Automatic loading and unloading unmanned vehicle loading and unloading method and system

Abstract

The present application relates to automatic loading and unloading technical field, especially to a kind of automatic loading and unloading unmanned vehicle loading and unloading method and system, method includes: according to loading and unloading task and operating area map, initial global shortest path and path point sequence are obtained by A algorithm;With any unmanned vehicle as target vehicle, the rest is matching vehicle, risk characteristic value of path point is quantified by DTW algorithm interactive analysis, adaptability score is obtained by combining direction deviation, and screening point is local target point;Based on the risk characteristic value and local target point of dynamic updating, the path is adjusted in real time using improved artificial potential field method, and the loading and unloading task is completed by cyclic iteration.The present application combines A algorithm and improved artificial potential field method, determines global path by A algorithm, quantifies risk characteristic value by DTW algorithm and dynamically updates local target point, adjusts potential field intensity to avoid dynamic obstacles, guarantees path optimization and driving stability, and reduces conflict congestion.

Description

Technical Field

[0001] This invention relates to the field of automated loading and unloading technology. In particular, it relates to an automated loading and unloading method and system for unmanned vehicles. Background Technology

[0002] Automated handling vehicles (AWDVs) are devices that enable unmanned material handling in various scenarios. With the development of automation technology, AWDVs can automatically receive handling tasks via the cloud and perform unmanned handling throughout the entire process, following designated areas and routes, thus creating a new, more efficient, safer, and more economical transportation model. These AWDVs have broad application prospects in logistics, warehousing, and other fields, significantly improving transportation efficiency and safety.

[0003] Path planning is a crucial step in achieving efficient operation of automated loading and unloading vehicles (AWDVs). Proper path planning not only ensures the safe and efficient completion of tasks in complex environments but also effectively avoids collisions and congestion, improving overall transportation efficiency. In scenarios involving multiple vehicles working collaboratively, path planning is even more critical, directly impacting the efficiency of collaboration between WADVs and the timeliness of task completion.

[0004] In the path planning process of automated loading and unloading vehicles, traditional A While the algorithm can effectively plan the globally optimal path to avoid static obstacles, it is based on a static map search and cannot handle dynamic obstacles (such as other autonomous vehicles in motion), easily leading to path conflicts and congestion. Traditional artificial potential field methods, although able to overcome the problem of dynamic obstacles, lack a global perspective, are prone to getting trapped in local minima, or leading to suboptimal paths, thus failing to guarantee the overall efficiency of automated loading and unloading vehicles. Summary of the Invention

[0005] To solve the traditional A Algorithms and artificial potential field methods have problems in autonomous vehicle path planning, namely, the inability to handle dynamic obstacles and the lack of a global perspective, which leads to path conflicts, congestion and low efficiency. This invention provides solutions in the following aspects.

[0006] In the first aspect, an automated loading and unloading unmanned vehicle (AWDV) loading and unloading method includes: obtaining an initial global shortest path and corresponding path point sequence based on the loading and unloading tasks and operating area map of each AWDV; taking any currently operating AWDV as the target vehicle and other currently operating AWDVs as matching vehicles, and using the DTW matching algorithm to perform interactive analysis between the path point sequence of the target vehicle and the path point sequence of the matching vehicles based on the path point sequence, quantifying the risk characteristic value of each path point, calculating the cosine similarity between the gravitational direction of each path point and the current driving direction, and obtaining the direction association weight; calculating the suitability score of the path point as a local target point based on the risk characteristic value and direction association weight of the path point, forming a score curve, and selecting the first inflection point closest to the current driving position as the local target point; adjusting the path in real time during driving using an improved artificial potential field method based on the dynamically updated path point risk characteristic value and the local target point; and iteratively executing the dynamic iteration process of global shortest path planning, risk characteristic value calculation, and real-time control until the AWDV reaches the destination and completes the automated loading and unloading task.

[0007] Preferably, obtaining the initial global shortest path and the corresponding path point sequence includes:

[0008] The running area map is gridded and obstacles are marked, using A The algorithm plans the global shortest path based on the cost function. It extracts the coordinates of all grid nodes on the global shortest path in the order in which the autonomous vehicles travel, forming an ordered set of path points. The ordered set of path points is then output as a sequence to obtain the path point sequence.

[0009] Preferably, the risk characteristic value is calculated in the following ways:

[0010] Obtain the number of automated loading and unloading vehicles currently in operation, and use any path point in the initial global shortest path of the target vehicle as the analysis path point, and use any path point in the initial global shortest path of any matching vehicle as the matching path point.

[0011] Calculate the cosine similarity between the driving direction of the analysis path point of the target vehicle and the driving direction of the driving direction of the matching path point of the matching vehicle, and use 1 minus the cosine similarity as the weight of the direction difference.

[0012] Calculate the normalized Euclidean distance between the analysis path points of the target vehicle and the matching path points of the matching vehicle, and use a negative exponential function to perform an exponential mapping on the normalized Euclidean distance to obtain the distance association weight.

[0013] Calculate the normalized Euclidean distance between the analysis path point of the target vehicle and the matching path point of the matching vehicle, and the normalized Euclidean distance between the next path point of the analysis path point of the target vehicle and the next path point of the matching path point of the matching vehicle, and calculate the ratio to obtain the path distance change trend weight.

[0014] Multiply the directional difference weight, distance correlation weight, and path distance change trend weight respectively to obtain the comprehensive interaction risk value between the analysis path point and the matching path point. Then, iterate through the comprehensive interaction risk values ​​of all matching vehicles other than the target vehicle, sum them, and take the average value to obtain the risk characteristic value of the analysis path point of the target vehicle.

[0015] Preferably, the calculation of the path points as the suitability score for local target points includes:

[0016] Using any path point in the initial global shortest path of the target vehicle as the analysis path point, the real-time driving direction of the target vehicle, the analysis path point, and the risk characteristic value of the analysis path point are obtained. The risk characteristic value is substituted into a negative exponential function to obtain the risk association weight. The similarity between the driving direction of the target vehicle and the gravitational direction of the analysis path point is calculated to obtain the direction association weight. The risk association weight is multiplied by the direction association weight to obtain the suitability score of the analysis path point as a local target point.

[0017] Preferably, the method of adjusting the path in real time during driving using an improved artificial potential field includes:

[0018] Calculate the path distance between the target vehicle's current position and the local target point. Use a negative exponential function to exponentially map the mean risk characteristic value of all path points between the target vehicle's current position and the local target point to obtain the gravity adjustment factor. Multiply the preset gravity gain coefficient, the path distance between the target vehicle's current position and the local target point, the gravity adjustment factor, and the unit direction vector from the target vehicle's current position to the local target point in sequence to obtain the gravity vector.

[0019] For each obstacle, a distance correlation factor is calculated using the correlation function between repulsion force and obstacle distance; the average risk characteristic value of all path points between the current position of the target vehicle and the local target point is added to 1 as the repulsion force adjustment factor; the preset repulsion force gain coefficient, repulsion force adjustment factor, distance correlation factor, and unit direction vector from the obstacle to the target vehicle are multiplied in sequence to obtain the repulsion force vector of a single obstacle.

[0020] The total repulsive force vector is obtained by summing the repulsive force vectors of all obstacles within the sensing range. The attractive force vector is then added to the total repulsive force vector to obtain the combined resultant force vector. Based on the combined resultant force vector, the real-time driving direction and adjustment intensity of the target vehicle are determined.

[0021] Preferably, when the DTW matching algorithm is executed, it only captures the path points where each automated loading and unloading unmanned vehicle is located at the current moment and the subsequent path point sequence for matching, excluding the path segments that have been traveled, so as to ensure the real-time and accuracy of risk assessment.

[0022] Preferably, the update frequency of the local target point is consistent with the update frequency of the path point risk feature value; after the automatic loading and unloading unmanned vehicle travels to the current local target point, it immediately recalculates the risk feature value of each path point in the remaining path segment, and selects the next local target point based on the updated risk feature value, repeating the loop until the global endpoint is reached.

[0023] Secondly, an automated loading and unloading unmanned vehicle loading and unloading system includes: a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the above-mentioned automated loading and unloading unmanned vehicle loading and unloading method is implemented.

[0024] The present invention has the following effects:

[0025] 1. This invention integrates A The advantages of algorithms and improved artificial potential field methods, through A The algorithm plans the global shortest path to solve the problems of traditional artificial potential field methods, which lack a global perspective and are prone to getting trapped in local minima. At the same time, it uses risk feature values ​​to adjust the potential field strength to achieve precise avoidance of dynamic obstacles, effectively reducing path conflicts and congestion.

[0026] 2. This invention quantifies the risk feature value of path points through the DTW matching algorithm and filters local target points by combining directional correlation weights. This ensures that the automated loading and unloading unmanned vehicle follows the globally optimal path, and improves driving stability and path adaptability by dynamically updating target points and adjusting the path, thus avoiding efficiency losses caused by suboptimal paths. Attached Figure Description

[0027] Figure 1 This is a flowchart of steps S1-S4 in an automatic loading and unloading unmanned vehicle loading and unloading method according to an embodiment of the present invention.

[0028] Figure 2 This is a structural block diagram of an automatic loading and unloading unmanned vehicle loading and unloading system according to an embodiment of the present invention. Detailed Implementation

[0029] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are some embodiments of the present invention, but not all embodiments.

[0030] Reference Figure 1 An automated loading and unloading method for unmanned vehicles includes steps S1-S4, as detailed below:

[0031] S1: Based on the loading and unloading tasks and operating area map of each automated loading and unloading vehicle, obtain the initial global shortest path and the corresponding path point sequence.

[0032] Obtain specific loading and unloading task information for the target vehicle, including cargo loading point (starting point), unloading point (ending point), and transportation task-related constraints (such as speed limits, loading and unloading area boundary requirements, etc.).

[0033] Obtain a complete electronic map of the transportation area, which must include the location information of static obstacles such as fixed buildings, equipment, and road boundaries within the area; the unmanned vehicle has completed initialization and has the basic control capabilities of map reception, path parsing, and path-based driving.

[0034] First, the acquired electronic map of the transportation area is divided into grids. The entire transportation area is divided into several uniform square grid units according to a preset fixed side length (e.g., 1 meter × 1 meter, which can be adjusted according to the size of the transportation area, the size of the unmanned vehicle, and the driving accuracy requirements).

[0035] Each grid cell is treated as an independent spatial node, defined by coordinates (such as two-dimensional planar coordinates). The grid cells containing static obstacles are uniquely identified in the map. At the same time, based on the static obstacle information in the map, grid cells containing static obstacles are marked (e.g., marked as "impassable"), while unmarked grid cells are marked as "passable", forming a grid map that can be used for path planning.

[0036] In the gridded map, the starting grid (corresponding to the loading point) and ending grid (corresponding to the unloading point) of the autonomous vehicle are clearly defined, using A... Algorithm for global shortest path 。 A The algorithm evaluates the priority of each walkable grid using a cost function. Specifically, A The algorithm is well-known in the field and will not be described in detail further. The initial global shortest path is obtained, whereby it is output as a sequence of path points, denoted as . ,in, The path points corresponding to the starting grid. The path point corresponding to the endpoint grid.

[0037] After completing the global shortest path and obtaining the path point sequence, the autonomous vehicle's control system analyzes the path point sequence and controls the autonomous vehicle's driving direction and speed according to the order of each path point in the sequence, so that the autonomous vehicle can initially drive along the planned global path.

[0038] During the initial driving process, the autonomous vehicle does not consider the influence of dynamic obstacles (such as other autonomous vehicles that are driving), and only maintains its driving direction based on the global path, laying the foundation for path adjustment in combination with dynamic risks in subsequent steps.

[0039] S2: Taking any operating automated loading and unloading vehicle as the target vehicle, and other operating automated loading and unloading vehicles other than the target vehicle as matching vehicles, based on the path point sequence, the DTW matching algorithm is used to perform interactive analysis on the path point sequence of the target vehicle and the path point sequence of the matching vehicle, quantify the risk characteristic value of each path point, calculate the cosine similarity between the gravitational direction of each path point and the current driving direction, and obtain the direction association weight.

[0040] The risk of waypoints stems from the path interaction between the target vehicle and other operating autonomous vehicles. It is mainly related to three key dimensions: the consistency of the two vehicles' directions of motion, the distance between waypoints, and the changing trend of the path distance. These three factors together determine the level of collision or congestion risk.

[0041] Obtain the global path point sequence of the target vehicle. At the same time, it extracts the planned path point sequences of all other operating unmanned vehicles, and only retains the current path point and subsequent path points for matching (excluding path segments that have been completed to ensure real-time evaluation).

[0042] The DTW (Dynamic Time Warping) matching algorithm is used to interactively match the path point sequence of the target vehicle with the path point sequence of each other autonomous vehicle, so as to accurately locate the corresponding matching point of each path point on the path of other vehicles.

[0043] Specifically, the total number of automated loading and unloading vehicles currently in operation in the transportation scenario is obtained, and any one of them is selected as the target vehicle. All other automated loading and unloading vehicles currently in operation are defined as matching vehicles.

[0044] Any path point in the initial global shortest path of the target vehicle is selected as the analysis path point, and any path point in the initial global shortest path of any matching vehicle is selected as the matching path point. Risk quantification is then carried out around this set of "analysis path points - matching path points".

[0045] The degree of conflict between the driving directions of the target vehicle and the matching vehicle is quantified by using cosine similarity: First, the cosine similarity between the driving direction of the target vehicle at the analysis path point and the driving direction of the matching vehicle at the matching path point is calculated, and its value range is determined by... Then, subtract the cosine similarity from 1 to obtain the directional difference weight. The closer the cosine similarity is to -1 (two vehicles traveling in opposite directions), the larger the weight, which represents a higher risk of collision at the directional level; the closer it is to 1 (two vehicles traveling in the same direction), the smaller the weight, which represents a lower risk of directional conflict.

[0046] Strengthening the proximity risk weight based on the spatial distance between the two vehicle path points: First, calculate the normalized Euclidean distance between the analysis path point of the target vehicle and the matching path point of the matching vehicle; then substitute the normalized Euclidean distance into a natural number... An exponential mapping is performed using a negative exponential function with a base of 0 to obtain distance-related weights. The closer the distance, the larger the mapping result and the higher the weight ratio, highlighting the risk of short-distance path interactions; the farther the distance, the smaller the mapping result, weakening invalid interference.

[0047] Analyze the approaching and diverging trends of the two vehicle paths: First, calculate the normalized Euclidean distance between the target vehicle's analyzed path point and the matching vehicle's matched path point, as well as the normalized Euclidean distance between the next consecutive path point of the target vehicle's analyzed path point and the next consecutive path point of the matching vehicle's matched path point. Then, divide the first set of distances by the second set of distances to obtain the weight of the path distance change trend. The larger the ratio, the more the path distance between the two vehicles is decreasing, and the higher the risk of subsequent path overlap and congestion. The smaller the ratio, the more the paths are diverging, and the lower the risk.

[0048] It should be noted that the Euclidean distance between all waypoints is calculated based on the distance of the gridded map. The normalized Euclidean distance is obtained as follows: First, obtain the pathpoint sequence of the target vehicle and the pathpoint sequences of all matching vehicles. Then, calculate the sum of the Euclidean distances between the pathpoints of the target vehicle and the pathpoints of all matching vehicles to obtain the total interaction distance, which serves as the basis for normalization. Finally, calculate the Euclidean distance between the analysis pathpoints of the target vehicle and the matching pathpoints of the matching vehicles. The ratio of the Euclidean distance to the total interaction distance is used as the normalized Euclidean distance, which is helpful in quantifying the interaction relationship between the pathpoints of the target vehicle and all matching vehicles.

[0049] Multiply the directional difference weight, distance correlation weight, and path distance change trend weight obtained above by one to obtain the comprehensive interaction risk value of the analysis path point-matching path point. The comprehensive interaction risk value fully reflects the interaction risk between the target vehicle and a single matched vehicle at that path point.

[0050] Iterate through all matched vehicles and repeat the above steps to calculate the comprehensive interaction risk value between the target vehicle's analysis path point and the corresponding matching path point of each matched vehicle. Sum all comprehensive interaction risk values ​​and then divide by the total number of matched vehicles (i.e., the number of operating unmanned vehicles other than the target vehicle) and take the average value to finally obtain the risk characteristic value of the target vehicle's analysis path point. Taking the average value can eliminate the impact of the difference in the number of matched vehicles on the risk assessment and ensure that the risk quantification results are objective and consistent.

[0051] Specifically, the risk eigenvalues ​​satisfy the following relationship:

[0052] ;

[0053] In the formula, Indicates the target vehicle's first Path points The risk characteristic value, Indicates the number of matched cars. Indicates the target vehicle's first Path points With the The direction of the matching path point of the matching vehicle and the first matching point of the target vehicle Path points The cosine similarity of the driving directions can reflect the consistency of the two vehicles' directions of motion, indicating a higher risk of collision when the two vehicles are traveling in opposite directions. Indicates the target vehicle's first Path points With the The normalized Euclidean distance between the matching path points of the matched vehicles. Indicates the target vehicle's first Path points With the The normalized Euclidean distance between the matching path points of the matched vehicles. Represented by natural numbers An exponential function with base 0. This indicates the total number of matched cars. This represents the local minimum constant, with a value of 0.01.

[0054] The degree of conflict between the two vehicles is quantified by using cosine similarity. The weight is increased when the vehicles are traveling in opposite directions, which can accurately capture the collision risk at the directional level. The normalized Euclidean distance between the two vehicle waypoints is nonlinearly weighted to enhance the risk proportion of close-range interactions and weaken the invalid interference at long distances. It reflects the changing trend of the path distance between two vehicles. The larger the ratio, the more likely the paths are to be close or even overlap, effectively identifying potential congestion risks. The summation operation summarizes the interaction risks between the current path point and all matched vehicles, avoiding the omission of safety hazards caused by individual vehicles. The average operation eliminates the impact of differences in the number of matched vehicles on risk quantification, ensuring the consistency of risk assessment under different vehicle density scenarios.

[0055] By integrating multi-dimensional weights and performing standardized calculations, the dynamic risks of path points are accurately and comprehensively quantified, providing a reliable quantitative basis for subsequent local target point selection and dynamic path adjustment. This effectively avoids collision and congestion risks and ensures the safety and efficiency of autonomous vehicle operation.

[0056] To further explain, after quantifying the risk characteristic values ​​of all path points in the global path of the target vehicle, it is necessary to dynamically set local target points based on the risk characteristic values ​​and calculate the suitability score of the path points as local target points.

[0057] Specifically, the fit score is calculated in the following ways:

[0058] Obtain the target vehicle's real-time driving direction, path points, and risk characteristic values ​​of the path points; substitute the risk characteristic values ​​into a negative exponential function to obtain the risk association weight; where, the smaller the risk association weight (lower risk), the closer the weight is to 1 (higher weight); the larger the risk association weight (higher risk), the closer the weight is to 0 (lower weight).

[0059] The similarity between the target vehicle's driving direction and the gravitational direction of the analysis path point is calculated using a cosine function to obtain the direction association weight. The smaller the direction deviation, the closer the weight is to 1; the larger the direction deviation, the closer the weight is to 0. The risk association weight is multiplied by the direction association weight to obtain the suitability score of the analysis path point, realizing the dual screening logic of "low risk and consistent direction".

[0060] The fit score satisfies the following relationship:

[0061] ;

[0062] In the formula, Indicates the target vehicle's first Path points As a fitness score for local target points Indicates the target vehicle's first Path points Based on the risk characteristic value, low-risk points should be prioritized when determining local target points. Represented by natural numbers An exponential function with base 0. This indicates the real-time driving direction of the target vehicle. Indicates the target vehicle's first Path points, This represents the function for calculating cosine similarity. Is the current driving direction and the first The cosine similarity between the gravitational directions of each path point as a local target point needs to be minimized to ensure that the path does not deviate from the global shortest path and cause a conflict during the journey.

[0063] Using path points as local target points must simultaneously meet the following core constraints to balance path stability and obstacle avoidance safety: Local target points must be selected from the path point sequence of the global path, and it is strictly forbidden to set target points deviating from the global path to avoid the autonomous vehicle deviating from the predetermined route and causing new path conflicts; The deviation between the gravitational direction of the local target point (i.e., the direction from the current position of the autonomous vehicle to the target point) and the current driving direction of the autonomous vehicle must be controlled within a preset range. By ensuring directional consistency, the frequency of path adjustment is reduced, ensuring a smooth driving process; Path points with lower risk characteristic values ​​are preferentially selected as local target points. By avoiding high-risk road sections, the probability of collisions or congestion with the matched vehicle is reduced, ensuring driving safety and efficiency.

[0064] S3: Based on the risk characteristic value and directional association weight of the waypoint, calculate the adaptability score of the waypoint as a local target point, form a score curve, and select the first inflection point closest to the current driving position as the local target point; based on the dynamically updated waypoint risk characteristic value and local target point, use the improved artificial potential field method to adjust the path in real time during driving.

[0065] When the target vehicle moves towards a local target point, its trajectory is determined by the combined gravitational force of the local target point and the repulsive force of surrounding obstacles. The calculation method for the combined resultant force vector includes:

[0066] The system obtains the current location of the autonomous vehicle, local target points, the average risk characteristic values ​​of all path points between the current location and local target points, and the real-time location and distance of all obstacles within the sensing range.

[0067] Calculate the path distance between the target vehicle's current position and the local target point. Use a negative exponential function to exponentially map the mean risk characteristic value of all path points between the target vehicle's current position and the local target point to obtain the gravity adjustment factor. Multiply the gravity gain coefficient, the path distance between the target vehicle's current position and the local target point, the gravity adjustment factor, and the unit direction vector from the target vehicle's current position to the local target point in sequence to obtain the gravity vector.

[0068] For each obstacle, a distance correlation factor is calculated using the correlation function between repulsive force and obstacle distance. The average risk characteristic value of all path points between the target vehicle's current position and the local target point is added to 1 and used as the repulsive force adjustment factor. The repulsive force gain coefficient, repulsive force adjustment factor, distance correlation factor, and unit direction vector from the obstacle to the target vehicle are multiplied sequentially to obtain the repulsive force vector of a single obstacle. The total repulsive force vector is obtained by summing the repulsive force vectors of all obstacles within the sensing range.

[0069] The gravitational vector is added to the total repulsive vector to obtain the combined resultant force vector, which serves as the real-time driving control command for the autonomous vehicle.

[0070] Specifically, the formula for calculating the combined resultant force vector is as follows:

[0071] ;

[0072] ;

[0073] ;

[0074] In the formula, This represents the combined resultant force vector of the target vehicle. This represents the gravitational vector of the local target point on the target vehicle. Indicates the first [unit] within the target vehicle's sensing range The repulsive force vector of an obstacle on the autonomous vehicle. This represents the gravitational gain coefficient, with a value of 0.5. This represents the path distance between the current position of the target vehicle and the local target point. The greater the distance, the greater the gravitational force, prompting the autonomous vehicle to follow the global shortest path. This represents the average risk characteristic value of all path points between the current position of the target vehicle and the local target point. When the risk of the path ahead of the vehicle is high, the gravity is appropriately reduced to focus more on obstacle avoidance and avoid collisions. This represents the unit direction vector pointing from the current position of the target vehicle to the local target point, used to define the direction of gravity; This represents the repulsion gain coefficient, with a value of 10. The function representing the relationship between repulsive force and distance to the obstacle is well-known to those skilled in the art and will not be described in detail here. This represents the unit direction vector from the obstacle to the target vehicle. Represented by natural numbers An exponential function with base 0.

[0075] By adjusting the attraction and repulsion forces in the artificial potential field method based on risk eigenvalues, while ensuring that the autonomous vehicle follows the global shortest path macroscopically, the sensitivity of the autonomous vehicle to obstacles is adjusted based on risk eigenvalues. This can minimize conflicts, ensure a smooth driving path, and improve driving efficiency.

[0076] S4: Iterates through the planning of the global shortest path, the calculation of risk characteristics, and the real-time control until the unmanned vehicle reaches the destination and completes the automatic loading and unloading task.

[0077] Following the method in step S1, the transportation area map is gridded and the global shortest path is planned to obtain a global path point sequence; step S2 is executed simultaneously to quantify the risk characteristic values ​​of each path point on the current road segment to be traveled and to select the first local target point.

[0078] Throughout the entire process of the target vehicle traveling from its current position to the first local target point, the improved artificial potential field method in step S3, combined with obstacle information and path point risk characteristic values ​​perceived by real-time sensors, dynamically adjusts the intensity of attraction and repulsion to achieve real-time optimization of the driving path and avoid dynamic risks such as collisions and congestion.

[0079] When the target vehicle reaches the current local target point, an iterative update mechanism is triggered: the risk characteristic values ​​of all path points in the remaining path segment are requantified (to adapt to real-time situations such as changes in the positions of other unmanned vehicles and dynamic adjustments to the environment); based on the updated risk characteristic values, a new local target point is determined according to the screening rules in step S2; the process of "local path control - reaching the target point - updating risk values ​​and local target points" is repeated to form continuous iteration. When the unmanned vehicle reaches the end point of the global path (i.e., the target location of the loading and unloading task) through the above iterative process, the dynamic iteration process terminates, and the unmanned vehicle completes the automatic loading and unloading task.

[0080] By constructing a dynamic iterative closed loop of "global shortest path - path point risk assessment - local path real-time control", the unmanned vehicle can continuously adapt and adjust in a dynamic environment: with the global path as the macro guide, the risk characteristic value as the quantitative basis, and the improved artificial potential field method as the real-time control means, by cyclically updating the risk assessment results and local target points, the unmanned vehicle can always move closer to the global endpoint while avoiding dynamic conflicts, and finally complete the loading and unloading task efficiently.

[0081] This invention also provides an automated loading and unloading system for unmanned vehicles. For example... Figure 2 As shown, the system includes a processor and a memory. The memory stores computer program instructions, which, when executed by the processor, implement an automated loading and unloading method for unmanned vehicles according to the first aspect of the present invention. The system also includes other components well-known to those skilled in the art, such as a communication bus and a communication interface. Their configurations and functions are known in the art and will not be described further here.

[0082] It should be noted that those skilled in the art can make various modifications and improvements without departing from the inventive concept, and these all fall within the scope of protection of this invention. Therefore, the scope of protection of this patent should be determined by the appended claims.

Claims

1. An automated loading and unloading unmanned vehicle loading and unloading method characterized by, The application relates to an automatic loading and unloading unmanned vehicle and a global shortest path planning method thereof. According to the loading and unloading tasks of the automatic loading and unloading unmanned vehicle and a running area map, an initial global shortest path and a corresponding path point sequence are obtained; Any running automatic loading and unloading unmanned vehicle is taken as a target vehicle, and the running automatic loading and unloading unmanned vehicles except the target vehicle are taken as matching vehicles; based on the path point sequence, the path point sequence of the target vehicle and the path point sequence of the matching vehicles are interactively analyzed by using a DTW matching algorithm, the risk characteristic values of the path points are quantified, the cosine similarity of the direction of the gravitational force of each path point and the current driving direction is calculated, and the direction correlation weight is obtained; According to the risk characteristic values of the path points and the direction correlation weight, the adaptability score of the path points as local target points is calculated, a score curve is formed, the first inflection point closest to the current driving position is selected as the local target point, the local target point is dynamically set based on the risk characteristic values, and the local target point is selected from the path point sequence of the global path; based on the dynamically updated path point risk characteristic values and the local target point, the improved artificial potential field method is used to adjust the path in real time during driving; The dynamic iteration process of the global shortest path planning, the risk characteristic value calculation and the real-time control is cyclically executed until the unmanned vehicle reaches the terminal point and the automatic loading and unloading task is completed.

2. The automatic loading and unloading unmanned vehicle loading and unloading method according to claim 1, characterized in that, The initial global shortest path and the corresponding path point sequence are obtained by: The running area map is gridded and obstacles are marked, and A The algorithm plans a global shortest path based on a cost function, extracts the coordinates of all grid nodes on the global shortest path in the order of the unmanned vehicle driving, forms an ordered path point set, outputs the ordered path point set in sequence, and obtains a path point sequence.

3. The automated loading and unloading method of claim 1, wherein, The risk characteristic value calculation method comprises: The number of the running automatic loading and unloading unmanned vehicles is obtained, any path point in the initial global shortest path of the target vehicle is taken as an analysis path point, and any path point in the initial global shortest path of any matching vehicle is taken as a matching path point; The cosine similarity between the driving direction of the analysis path point of the target vehicle and the driving direction of the matching path point of the matching vehicle is calculated, and the direction difference weight is obtained by subtracting the cosine similarity from 1; The normalized Euclidean distance between the analysis path point of the target vehicle and the matching path point of the matching vehicle is calculated, the normalized Euclidean distance is exponentially mapped by using a negative exponential function, and the distance correlation weight is obtained; The normalized Euclidean distance between the analysis path point of the target vehicle and the matching path point of the matching vehicle and the normalized Euclidean distance between the next path point of the analysis path point of the target vehicle and the next path point of the matching path point of the matching vehicle are calculated, and the ratio is calculated to obtain the path distance change trend weight; The direction difference weight, the distance correlation weight and the path distance change trend weight are multiplied respectively to obtain the comprehensive interaction risk value of the analysis path point and the matching path point, the comprehensive interaction risk values of all the matching vehicles except the target vehicle are traversed and summed to obtain the average value, and the risk characteristic value of the analysis path point of the target vehicle is obtained.

4. The automated loading and unloading method of claim 1, wherein, The adaptability score of the path points as the local target points is calculated by: The risk correlation weight is obtained by taking any path point in the initial global shortest path of the target vehicle as an analysis path point, obtaining the real-time driving direction of the target vehicle, the analysis path point, and a risk characteristic value of the analysis path point, and substituting the risk characteristic value into a negative exponential function; the direction correlation weight is obtained by calculating the similarity between the driving direction of the target vehicle and the direction of the gravity of the analysis path point; and the adaptability score of the analysis path point as a local target point is obtained by multiplying the risk correlation weight and the direction correlation weight.

5. The automated loading and unloading method of claim 1, wherein, The improved artificial potential field method is used to adjust the path in real time during driving, and includes: The path distance between the current position of the target vehicle and the local target point is calculated, the mean value of the risk characteristic values of all path points between the current position of the target vehicle and the local target point is exponentially mapped using a negative exponential function to obtain a gravity adjustment factor, and the gravity gain coefficient, the path distance between the current position of the target vehicle and the local target point, the gravity adjustment factor, and a unit directional vector from the current position of the target vehicle to the local target point are sequentially multiplied to obtain a gravity vector; For each obstacle, a distance correlation factor is calculated by an association function of repulsion and obstacle distance; 1 is added to the mean value of the risk characteristic values of all path points between the current position of the target vehicle and the local target point to obtain a repulsion adjustment factor; and the repulsion gain coefficient, the repulsion adjustment factor, the distance correlation factor, and a unit directional vector from the obstacle to the target vehicle are sequentially multiplied to obtain a repulsion vector of a single obstacle; The repulsion vectors of all obstacles in the sensing range are summed to obtain a total repulsion vector, the gravity vector and the total repulsion vector are added to obtain a comprehensive resultant vector, and the real-time driving direction and the adjustment intensity of the target vehicle are determined based on the comprehensive resultant vector.

6. The automated loading and unloading method of claim 1, wherein, When the DTW matching algorithm is executed, only the path point and the subsequent path point sequence of each automatic loading and unloading unmanned vehicle at the current time are intercepted for matching, and the path segment that has been driven is excluded to ensure the real-time and accuracy of risk assessment.

7. The automated loading and unloading method of claim 1, wherein, The update frequency of the local target point and the update frequency of the path point risk characteristic value are consistent; after the automatic loading and unloading unmanned vehicle drives to the current local target point, the risk characteristic values of the path points in the subsequent remaining path segment are immediately recalculated, and the next local target point is selected based on the updated risk characteristic values, and the cycle is repeated until the global end point is reached.

8. An automated loading and unloading unmanned vehicle loading and unloading system characterized by, The method comprises the following steps: A processor and a memory, the memory stores computer program instructions, when the computer program instructions are executed by the processor, the automatic loading and unloading unmanned vehicle loading and unloading method according to any one of claims 1-7 is realized.

Images