Scheduling method based on improved A star algorithm

Abstract

According to the scheduling method based on the improved star-A algorithm, path nodes planned based on the star-A algorithm are proposed to generate an expansion model, and then collision avoidance expansion detection is implemented to judge whether there is model overlap; therefore, the situation that the tool collides, shifts, even falls off and is damaged due to the fact that the size of the bearing tool is large or the steering radius is large in the operation process of the mobile robot is effectively avoided. The influence of interference blocking is considered, and a better path is dynamically selected. Comprising the following steps: step 1, path planning; step 2, collision avoidance expansion detection; step 2.1, searching a previously generated known vehicle expansion model; step 2.2, creating a detection queue; step 2.3, obtaining calculation parameters of the expansion model; 2.4, generating an expansion model of the current vehicle; step 2.5, vehicle expansion model interference judgment; 2.6, constructing a time window model to calculate the conflict occupation time of the current vehicle; step 2.7, calculating a congestion estimation value coefficient; and 2.8, carrying out secondary path planning by adopting an improved A star algorithm.

Description

Technical Field

[0001] The invention relates to a scheduling method applied to a mobile robot such as AGV. The method implements a collision avoidance and expansion detection strategy based on a path planning result of an A-star algorithm, and belongs to the field of logistics warehousing and automation control. Background Art

[0002] With the development and progress of industrial automation technology, various types of mobile robots such as AGVs have been widely used in logistics warehousing transfer yards and automated production sites to perform material turnover and transportation operations through omnidirectional driving functions.

[0003] Usually, logistics transfer sites have large workloads, narrow sites, and complex robot paths. In order to cope with the tight production cycle and the increasingly serious on-site vehicle congestion, the currently used scheduling methods must face the interference and conflict between the planned paths. The existing A* algorithm generally only uses the path length as the evaluation function when planning the path, resulting in similar routes for multiple AGVs, which further aggravates the congestion.

[0004] For example, the previously disclosed domestic patent application, publication number CN 114141034B, is entitled a traffic control method, system and electronic device. Upon receiving a request instruction from a vehicle, the road status information of the traffic control point is obtained. When the traffic control point is in an occupied state, the driving status information of all vehicles applying to pass through the same traffic control point is obtained, and the priority of the vehicles is divided according to the driving status information of the vehicles, and the vehicle with the highest priority is granted access rights. When multiple vehicles apply to pass through the same traffic control point, the processing system divides the priority of the vehicles according to the driving status information of the vehicles, wherein the driving status information includes the distance between the vehicle and the traffic control point, and the corresponding access rights are granted in sequence according to the priority sorting of the vehicles, thereby saving the waiting time for passing and improving the traffic efficiency. Before a vehicle with a high priority passes through the traffic control point, it is also determined whether the distance between the remaining vehicles to be passed and the traffic control point is within the safety threshold range, so as to take further deceleration and give way or redefine the priority order.

[0005] The above solution avoids collisions by occupying and waiting at conflicting points, but in complex situations, it will lead to long waiting times and cause congestion. For complex and narrow environments and paths with many vehicles, the existing technology does not address the collision problem of different paths. It only uses occupation to avoid collisions at the same resource point without interference from nearby vehicles and vehicles, and does not consider the efficiency problems caused by congestion.

[0006] In view of this, this patent application is proposed. Summary of the invention

[0007] The scheduling method based on the improved A-star algorithm described in the present invention aims to solve the problems existing in the above-mentioned prior art and proposes an expansion model for generating path nodes based on the A-star algorithm planning, and then implements collision avoidance expansion detection to determine whether there is model overlap, thereby effectively avoiding the occurrence of collision, displacement or even falling and damage of tooling due to the large volume of the tooling carried or the large turning radius of the mobile robot during operation, and further dynamically optimizes the path according to the congestion waiting situation.

[0008] To achieve the above design objectives, this application proposes a scheduling method based on an improved A-star algorithm, comprising the following steps:

[0009] Step 1: Path planning;

[0010] The A-star algorithm is used to calculate the path from the starting point of the task to the target point;

[0011] Step 2: Collision avoidance expansion detection;

[0012] Analyze and calculate the expansion model of the planning path nodes generated in step 1, build a real-time evaluation function based on the model interference and congestion conditions, and perform secondary planning to obtain a better path;

[0013] Step 2.1, find the known vehicle expansion model generated previously;

[0014] Step 2.2, create a detection queue;

[0015] The detection range is constructed with the path step length of the current vehicle as the radius and the position of the current vehicle as the center. The expansion models of all known vehicles in the detection range are found and a detection queue is created.

[0016] Step 2.3, obtaining calculation parameters of the expansion model;

[0017] The calculation parameters include path segment attribute parameters and expansion coefficient;

[0018] Step 2.4, generating an expansion model of the current vehicle;

[0019] Generate an expansion model based on the current vehicle's body, tooling dimensions and steering conditions;

[0020] When the tooling is not loaded, the expansion model is generated by multiplying the AGV vehicle size by the expansion coefficient;

[0021] When the tooling is loaded, the outer dimensions of the tooling are expanded;

[0022] Step 2.5, generate an expansion model of all vehicles in the detection queue and make a judgment;

[0023] According to step 2.4, the expansion models of all vehicles in the detection queue are generated in sequence; and the expansion models of the current vehicle are compared one by one to determine whether there is overlap interference;

[0024] Step 2.6, construct a time window model to calculate the conflict occupancy time of the current vehicle;

[0025] The time window model is:

[0026]

[0027] Among them, t ak represents the time window of the AGV marked as a on the road section k, represents the time when the AGV marked as a enters the road section k, Indicates the time when the AGV marked as a leaves the road section k;

[0028] Calculate the conflict waiting time according to the conflict type and calculate the conflict waiting estimate as follows:

[0029] goal.w = κvT

[0030] Among them, κ is the expansion waiting coefficient, which is configured according to the actual situation on site and has a value range of 0.5-0.8; v is the set AGV travel speed; T is the vehicle conflict waiting time;

[0031] Step 2.7, calculate the congestion estimation coefficient;

[0032] Evaluate each map node based on the map's real-time heat map and construct the node congestion cost;

[0033] According to the AGV heartbeat reporting status, the occupied thermal value B of each node within a period of time T is calculated. If there is a vehicle occupying the node every second, the thermal value is increased by one to calculate the congestion waiting coefficient as shown in the following formula:

[0034]

[0035] Among them, λ is the congestion waiting coefficient, and v is the AGV travel speed set at the node; the design idea is the reverse application of the pheromone strategy of the ant colony algorithm;

[0036] Step 2.8: Use the A-star algorithm for secondary path planning;

[0037] The improved A-star algorithm heuristic evaluation function is:

[0038] h(n)=(1+λ)(nx-goal.x|+|ny-goal.y)+goal.w

[0039] The expansion model of the AGV is generated according to the final planned path, and the expansion model of each node is deleted after the AGV passes through.

[0040] Furthermore, in the step 2.3, the path segment attribute parameters include angle, length, forward expansion area, backward expansion area, expansion area width, safety distance and vehicle reverse length.

[0041] Furthermore, in the step 2.4, when the vehicle passes the turning point, the size of the diagonal is used as the calculation input parameter of the model; the length and width values ​​of the tooling size are set to calculate the length of the diagonal, and the model is constructed with the length of the diagonal as the radius and the node as the center of the circle.

[0042] Furthermore, in step 2.4, when the AGV is traveling on a curve, the center point is calculated based on the reverse length l of the line segment, (x, y) is the position coordinate of the node, yaw is the angle of the line segment, and the calculated AGV center coordinate (x AGV ,Y AGV ); that is, the vehicle model size is taken as the length and width, and the line segment angle is taken as the model angle to construct a rectangular expansion model as shown in the following formula:

[0043] X AGV =xl*cos(yaw)

[0044] Y AGV =yl*sin(yaw)

[0045] Furthermore, in the step 2.6, the time window is defined as the time period from entering to leaving a node, which is divided into a reserved time period and an idle time period; the reserved time period is the time period occupied by the AGV passing through the node, and the idle time period is the middle interval time of the reserved time period for planning other AGVs.

[0046] As mentioned above, the scheduling method based on the improved A-star algorithm has the following advantages:

[0047] 1. This application analyzes and expands the nodes of the mobile robot's path planned by the A-star algorithm, so as to effectively determine whether a collision of the vehicle body or the carrying tooling will occur. This changes the disadvantage of the existing algorithm that only uses path distance as the evaluation criterion, and can reduce congestion in complex and dense operating environments, and avoid interference and collision of multiple mobile robots on different paths.

[0048] 2. This application takes into account the comprehensive conditions of various vehicle models and tooling, significantly improving the efficiency of turnover or transportation of tooling under complex working conditions, and avoiding damage such as tooling displacement and falling due to collisions. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] The present application will now be further described with reference to the following drawings.

[0050] Figure 1 It is a schematic diagram of node parameter configuration;

[0051] Figure 2 It is a schematic diagram of the AGV expansion entity calculation process;

[0052] Figure 3 This is a schematic diagram of center coordinate calculation;

[0053] Figure 4 is a schematic diagram of the process of generating an expansion model;

[0054] Figure 5 It is a time window three-dimensional model diagram;

[0055] Figure 6 It is a two-dimensional schematic diagram of the time window model; DETAILED DESCRIPTION

[0056] Embodiment 1, as Figures 1 to 6 As shown, the present application proposes a scheduling method based on an improved A-star algorithm, which can be applied to various types of mobile robots, such as the AGV described in this embodiment for transporting and circulating various types of goods.

[0057] The scheduling method based on the improved A-star algorithm includes the following steps:

[0058] Step 1: Path planning;

[0059] After accepting the task, each AGV obtains the starting point and target point to perform the task. Path planning refers to finding an optimal path from the starting point to the target point. The A-star algorithm can be used to calculate the optimal solution for the path.

[0060] Specifically, the A-star algorithm adds a heuristic function to the evaluation function of the Dijkstra algorithm, so that it moves toward the target point when searching for a path. First, the static map is converted into a grid map, the current point is taken as the parent node, the estimated values ​​of all the surrounding adjacent nodes are calculated, and the node with the smallest estimated value is selected to move in the direction; secondly, the node is taken as the new parent node and the search continues until the end point is reached, and finally the node sequence from the starting point to the target point is obtained.

[0061] When calculating the estimated value, it is necessary to calculate according to the valuation function. The expression of the valuation function is:

[0062] f(n)=g(n)+h(n)

[0063] Where n is a node in the map, g(n) is the minimum cost estimate from the starting point to node n, that is, the minimum distance from the starting point to the current point, and h(n) is the minimum estimated cost from the current node n to the target point, which is called the heuristic function. h(n) can choose a variety of functions in the map to evaluate the distance between two nodes. Let the coordinates of node n in the map be (nx, ny), and the coordinates of the destination in the map be (goal.x, goal.y).

[0064] The Manhattan distance formula is preferred to represent the heuristic function of the A* algorithm. If the coordinates of the first point in the plane coordinate system are (x1, y1) and the coordinates of the second point are (x2, y2), then the Manhattan distance between the two points is:

[0065] d=|x1-x2|+|y1-y2|

[0066] The heuristic function corresponding to the A* algorithm can be expressed as:

[0067] h(n)=|nx-goal.x|+|ny-goal.y|

[0068] When an AGV starts path planning, it has the global map information and relevant information of the task to be performed transmitted by the central processing system. It uses the grid map constructed by the global map and the starting and target points of the task as input, and obtains a collision-free path sequence from the starting point to the target point through the A* algorithm.

[0069] The specific steps are as follows:

[0070] Step 1.1, first set up two sets, namely the open table and the close table;

[0071] Step 1.2, put the starting point into the open table, set the parent node of the starting point to be empty, calculate the estimated value of the points that can be reached around the starting point, and put the estimated value into the open table;

[0072] Step 1.3: If the open table is empty, it means that the path does not exist and the search stops; otherwise, select the point with the smallest estimated value from the open table as the new node, add the node to the close table, and remove it from the open table;

[0073] Step 1.4: If the current node position is the same as the target point, the path search is completed and step 1.7 is executed; otherwise, step 1.5 is continued;

[0074] Step 1.5: Determine the points that can be reached around the current node;

[0075] If a neighboring node is recorded in the close table or is an obstacle grid point, skip the node; if the neighboring node is not in the open table, add the neighboring node to the open table, set the parent node of the node as the current node, and calculate the estimated value of the node; if the node is already in the open table and the previously calculated g value is greater than the g value of the current node, reset the parent node of the neighboring node to the current node;

[0076] Step 1.6, jump to step 1.3;

[0077] Step 1.7: Backtrack the path according to the close table, take the target point as the first node, select the parent node of each node in the close table in turn until the starting point is traced back. The nodes selected from the starting point to the target point are the planned path sequence; finally, a path from the starting point to the target point is obtained.

[0078] Step 2: Collision avoidance expansion detection;

[0079] Usually, the tooling carried by the AGV during operation is large in volume, and collisions may occur between the tooling during turning. The following is an analytical calculation of the planning path nodes generated in step 1 to perform volume expansion;

[0080] Analyze and calculate the expansion model of the planning path nodes generated in step 1, build a real-time evaluation function based on the model interference and congestion conditions, and perform secondary planning to obtain a better path;

[0081] Step 2.1, find the known vehicle expansion model generated previously;

[0082] Step 2.2, create a detection queue;

[0083] The detection range is constructed with the path length of the current vehicle as the radius and the position of the current vehicle as the center. The expansion models of all known vehicles in the detection range are found and a detection queue is created.

[0084] Step 2.3, obtaining calculation parameters of the expansion model;

[0085] like Figure 1 As shown, the calculation parameters include path segment attribute parameters and expansion coefficient;

[0086] The path segment attribute parameters include angle, length, forward expansion area, backward expansion area, expansion area width, safety distance and vehicle reverse length;

[0087] The expansion coefficient is used to slightly expand the model to prevent tooling misalignment or vehicle yaw. The expansion coefficient can be manually adjusted and configured according to the on-site environment.

[0088] Specifically, the vehicle parameters include vehicle size data (length, width); when the path segment is a curve, calculate the forward length; path length Length = segment length * sin45° + safety distance; reverse extension line, used to calculate the forklift model, reverse length = length * sin45° + vehicle width / 2; angle is the angle of the line; when moving forward, the length of the expansion model is equal to the length of the forward expansion area, and when moving backward, the length of the expansion model is equal to the length of the backward expansion area; the width of the expansion model is equal to the expansion area width preset by the system;

[0089] Step 2.4, generating an expansion model of the current vehicle;

[0090] like Figure 2 As shown, an expansion model is generated according to the body, tooling size and steering conditions of the current vehicle;

[0091] When the tooling is not loaded, the expansion model is generated by multiplying the AGV vehicle size by the expansion coefficient;

[0092] When the tooling is loaded, the tooling's dimensions are expanded; when the vehicle passes the turning point, the tooling's dimensions are further expanded, that is, the diagonal dimensions are used as the calculation parameters of the model;

[0093] For example, when the AGV carrying the tooling is located at the turning node, if the tooling size is set, the length and width of the tooling are used to calculate the length of the diagonal, and the model is constructed with the diagonal length as the radius and the node as the center;

[0094] When the AGV is traveling on a curve, the center point is calculated based on the reverse length l of the line segment, (x, y) is the position coordinate of the node, yaw is the angle of the line segment, and the calculated AGV center coordinate (X AGV ,Y AGV ), that is, the vehicle model size is taken as the length and width, and the line segment angle is taken as the model angle to construct a rectangular expansion model; as shown in the following formula,

[0095] X AGV =xl*cos(yaw)

[0096] Y AGV =yl*sin(yaw)

[0097] The coordinate calculation diagram is as follows: Figure 3 As shown, the generated expansion model is Figure 4 shown.

[0098] Step 2.5, generate an expansion model of all vehicles in the detection queue and make a judgment;

[0099] According to step 2.4, the expansion models of all vehicles in the detection queue are generated in sequence;

[0100] Compare each with the expansion model of the current vehicle to determine whether there is overlapping interference;

[0101] Since the expansion models of vehicles are divided into two types: circular and rectangular, the judgment of overlapping interference is divided into three cases: judgment between circles and circles, circles and rectangles, and rectangles and rectangles;

[0102] Step 2.6, construct a time window model to calculate the conflict occupancy time of the current vehicle;

[0103] like Figure 5 and Figure 6 As shown, the time window can be defined as the time period from entering to leaving a node, which can be divided into a reserved time period and an idle time period. The reserved time period is the time period occupied by the AGV passing through the node, and the idle time period is the middle interval time of the reserved time period for planning other AGVs.

[0104] The time window model is:

[0105]

[0106] where t ak represents the time window of the AGV marked as a on the road section k, represents the time when the AGV marked as a enters the road section k, Indicates the time when the AGV marked as a leaves the road section k;

[0107] The following conflict types are included:

[0108] (1) Conflict;

[0109] If the time window of section k is occupied by any two AGVs p and q at the same time, and the two AGVs enter and exit the node in opposite directions, it is proved that a conflict occurs. The time window criterion for a conflict is:

[0110]

[0111] (2) Node conflict;

[0112] If the difference between the departure time of any AGVp and the entry time of AGVq at node v is less than the threshold and the movement directions are different, it means that a node conflict occurs. The time window criterion when a node conflict occurs is:

[0113]

[0114] Where δ is the time safety threshold.

[0115] (3) Chasing conflict;

[0116] If the time window sequence of the intersection node v and the previous node of the two AGVs changes, it means that a chase conflict occurs. The time window judgment criterion when a chase conflict occurs is:

[0117]

[0118] Calculate the conflict waiting time according to the conflict type and calculate the conflict waiting estimate as follows:

[0119] goal.w = κvT

[0120] Among them, κ is the expansion waiting coefficient, which is configured according to the actual situation on site and has a value range of 0.5-0.8; v is the set AGV travel speed; T is the vehicle conflict waiting time;

[0121] Step 2.7, calculate the congestion estimation coefficient;

[0122] Evaluate each map node based on the map's real-time heat map and construct the node congestion cost;

[0123] According to the AGV heartbeat reporting status, the occupied thermal value B of each node within a period of time T is calculated. If there is a vehicle occupying the node every second, the thermal value is increased by one to calculate the congestion waiting coefficient as shown in the following formula:

[0124]

[0125] Among them, λ is the congestion waiting coefficient, and v is the AGV travel speed set on the node; the design idea is the reverse application of the pheromone strategy of the ant colony algorithm.

[0126] Step 2.8, use the improved A-star algorithm for secondary path planning;

[0127] The improved A-star algorithm heuristic evaluation function is:

[0128] h(n)=(1+λ)(nx-goal.x|+|ny-goal.y)+goal.w

[0129] The expansion model of the AGV is generated according to the final planned path, and the expansion model of each node is deleted after the AGV passes through.

[0130] The path planned using the above-mentioned method steps of the present application can comprehensively consider factors such as interference, blocking and waiting in a dense environment, so as to plan a more reasonable path and effectively improve production and handling efficiency and reduce the occurrence of collision and deadlock.

[0131] The above contents and the embodiments given in the accompanying drawings are only preferred solutions for achieving the purpose of the present invention. Those skilled in the art can be inspired by this and directly deduce other alternative structures that conform to the design concept of the present invention. Other structural features obtained thereby should also belong to the scope of the solution described in the present invention.

Claims

1. A scheduling method based on an improved A-star algorithm, applied to a mobile robot, characterized in that: The following steps are included: Step 1: Path planning; The A-star algorithm is used to calculate the optimal path from the starting point of the task to the target point; Step 2: Collision avoidance expansion detection; Analyze and calculate the expansion model of the planning path nodes generated in step 1, build a real-time evaluation function based on the model interference and congestion conditions, and perform secondary planning to obtain a better path; Step 2.1, find the known vehicle expansion model generated previously; Step 2.2, create a detection queue; The detection range is constructed with the path step length of the current vehicle as the radius and the position of the current vehicle as the center. The expansion models of all known vehicles in the detection range are found and a detection queue is created. Step 2.3, obtaining calculation parameters of the expansion model; The calculation parameters include path segment attribute parameters and expansion coefficient; Step 2.4, generating an expansion model of the current vehicle; Generate an expansion model based on the current vehicle's body, tooling dimensions and steering conditions; When the tooling is not loaded, the expansion model is generated by multiplying the AGV vehicle size by the expansion coefficient; When the tooling is loaded, the outer dimensions of the tooling are expanded; Step 2.5, generate an expansion model of all vehicles in the detection queue and determine the interference; According to step 2.4, the expansion models of all vehicles in the detection queue are generated in sequence; and the expansion models of the current vehicle are compared one by one to determine whether there is overlap interference; Step 2.6, construct a time window model to calculate the conflict occupancy time of the current vehicle; The time window model is: Among them, t ak represents the time window of the AGV marked as a on the road section k, represents the time when the AGV marked as a enters the road section k, Indicates the time when the AGV marked as a leaves the road section k; Calculate the conflict waiting time according to the conflict type and calculate the conflict waiting estimate as follows: goal.w = κvT Among them, κ is the expansion waiting coefficient, which is configured according to the actual situation on site and has a value range of 0.5-0.8; v is the set AGV travel speed; T is the vehicle conflict waiting time; Step 2.7, calculate the congestion estimation coefficient; Evaluate each map node based on the map's real-time heat map and construct the node congestion cost; According to the AGV heartbeat reporting status, the occupied thermal value B of each node within a period of time T is calculated. If there is a vehicle occupying the node every second, the thermal value is increased by one to calculate the congestion waiting coefficient as shown in the following formula: Among them, λ is the congestion waiting coefficient, and v is the AGV travel speed set at the node; the design idea is the reverse application of the pheromone strategy of the ant colony algorithm; Step 2.8, use the improved A-star algorithm for secondary path planning; The improved A-star algorithm heuristic evaluation function is: h(n)=(1+λ)(nx-goal.x|+|ny-goal.y)+goal.w The expansion model of the AGV is generated according to the final planned path, and the expansion model of each node is deleted after the AGV passes through.

2. The scheduling method based on the improved A-star algorithm according to claim 1, characterized in that: In the step 2.3, the path segment attribute parameters include angle, length, forward expansion area, backward expansion area, expansion area width, safety distance and vehicle reverse length.

3. The scheduling method based on the improved A-star algorithm according to claim 1, characterized in that: In the step 2.4, when the vehicle passes the turning point, the size of the diagonal is used as the calculation input parameter of the model; the length and width values ​​of the tooling size are set to calculate the length of the diagonal, and the model is constructed with the length of the diagonal as the radius and the node as the center of the circle.

4. The scheduling method based on the improved A-star algorithm according to claim 1 is characterized in that: In step 2.4, when the AGV is traveling on a curve, the center point is calculated based on the reverse length l of the line segment, (x, y) is the position coordinate of the node, yaw is the angle of the line segment, and the calculated AGV center coordinate (x AGV ,Y AGV ); Taking the vehicle model size as the length and width and the line segment angle as the model angle, a rectangular expansion model is constructed as shown in the following formula: X AGV =x-l*cos(yaw) AND AGV =yl*sin(yaw) 5. The scheduling method based on the improved A-star algorithm according to claim 1 is characterized in that: In step 2.6, the time window is defined as the time period from entering to leaving a node, which is divided into a reserved time period and an idle time period; the reserved time period is the time period occupied by the AGV passing through the node, and the idle time period is the middle interval time of the reserved time period for planning other AGVs.

Images