change are evaluated. This makes the designed path not only the shortest distance, but also greatly reduces the deflection angle, has a high effective distance ratio during the driving process, and is less affected by dynamic and static obstacles, so that it can adapt well to complex environments.

[0132] The above are preferred embodiments of the present invention, but the embodiments of the present invention are not limited to the above content. Any changes, modifications, substitutions, combinations, or simplifications made without departing from the spirit and principle of the present invention shall be considered equivalent substitutions and shall be included within the protection scope of the present invention.

Claims

1. A path planning method based on three-level communication and forward / reverse expansion search strategy, characterized in that, Includes the following steps: S1. Divide the factory into different areas and deploy sensors accordingly; S2. Collect various parameters of the vehicle through sensors and send the collected information to the main control platform; S3. The main control platform analyzes the collected information to determine whether the vehicle's forward path should be adjusted. When it is necessary to adjust the forward path of the car, the main control platform plans the forward path of the car by using a forward and reverse expansion search mechanism based on the heuristic search algorithm to obtain the planned path. The main control platform plans the vehicle's forward path using a heuristic search algorithm with a forward and reverse expansion search mechanism, including the following steps: S31. Reduce the number of search nodes in the traditional heuristic search algorithm. Reduce the number of nodes in the traditional heuristic search algorithm from eight nodes in the original nine-square grid centered on the node, excluding the current node, to five nodes. S32. In the traditional heuristic search algorithm's node expansion process in both the forward and reverse directions, the minimum cost point in the opposite direction is considered as the target point for expansion in this search direction, and an optimal trajectory weight function is introduced. Establish the cost evaluation function shown below: ； In the formula: f1(n) is the cost function for expanding nodes from the starting point to the target point; f2(n) is the cost function for expanding nodes from the target point to the starting point; The actual cost from the starting point to node n. The actual cost from the target point to node n; Estimate the cost of the distance from node n to the target point. Estimate the cost of the distance from node n to the starting point; It is a constant, and ; in, ； In the formula, These are the angles between the current node in the forward and reverse search processes and the line connecting the minimum cost of the opposite search; Step S33: After obtaining the cost evaluation function, redundant points and turning points are considered as special nodes in the search process; all nodes are connected sequentially to generate a temporary path. If the current node is collinear with the nodes before and after it, it is a collinear node and needs to be removed; after the collinear nodes are removed, let the nodes in the track be... Let P1 be the parent node, and start from the parent node P1 to connect the next node in sequence until... Intersecting with an obstacle; at this point, the special node is P. k-1 The official flight path became Then update the parent node P1 to P. k-1 Repeat the above steps until all nodes have been traversed; S4. Repeat steps S2 and S3 to obtain multiple planned paths; S5. The multiple planned paths are evaluated by constructing a comprehensive evaluation function, and the planned path with the highest evaluation is selected as the optimal path. The construction of the comprehensive evaluation function includes the following steps: S51. Constructing a velocity evaluation function ; S52. Construct the deflection evaluation function for the car. ; S53. Constructing a path change rationality evaluation function The steps are as follows: S531, The vehicle is continuously scanned by sensors. The dynamic threat trajectory is constantly monitored. and generate local obstacle avoidance trajectories; in Dynamic threat trajectory at any given moment and Does the path change meet the reasonableness requirement? ； In the formula , ; It is a constant, used as the criterion for deciding whether to replan the route; S532, Evaluation function for the rationality of path changes : ； When the maximum value of the dynamic threat trajectory change is smaller than the set value... The higher the value, the higher the evaluation of the path; if its maximum value is equal to or greater than a certain value, the higher the evaluation. If the value is not specified, the path segment will be cancelled and replanned. S54. Constructing a comprehensive evaluation function ,in, ； In the formula, These are the weighting coefficients for each evaluation function.

2. The path planning method based on three-level communication and forward / reverse expansion search strategy according to claim 1, characterized in that, In step S1, the parameters include, but are not limited to, the position and speed of the vehicle in different areas.

3. The path planning method based on three-level communication and forward / reverse expansion search strategy according to claim 2, characterized in that, In step S3, when the sensor reports an obstacle ahead, the vehicle's path is adjusted again at the current position to achieve local obstacle avoidance. At the same time, the sensor should transmit the vehicle's position information in real time and display it as an icon on the main control platform to facilitate monitoring of the vehicle's operation and timely adjustments if problems occur.

4. The path planning method based on three-level communication and forward / reverse expansion search strategy according to claim 3, characterized in that, In step S33, a cubic uniform B-spline curve is used to optimize the path at the inflection point. After smoothing the inflection point with the cubic uniform B-spline curve, the curve is used instead of the broken line to obtain the shortest path. After the redundant nodes and inflection points in the shortest path are removed, the shortest path only contains the starting point, special nodes and the ending point.

5. The path planning method based on three-level communication and forward / reverse expansion search strategy according to claim 4, characterized in that, In step S51, a velocity evaluation function is constructed. The steps are as follows: S521. Apply speed constraints to the trolley: ； In the formula, These are the minimum and maximum linear velocities of the vehicle, respectively. These represent the minimum and maximum angular velocities of the trolley, respectively. S522. Apply acceleration constraints to the trolley: ； In the formula, , These are the linear velocity and angular velocity at the current moment; , These are the maximum linear and angular accelerations of the vehicle. , These are the minimum linear and angular accelerations of the vehicle. S513. In order to ensure that the car stops before colliding with the obstacle, the speed must also meet the following requirements: ； In the formula, For the corresponding The closest distance between the end of the car's trajectory and an obstacle on the path; S514, Speed ​​Evaluation Function for: ； in, The constraint range for the car's speed, when the car's speed attribute meets... If the result is positive, it indicates that the current path planning has achieved the best evaluation; otherwise, a new path planning should be implemented.

6. The path planning method based on three-level communication and forward / reverse expansion search strategy according to claim 5, characterized in that, In step S52, the deflection evaluation function of the vehicle is constructed. for: ； In the formula, This represents the deflection angle of the next path segment relative to the previous path segment.

Images