Double-vehicle cooperative obstacle avoidance and path planning method based on adaptive weight and multi-sensor fusion

Abstract

The invention discloses a double-vehicle cooperative obstacle avoidance and path planning method based on self-adaptive weight and multi-sensor fusion, and relates to the field of path planning obstacle avoidance, and the method comprises the steps: building a map containing obstacles and ground friction force marks; sorting the transportation tasks based on a genetic algorithm considering a dependency relationship to obtain an optimal transportation sequence; performing risk judgment on current road safety, and calculating a road safety evaluation coefficient; calculating an optimal path of the vehicle based on a path planning algorithm considering adaptive weight; executing the current transportation task according to the optimal transportation sequence and the optimal path; an affected obstacle on the path is determined, and self-adaptive obstacle avoidance is carried out according to a following obstacle avoidance and self-adaptive vehicle guiding strategy of a rear vehicle; calculating a safe spacing threshold value; and controlling the speed of the rear vehicle according to the comparison result of the effective distance between the vehicles and the safe distance threshold. According to the invention, task allocation and path planning intelligence is ensured, and the efficiency and reliability of the double-vehicle control system are improved.

Description

Technical Field

[0001] The present invention relates to the field of path planning and obstacle avoidance, and more specifically, to a dual-vehicle collaborative obstacle avoidance and path planning method based on adaptive weights and multi-sensor fusion. Background Art

[0002] With the rapid development of intelligent manufacturing and logistics automation technologies, the application of Automated Guided Vehicle (AGV) systems in scenarios such as warehousing logistics and flexible production lines has become increasingly widespread. Traditional AGV systems mostly adopt a single-vehicle independent operation mode, and their task allocation and path planning algorithms are usually based on fixed priorities or static environment assumptions, making it difficult to meet the requirements of dynamic, multi-task, and high-complexity collaborative operations. Especially in the dual-vehicle collaborative scenario, problems such as the path coupling between the front and rear vehicles, dynamic obstacle avoidance, and safe following pose higher requirements for the real-time performance, robustness, and intelligence of the system.

[0003] In the prior art, the task allocation of dual-vehicle systems mostly uses a simple round-robin scheduling algorithm, lacking global optimization of the dependencies between tasks (such as the loading and unloading order), which easily leads to resource waste and logical conflicts. In terms of path planning, although traditional A* algorithms or Dynamic Window Approach (DWA) can achieve basic obstacle avoidance, they do not fully integrate the impact of road surface physical characteristics (such as friction, slope, water accumulation) on energy consumption and safety, resulting in large deviations in path cost estimation. In addition, existing dual-vehicle following control mostly adopts a fixed safety distance strategy, without considering the influence of load changes, road surface friction coefficient fluctuations, and dynamic obstacle interference, which is prone to collision risks under emergency braking or complex road conditions.

[0004] In terms of energy management, traditional AGV systems usually trigger charging based on the remaining battery capacity threshold, lacking the joint prediction of task execution duration and path energy consumption, which easily leads to task interruption due to battery charge estimation deviation. At the same time, existing battery health monitoring mechanisms mostly rely on simple threshold judgments of voltage or current, making it difficult to accurately evaluate the non-linear impact of battery aging on the endurance ability.

[0005] In response to the above problems, although existing research has tried to introduce optimization methods such as genetic algorithms and Model Predictive Control (MPC), there are still defects such as high computational complexity and insufficient real-time performance in multi-objective collaborative optimization (such as task priority, path cost, dynamic obstacle avoidance). Especially in the dual-vehicle collaborative scenario, a systematic solution for path replanning and role switching mechanisms under the coupling of the front and rear vehicle states has not been formed, resulting in limited reliability and adaptability of the system in complex environments.

[0006] Effective solutions have not been proposed for the problems in the related art. Summary of the Invention

[0007] In view of the problems in the related art, the present invention proposes a dual-vehicle collaborative obstacle avoidance and path planning method based on adaptive weights and multi-sensor fusion to overcome the above-mentioned technical problems existing in the existing related art.

[0008] To this end, the specific technical solution adopted by the present invention is as follows:

[0009] The dual-vehicle collaborative obstacle avoidance and path planning method based on adaptive weights and multi-sensor fusion includes:

[0010] Establish a map including obstacles and ground friction marks; sort the transportation tasks based on a genetic algorithm considering dependencies to obtain the optimal transportation order;

[0011] Use the method of multi-sensor fusion to judge the risk of the current road safety and calculate the road safety evaluation coefficient; calculate the optimal path of the vehicle based on a path planning algorithm considering adaptive weights;

[0012] Execute the current transportation task according to the optimal transportation order and the optimal path; determine the affected obstacles on the path according to the expansion radius of the vehicle and the obstacle, and perform adaptive obstacle avoidance according to the following vehicle following obstacle avoidance and adaptive leading vehicle strategy;

[0013] Calculate the safety distance threshold based on the vehicle's dynamic reaction distance, the vehicle's adaptive braking distance, and the road safety evaluation coefficient; control the speed of the following vehicle according to the comparison result between the effective distance between vehicles and the safety distance threshold.

[0014] Further, sorting the transportation tasks based on a genetic algorithm considering dependencies to obtain the optimal transportation order includes:

[0015] Construct a fitness function based on the Euclidean distance between the current point and the target point, the task priority, the task duration, and the average speed recorded by the leading vehicle, and ensure the correct order between tasks;

[0016] Based on the genetic algorithm, and through operations including natural selection, crossover, and mutation, screen and sort the transportation tasks to be executed based on the calculation result of the fitness function to obtain the optimal transportation order;

[0017] The fitness function is:

[0018] Fitness = Q(K1)*(d + d1)+Q(K2)*M(d)*P+Q(K3)*t*v;

[0019] In the formula, d is the Euclidean distance between the current point and the target point, and d1 is the Euclidean distance between the unloading point and the loading point;

[0020] Let \(M(d)\) be the amplification function, \(P\) be the task priority, \(t\) be the task duration, and \(v\) be the average speed recorded by the guiding vehicle;

[0021] Let \(Q(K1)\) be the weight value corresponding to the distance, \(Q(K2)\) be the weight value corresponding to the task priority, and \(Q(K3)\) be the weight value corresponding to the average speed.

[0022] Furthermore, using the multi-sensor fusion method, a risk judgment is made on the current road safety, and the calculated road safety evaluation coefficient includes:

[0023] Obtain multi-environment data on the road, where the multi-environment data includes the water accumulation depth, air humidity, environmental temperature, and rainfall;

[0024] Based on the multi-environment data, determine whether there is a risk of affecting vehicle functions or a risk of road icing. If so, set an obstacle impassable mark at the corresponding position on the map. If not, determine whether there is water accumulation or rainfall without the risk of icing. If so, reduce the weight of the corresponding route, otherwise do not adjust;

[0025] Based on the water accumulation depth and air humidity, calculate the road safety evaluation coefficient, and the calculation formula for the road safety evaluation coefficient \(w\) is:

[0026]

[0027] In the formula, \(H'\) is the normalized air humidity, and \(D'\) is the normalized water accumulation depth;

[0028] Let \(Q(K4)\) be the weight value corresponding to the air humidity, and \(Q(k5)\) be the weight value corresponding to the water accumulation depth.

[0029] Furthermore, based on the path planning algorithm considering adaptive weights, calculating the optimal path of the vehicle includes:

[0030] Based on the vehicle position, the theoretical optimal friction value between the vehicle tires and the ground, and the ground friction value of the current road section, construct the actual cost function;

[0031] Based on the vehicle position, the ground friction value of the current road section, the current road slope, the value of the current grid point, and the value of the end grid point, construct the heuristic cost function;

[0032] Based on the current task vehicle load, the average speed of the current task vehicle, and the road safety evaluation coefficient of the current road section, construct the vehicle condition weight function;

[0033] Using the actual cost function, the heuristic cost function, and the vehicle condition weight function, construct the objective function of the path planning algorithm, and the objective function is:

[0034] f(n) = α2 * g(n) + (1 - α2) * h(n) + γ2 * K(n);

[0035] Wherein, g(n) is the actual cost function, h(n) is the heuristic cost function, and K(n) is the vehicle condition weight function;

[0036] α2 is the dynamic weight function, and γ2 is the coefficient of the vehicle condition weight function;

[0037] Construct a path planning algorithm, and obtain the optimal path of the vehicle from the current point to the target point according to the calculated value of the objective function.

[0038] Furthermore, before executing the current transportation task according to the optimal transportation order and the optimal path, it also includes:

[0039] Detect the remaining power of the vehicle battery. When the remaining power is not enough to support the subsequent journey, select the nearest charging pile location and evaluate the required charging amount, and re - sort the remaining task order at the charging location;

[0040] When the charging amount of the vehicle battery is greater than the preset value each time, evaluate the vehicle battery efficiency, and when the vehicle battery efficiency is lower than the alarm threshold, alarm to the host computer.

[0041] Furthermore, according to the inflation radii of the vehicle and the obstacles, the affected obstacles on the path are determined as follows:

[0042] After adding the preset inflation radius of the vehicle itself and the preset inflation radius of the obstacle, compare the result with the Euclidean distance between the vehicle and the obstacle to obtain the current passable distance;

[0043] Determine the affected obstacles on the path according to the size of the current passable distance.

[0044] Furthermore, according to the size of the current passable distance, the affected obstacles on the path are determined as follows:

[0045] If the current passable distance is less than the predetermined value, the corresponding obstacle is the affected obstacle on the path;

[0046] If the current passable distance is greater than or equal to the predetermined value, the corresponding obstacle does not affect passage, and after the vehicle bypasses the obstacle, it proceeds along the original path.

[0047] Furthermore, the adaptive obstacle avoidance according to the following - vehicle following obstacle avoidance and adaptive leading vehicle strategy includes:

[0048] In the case where the following vehicle following the leading vehicle encounters an obstacle and cannot pass, if it is on a one - way road, the following vehicle adaptively becomes a leading vehicle, returns to the nearest passed marker point, and then performs path planning to the end point based on the path planning algorithm considering the adaptive weight;

[0049] If it is located at an intersection, determine whether the leading vehicle has passed through the intersection. If it has not passed through the intersection, the leading vehicle remains unchanged and returns to the previous marked point to re-plan the path;

[0050] If it has passed through the intersection, determine the distance between the leading vehicle and the center of the intersection and the distance between other vehicles and the center of the intersection. If the distance between the leading vehicle and the center of the intersection is less than the distance between other vehicles and the center of the intersection, the leading vehicle remains unchanged and returns to the previous marked point to re-plan the path;

[0051] If the distance between the leading vehicle and the center of the intersection is greater than the distance between other vehicles and the center of the intersection, other vehicles are adapted to be the leading vehicle and return to the previous marked point to re-plan the path.

[0052] Furthermore, based on the vehicle's dynamic response distance, vehicle adaptive braking distance, and road safety evaluation coefficient, calculating the safety distance threshold includes:

[0053] Obtain the effective distance between adjacent vehicles, and calculate the dynamic vehicle condition weight factor according to the current road safety evaluation coefficient. The calculation formula for the dynamic vehicle condition weight factor is:

[0054]

[0055] According to the dynamic vehicle condition weight factor, and combining the vehicle's dynamic response distance and vehicle adaptive braking distance, calculate the safety distance threshold at the current speed. The calculation formula for the safety distance threshold at the current speed is:

[0056]

[0057] In the formula, v is the speed of the current leading vehicle in front, m is the total mass of the current vehicle and its load, m0 is the standard unloaded mass of the current vehicle, μ is the ground friction coefficient reference value of the current section, g is the gravitational acceleration value, τ is the sensor delay time of the following vehicle, and α3 is the default value of the mass decay index;

[0058] w is the current road safety evaluation coefficient, γ is the mass response coefficient, β3 is the road condition mutation factor, and λ is the risk perception coefficient.

[0059] Furthermore, according to the comparison result between the effective distance between vehicles and the safety distance threshold, controlling the speed of the following vehicle includes:

[0060] If the effective distance between vehicles is greater than the safety distance threshold, the following vehicle adjusts to the same speed as the vehicle in front in a way of first accelerating and then decelerating on the premise of being less than the maximum speed of the preset vehicle;

[0061] If the effective distance between vehicles is less than the safety distance threshold, the following vehicle adjusts its speed to be the same as that of the preceding vehicle in a way of first decelerating and then accelerating on the premise of being less than the maximum speed of the preset vehicle.

[0062] The beneficial effects of the present invention are as follows:

[0063] The present invention provides a dual-vehicle collaborative obstacle avoidance and path planning method based on adaptive weight and multi-sensor fusion. By fusing multi-sensor data to construct a high-precision environment map, combining adaptive weight reinforcement learning to achieve real-time avoidance of dynamic obstacles, using the fusion technology of global and local path planning to significantly improve the adaptability to complex scenarios, and balancing the driving efficiency and energy consumption through a multi-objective optimization decision model. The environment perception system based on the fusion of lidar and vision effectively reduces the misjudgment rate, ensures the intelligence of task allocation and path planning, and finally forms a dual-vehicle collaborative operation system with high robustness, high efficiency and high reliability, providing an expandable solution for the intelligent logistics system. Description of the Drawings

[0064] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required to be used in the embodiments. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0065] Figure 1 is a flowchart of the dual-vehicle collaborative obstacle avoidance and path planning method based on adaptive weight and multi-sensor fusion according to an embodiment of the present invention;

[0066] Figure 2 is a calculation example diagram for determining whether an obstacle affects passage by expanding the distance according to an embodiment of the present invention;

[0067] Figure 3 is an example diagram of a single-lane road situation according to an embodiment of the present invention;

[0068] Figure 4 is an example diagram of an intersection situation according to an embodiment of the present invention;

[0069] Figure 5 is one of the flowcharts in the actual process according to an embodiment of the present invention;

[0070] Figure 6 is another flowchart in the actual process according to an embodiment of the present invention;

[0071] Figure 7 is the third flowchart in the actual process according to an embodiment of the present invention;

[0072] Figure 8 It is the fourth flow chart in the actual process according to an embodiment of the present invention;

[0073] Figure 9 It is the fifth flow chart in the actual process according to an embodiment of the present invention. Detailed implementation manners

[0074] To further illustrate each embodiment, the present invention provides accompanying drawings, which are part of the disclosure of the present invention. They are mainly used to illustrate the embodiments and can be combined with the relevant descriptions in the specification to explain the operating principles of the embodiments. With reference to these contents, those of ordinary skill in the art should be able to understand other possible implementation manners and the advantages of the present invention. The components in the figures are not drawn to scale, and similar component symbols are usually used to represent similar components.

[0075] According to an embodiment of the present invention, a dual-vehicle collaborative obstacle avoidance and path planning method based on adaptive weight and multi-sensor fusion is provided.

[0076] Now, the present invention will be further described in combination with the accompanying drawings and specific implementation manners. As Figure 1 shown, the dual-vehicle collaborative obstacle avoidance and path planning method based on adaptive weight and multi-sensor fusion according to an embodiment of the present invention includes:

[0077] Establish a map including obstacles and ground friction marks; sort the transportation tasks based on a genetic algorithm considering dependencies to obtain the optimal transportation order.

[0078] Use the multi-sensor fusion method to judge the risk of the current road safety and calculate the road safety evaluation coefficient; calculate the optimal path of the vehicle based on a path planning algorithm considering adaptive weights.

[0079] Execute the current transportation task according to the optimal transportation order and the optimal path; determine the affected obstacles on the path according to the expansion radii of the vehicle and the obstacles, and perform adaptive obstacle avoidance according to the following vehicle following obstacle avoidance and adaptive leading vehicle strategy.

[0080] Calculate the safety distance threshold based on the vehicle's dynamic reaction distance, the vehicle's adaptive braking distance, and the road safety evaluation coefficient; control the speed of the following vehicle according to the comparison result between the effective distance between vehicles and the safety distance threshold.

[0081] In one embodiment, sorting the transportation tasks based on a genetic algorithm considering dependencies to obtain the optimal transportation order includes:

[0082] Based on the Euclidean distance between the current point and the target point, the task priority, the task duration, and the average speed recorded by the guiding vehicle, while ensuring the correct order of tasks, a fitness function is constructed; based on the genetic algorithm, through operations including natural selection, crossover, and mutation, and taking the calculation result of the fitness function as the basis, the transportation tasks to be executed are screened and sorted to obtain the best transportation order;

[0083] The fitness function is as follows:

[0084] Fitness = Q(K1)*(d + d1)+Q(K2)*M(d)*P+Q(K3)*t*v;

[0085] In the formula, d is the Euclidean distance between the current point and the target point, d1 is the Euclidean distance between the unloading point and the loading point; M(d) is the amplification function, P is the task priority, t is the task duration, v is the average speed recorded by the guiding vehicle; Q(K1) is the weight value corresponding to the distance, Q(K2) is the weight value corresponding to the task priority, Q(K3) is the weight value corresponding to the average speed.

[0086] In one embodiment, by using the method of multi-sensor fusion, a risk judgment is made on the current road safety, and the calculated road safety evaluation coefficient includes:

[0087] Obtain multi-environment data in the road, and the multi-environment data includes water accumulation depth, air humidity, environmental temperature, and rainfall; according to the multi-environment data, judge whether there is a risk of affecting vehicle functions or a risk of road icing. If so, set an obstacle impassable mark at the corresponding position on the map. If not, judge whether there is water accumulation or rainfall without the risk of icing. If so, reduce the weight of the corresponding route, otherwise do not adjust; according to the water accumulation depth and air humidity, calculate the road safety evaluation coefficient, and the calculation formula of the road safety evaluation coefficient w is:

[0088]

[0089] In the formula, H′ is the normalized air humidity, D′ is the normalized water accumulation depth; Q(K4) is the weight value corresponding to the air humidity, Q(k5) is the weight value corresponding to the water accumulation depth.

[0090] In one embodiment, based on the path planning algorithm considering adaptive weights, calculating the best path of the vehicle includes:

[0091] Construct an actual cost function based on the vehicle position, the theoretical optimal friction value between the vehicle tires and the ground, and the ground friction value of the current road section; construct a heuristic cost function based on the vehicle position, the ground friction value of the current road section, the current road slope, the value of the current grid point, and the value of the end grid point; construct a vehicle condition weight function based on the current load of the vehicle for the current task, the average speed of the vehicle for the current task, and the road safety evaluation coefficient of the current road section; use the actual cost function, the heuristic cost function, and the vehicle condition weight function to construct the objective function of the path planning algorithm, and the objective function is:

[0092] f(n) = α2 * g(n) + (1 - α2) * h(n) + γ2 * K(n);

[0093] In the formula, g(n) is the actual cost function, h(n) is the heuristic cost function, K(n) is the vehicle condition weight function; α2 is the dynamic weight function, and γ2 is the coefficient of the vehicle condition weight function.

[0094] Construct a path planning algorithm, and obtain the optimal path of the vehicle from the current point to the target point according to the calculated value of the objective function.

[0095] In one embodiment, before performing the current transportation task according to the optimal transportation order and the optimal path, it further includes:

[0096] Detect the remaining power of the vehicle battery. When the remaining power is not enough to support the subsequent journey, select the position of the nearest charging pile and evaluate the required charging amount, and re - sort the remaining task order at the charging position; when the charging amount of the vehicle battery is greater than the preset value each time, evaluate the vehicle battery efficiency, and when the vehicle battery efficiency is lower than the alarm threshold, alarm to the host computer.

[0097] In one embodiment, according to the expansion radii of the vehicle and the obstacles, the affected obstacles on the path are determined as follows:

[0098] After adding the expansion radius preset for the vehicle itself and the expansion radius preset for the obstacle, compare the result with the Euclidean distance between the vehicle and the obstacle to obtain the current passable distance; determine the affected obstacles on the path according to the size of the current passable distance.

[0099] In one embodiment, according to the size of the current passable distance, the affected obstacles on the path are determined as follows:

[0100] If the current passable distance is less than a predetermined value, the corresponding obstacle is the affected obstacle on the path; if the current passable distance is greater than or equal to the predetermined value, the corresponding obstacle does not affect passage, and after the vehicle bypasses the obstacle, it proceeds along the original path.

[0101] In one embodiment, adaptive obstacle avoidance according to the following-vehicle following obstacle avoidance and adaptive guiding vehicle strategy includes:

[0102] When the following vehicle follows the leading vehicle and encounters an obstacle and cannot pass, if it is on a one-way road, the following vehicle adaptively becomes a guiding vehicle. After returning to the nearest passed marker point, path planning is performed based on a path planning algorithm considering adaptive weights to reach the end point. If it is at an intersection, it is judged whether the guiding vehicle passes through the intersection. If it does not pass through the intersection, the guiding vehicle remains unchanged and returns to the previous marker point to re-perform path planning. If it passes through the intersection, the distances between the guiding vehicle and the center of the intersection and between other vehicles and the center of the intersection are judged. If the distance between the guiding vehicle and the center of the intersection is less than the distance between other vehicles and the center of the intersection, the guiding vehicle remains unchanged and returns to the previous marker point to re-perform path planning. If the distance between the guiding vehicle and the center of the intersection is greater than the distance between other vehicles and the center of the intersection, the other vehicle adaptively becomes a guiding vehicle and returns to the previous marker point to re-perform path planning.

[0103] In one embodiment, calculating the safety spacing threshold based on the vehicle's dynamic reaction distance, vehicle adaptive braking distance, and road safety evaluation coefficient includes:

[0104] Obtain the effective distance between adjacent vehicles, and calculate the dynamic vehicle condition weight factor according to the current road safety evaluation coefficient. The calculation formula for the dynamic vehicle condition weight factor is:

[0105]

[0106] According to the dynamic vehicle condition weight factor, and in combination with the vehicle's dynamic reaction distance and vehicle adaptive braking distance, calculate the safety spacing threshold at the current speed. The calculation formula for the safety spacing threshold at the current speed is:

[0107]

[0108] In the formula, v is the speed of the current leading vehicle, m is the total mass of the current vehicle and its load, m0 is the standard unloaded mass of the current vehicle, μ is the reference value of the ground friction coefficient of the current section, g is the acceleration due to gravity value, τ is the sensor delay time of the following vehicle, α3 is the default value of the mass decay exponent; w is the current road safety evaluation coefficient, γ is the mass response coefficient, β3 is the road condition mutation factor, and λ is the risk perception coefficient.

[0109] In one embodiment, controlling the speed of the following vehicle according to the comparison result between the effective distance between vehicles and the safety spacing threshold includes:

[0110] If the effective distance between vehicles is greater than the safety distance threshold, the following vehicle adjusts its speed to match that of the preceding vehicle in a way of first accelerating and then decelerating, provided that the speed is less than the maximum speed of the vehicle set in advance; if the effective distance between vehicles is less than the safety distance threshold, the following vehicle adjusts its speed to match that of the preceding vehicle in a way of first decelerating and then accelerating, provided that the speed is less than the maximum speed of the vehicle set in advance.

[0111] To facilitate the understanding of the above technical solutions of the present invention, the working principle of the present invention in the actual process will be described in detail below. As Figures 5 - 9 shown. Figure 5 In, line segment ④ and line segment ⑤ are respectively connected to Figure 6 in line segment ④ and line segment ⑤. Figure 6 In, line segment ⑥ is connected to Figure 7 in line segment ⑥. Figure 7 In, line segment ⑦ is connected to Figure 8 in line segment ⑦. Figure 8 In, line segment ①, line segment ② and line segment ③ are respectively connected to Figure 9 in line segment ①, line segment ② and line segment ③.

[0112] The present invention includes a dual-vehicle collaborative system integrating dynamic task scheduling, multi-factor path optimization, adaptive following control and intelligent energy management to optimize the operation efficiency, safety and system robustness in complex industrial scenarios.

[0113] The present invention provides a method for dual-vehicle collaborative obstacle avoidance and path planning based on adaptive weight and multi-sensor fusion. This process ensures the intelligence of task allocation and path planning, and improves the efficiency and reliability of the dual-vehicle control system, including the following steps:

[0114] 1. Detect the surrounding environment information, and establish a map containing ground friction marks using the grid method, including shelf grids, obstacle grids and path grids, and mark the corresponding ground friction in the path.

[0115] 2. Receive task requests, and each task includes a starting point and an ending point. The platform first uses a hierarchical optimization strategy to screen and sort all tasks using a genetic algorithm with added dependency relationships to ensure the correct order of precedence between tasks and sort the optimal execution order. Then, the system checks whether the vehicle has sufficient power to complete the task through the battery management system. If the power is insufficient, charging is arranged until the task requirements are met. After confirming that the vehicle is ready, the corresponding algorithm is used to plan the best path from the current position to the task starting point and then to the ending point, and the real-time obstacle avoidance module is used to handle the obstacles encountered during operation, and a time window is added.

[0116] III. Hierarchical optimization strategy, which includes using a genetic algorithm to preliminarily screen and sort all tasks to be executed, incorporating dependency relationships, and finding a relatively optimal task sequence through operations such as natural selection, crossover, and mutation. The fitness function comprehensively considers factors such as task priority, the starting and ending points of AGVs, the distances between unloading points and loading points, and task duration.

[0117] Specifically, it includes:

[0118] Incorporate dependency relationship management: Ensure the correct sequence of tasks. For example, first arrive at the loading point and then go to the unloading point to avoid logical errors or resource waste.

[0119] Task priority P: Classify tasks into different levels according to importance (such as levels 1, 2, and 3 with decreasing importance in sequence) to ensure that high-priority tasks are processed first.

[0120] Task duration t: Prioritize tasks with shorter durations to reduce overall waiting time and improve efficiency.

[0121] v is the average speed recorded by the guiding vehicle, and d1 is the Euclidean distance between the unloading point and the loading point.

[0122] Among them, the calculation formula of the fitness function used is:

[0123] The calculation amplification function M(d) is:

[0124]

[0125] In the formula, d is the Euclidean distance between the current point and the target point.

[0126] The corresponding weight function Q(K) is:

[0127] Q(K) = 1 - 0.1 * (K - 1) when K < 5;

[0128] K is the importance value of each module (less than 5).

[0129] The fitness function is:

[0130] Fitness = Q(K1) * (d + d1) + Q(K2) * M(d) * P + Q(K3) * t * v;

[0131] Under the default setting, the distance has the greatest impact on the fitness function. Therefore, the importance of K1 is default set to 1, the task priority is the second, the importance of K2 is default set to 2, and the average speed has the relatively smallest impact. Therefore, the importance of K3 is default set to 3. It can also be modified according to actual needs to achieve the optimal state. The weight function Q(K) is used to set the corresponding parameters. The importance levels are 1, 2, and 3, and the weight function values corresponding to different importance levels are different (the maximum importance does not exceed 5).

[0132] Using a genetic algorithm, starting from the current point, continuously perform natural selection, crossover, and mutation to select the next best target point, sort the tasks, and finally form the optimal transportation sequence.

[0133] IV. Analysis of environmental multi-sensor fusion algorithm. In areas prone to water accumulation (low-lying sections or uphill / downhill sections), use buried water accumulation monitors to detect the water accumulation depth, use temperature and humidity sensors to detect the environmental humidity, and use rain sensors to detect whether it is raining, and transmit information to the upper computer in real time. Considering possible adverse situations, classify the road condition coefficients according to the risk of water accumulation or road icing.

[0134] There is water accumulation or rainfall:

[0135] Temperature > 5 degrees, no icing risk; water accumulation or rainfall < 15 cm, reduce the weight of this route to prevent the vehicle from passing through here.

[0136] First, normalize each parameter:

[0137]

[0138] In the formula, H is the humidity of the current section, the range of air humidity is 0% to 90%, D is the current water accumulation depth, and the range is 0 to 15 cm.

[0139] At this time, the road safety evaluation coefficient w is:

[0140]

[0141] In the formula, H′ is the normalized air humidity, D′ is the normalized water accumulation depth. Here, the importance of air humidity is default set to 3, and the importance of water accumulation depth is default set to 1. The importance can also be modified according to actual needs to achieve the optimal effect. Similarly, define the weight adjustment function through the weight function Q(K).

[0142] Water accumulation or rainfall > 15 cm (extreme situation), which may affect the vehicle's function, is set as an impassable obstacle.

[0143] Temperature < 5 degrees, there is a risk of road icing; set the grid here as an impassable obstacle.

[0144] If there is no water accumulation or rainfall, there is no impact and it does not affect the vehicle condition weight function.

[0145] Temperature < 5 degrees and humidity > 90%, then there is a risk of road icing, and set the grid here as an impassable obstacle.

[0146] V. The objective function of the improved path planning algorithm, the added path weight function, calculates the optimal path for the vehicle to move from the current point to the target point. The path optimization function takes into account factors such as road slope, introduces real-time analysis of road surface water accumulation and road surface friction factors, calculates the appropriate weights of the actual cost function and the heuristic cost function, and sets the weights of each cost function. The objective function of the improved path planning algorithm is as follows:

[0147] f(n) = α2 * g(n) + (1 - α2) * h(n) + γ2 * K(n);

[0148] Among them, g(n) is the actual cost function, h(n) is the heuristic cost function, and K(n) is the vehicle condition weight function.

[0149]

[0150] Among them, s is the distance from the starting point to the ending point, d is the Euclidean distance from the current point to the ending point, α2 is the dynamic weight function at this time, and γ2 is the coefficient of the vehicle condition weight function at this time.

[0151] Among them, the actual cost function:

[0152]

[0153] Among them, x n , y n are the positions of the current vehicle, x n-1 , y n-1 are the positions of a point on the vehicle, f is the optimal theoretical friction value between the vehicle tire and the ground, and f n is the ground friction value of the current road section.

[0154] The heuristic cost function is:

[0155]

[0156] Among them, f i is the ground friction value of the current road section, θ is the current road slope, x i , y i are the positions of the current vehicle, x i-1 , y i-1 are the positions of a point on the vehicle, n is the value of the current grid point, and m is the value of the ending grid point.

[0157] The vehicle condition weight function is:

[0158]

[0159] Among them, M is the load of the vehicle for the current task, V is the average speed of the vehicle for the current task, and w is the road safety evaluation coefficient of the current road section.

[0160] VI. Real-time obstacle avoidance module. Both the front and rear vehicles have real-time obstacle avoidance functions (implemented based on lidar) to ensure that obstacles can be detected and avoided in a timely manner during driving.

[0161] When driving to the next calibration point, an obstacle is detected by the radar. The expandable radius R preset by the vehicle itself 小车 and the expandable radius R preset for the obstacle 障碍 are used to calculate the current passable distance, and it is determined whether the obstacle affects passage (as Figure 2 shown).

[0162] D = distance(C, O) - (R 障碍 + R 小车 );

[0163] where C is the center point of the current position of the vehicle, O is the center point position of the obstacle, and distance(C, O) is the Euclidean distance between the two.

[0164] An example calculation of determining whether an obstacle affects passage by the expansion distance is shown as Figure 2 shown.

[0165] If D < 0, then the obstacle affects passage at this time:

[0166] In step nine, the following vehicle uses the follow-up obstacle avoidance and adaptive guidance vehicle strategy for adaptive obstacle avoidance.

[0167] If D ≥ 0, indicating that the path is safe, then the obstacle does not affect passage:

[0168] Perform obstacle bypass processing, continue to move forward along the original path, and alarm the host computer to prompt to remove the obstacle.

[0169] VII. The described energy management mechanism, that is, using a battery management system to ensure the endurance ability, evaluating whether the remaining battery power is sufficient to complete the current task. If it is estimated that the power is insufficient, it will automatically arrange to charge to an appropriate power to ensure that the vehicle is always in the best working state.

[0170] At the same time, the system is equipped with intelligent charging decision-making. When it detects that the power is insufficient to support the subsequent journey, the system will use an improved A* algorithm to plan the optimal charging plan in advance, including selecting the location of the nearest charging pile and evaluating the required charging amount, and re-sorting the remaining tasks at the charging location.

[0171] And evaluate the battery efficiency coefficient of the vehicle when the charging amount is greater than 70% each time.

[0172]

[0173] When this value is lower than 80%, the vehicle alarms the host computer to prompt to replace the vehicle battery.

[0174] 8. Adaptive following safety distance for rear vehicles. In the double-vehicle following part, the rear vehicle can intelligently control the distance with the front vehicle according to the speed of the front vehicle, so as to avoid the possibility of accidents in case of emergency braking of the front vehicle when following at a faster speed.

[0175] Both the front and rear vehicles are equipped with laser distance sensors to measure the distance between the two vehicles. The maximum speed Vmax of the vehicle is pre-set, and the measured data are L1 and L2. L1 is the distance from the front vehicle to the rear vehicle measured by the laser distance sensor, and L2 is the distance from the rear vehicle to the front vehicle measured by the laser distance sensor. The geometric mean is calculated as follows:

[0176]

[0177] The L value is the current effective distance between the two vehicles.

[0178] Then calculate the safety distance threshold L at the current speed:

[0179]

[0180] k is the dynamic vehicle condition weight factor.

[0181]

[0182] Wherein, the first half is the dynamic reaction distance, the second half is the adaptive braking distance, v is the current speed of the leading vehicle, m is the total mass of the vehicle and the load of the current vehicle, m0 is the current standard unloaded mass of the vehicle, w is the current road safety evaluation coefficient, τ is the sensor delay time of the following vehicle, γ is the mass response coefficient, the default value is 0.5, and the nonlinear system compensation is performed on the current system, μ is the reference value of the ground friction coefficient of the current road section, and g is 9.8m / s 2 gravitational acceleration, α3 is the mass attenuation index, the default value is 0.15, β3 is the road condition mutation factor, the default value is 0.2, λ is the risk perception coefficient, the default value is 2.5, L 默 For the pre-set default following distance, a combination of mass logarithmic compensation and road condition exponential decay is introduced, and the dynamic risk perception factor realizes safety margin adaptation, which can improve the safety margin of the car compared to the traditional method.

[0183] If the current L>Ltarget, the rear vehicle will slowly accelerate under the premise of being less than Vmax to keep a safe distance from the front vehicle, and then gradually decelerate to the same speed as the front vehicle.

[0184] If the current L <L目,则后车缓慢减速与前车保持安全车距后逐渐加速到与前车速度一致。

[0185] IX. Rear - vehicle following obstacle avoidance and adaptive guide - vehicle strategy: When the rear vehicle following the front vehicle encounters an obstacle and cannot pass through, the front vehicle immediately transmits information to the rear vehicle to prompt it to avoid the obstacle. The rear vehicle selects obstacle avoidance according to the following situations to improve the operation efficiency.

[0186] 1. When encountering an obstacle (such as Figure 3 ) on single - lane roads like straight roads, curved roads, and turning roads, the rear vehicle adaptively becomes a guide vehicle, returns to the nearest passed marker point, and then re - uses the path - planning method of the guide vehicle mentioned above to plan the path to the end point. If it still fails, an error report is sent to enable relevant personnel to come for maintenance in time.

[0187] 2. When encountering an obstacle near an intersection (including but not limited to crossroads and T - shaped intersections), as Figure 4 shown, the road width is 2d at this time, and the center of the intersection is marked as the origin O1.

[0188] When an obstacle is found and cannot be passed through:

[0189] If the guide vehicle A has not passed through the intersection, the guide vehicle A remains unchanged and returns to the previous marker point to re - plan the path.

[0190] If the guide vehicle A has passed through the intersection, but the distance of A from the origin O1 < the distance of B from O1, the guide vehicle A still remains unchanged and returns to the previous marker point to re - plan the path.

[0191] If the guide vehicle A has passed through the intersection, and the distance of A from the origin O1 > the distance of B from O1, then vehicle B adaptively becomes the guide vehicle, and A becomes the rear vehicle, and they return to the previous marker point to re - plan the path. If it still fails, an error report is sent to remind relevant personnel to come for maintenance in time.

[0192] The above is only the preferred embodiment of the present invention and is not intended to limit the present invention. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for cooperative obstacle avoidance and path planning of two vehicles based on adaptive weights and multi-sensor fusion, characterized in that Including: Establish a map including obstacles and ground friction marks; Sort the transportation tasks based on a genetic algorithm considering dependencies to obtain the optimal transportation order; Use multi-sensor fusion to judge the risk of current road safety and calculate the road safety evaluation coefficient; Based on a path planning algorithm considering adaptive weights, calculate the optimal path of the vehicle; Execute the current transportation task according to the optimal transportation order and the optimal path; Determine the affected obstacles on the path according to the inflation radius of the vehicle and the obstacles, and perform adaptive obstacle avoidance according to the following vehicle following obstacle avoidance and adaptive leading vehicle strategy; Calculate the safety distance threshold based on the vehicle's dynamic reaction distance, the vehicle's adaptive braking distance, and the road safety evaluation coefficient; Control the speed of the following vehicle according to the comparison result between the effective distance between vehicles and the safety distance threshold.

2. The method for dual-vehicle cooperative obstacle avoidance and path planning based on adaptive weight and multi-sensor fusion according to claim 1, wherein The sorting of the transportation tasks based on the genetic algorithm considering dependencies to obtain the optimal transportation order includes: Based on the Euclidean distance between the current point and the target point, the task priority, the task duration, and the average speed recorded by the leading vehicle, and ensuring the correct order between tasks, construct a fitness function; Based on the genetic algorithm, and through operations including natural selection, crossover, and mutation, and based on the calculation result of the fitness function, screen and sort the transportation tasks to be executed to obtain the optimal transportation order; The fitness function is: Fitness = Q(K1)*(d + d1)+Q(K2)*M(d)*P+Q(K3)*t*v; In the formula, d is the Euclidean distance between the current point and the target point, and d1 is the Euclidean distance between the unloading point and the loading point; M(d) is the amplification function, P is the task priority, t is the task duration, and v is the average speed recorded by the leading vehicle; Q(K1) is the weight value corresponding to the distance, q(K2) is the weight value corresponding to the task priority, and Q(K3) is the weight value corresponding to the average speed.

3. The method for dual-vehicle collaborative obstacle avoidance and path planning based on adaptive weight and multi-sensor fusion according to claim 1, wherein The use of multi-sensor fusion to judge the risk of current road safety and calculate the road safety evaluation coefficient includes: Obtain multi-environment data in the road, and the multi-environment data includes the water accumulation depth, air humidity, environmental temperature, and rainfall; According to the multi-environment data, judge whether there is a risk of affecting vehicle functions or a risk of road icing. If so, set an obstacle impassable mark at the corresponding position on the map. If not, judge whether there is water accumulation or rainfall without the risk of icing. If so, reduce the weight of the corresponding route, otherwise do not adjust; Calculate the road safety evaluation coefficient according to the water accumulation depth and air humidity, and the calculation formula of the road safety evaluation coefficient w is: In the formula, H′ is the normalized air humidity, and D′ is the normalized water accumulation depth; Q(K4) is the weight value corresponding to the air humidity, and Q(k5) is the weight value corresponding to the water accumulation depth.

4. The method for dual-vehicle cooperative obstacle avoidance and path planning based on adaptive weight and multi-sensor fusion according to claim 1, characterized in that The calculation of the optimal path of the vehicle based on the path planning algorithm considering adaptive weights includes: Construct an actual cost function according to the vehicle position, the optimal value of the theoretical friction between the vehicle tire and the ground, and the ground friction value of the current section; Construct a heuristic cost function based on the vehicle position, the ground friction value of the current road section, the current road slope, the value of the current grid point, and the value of the end grid point; Construct a vehicle condition weight function based on the current task vehicle load, the average speed of the current task vehicle, and the road safety evaluation coefficient of the current road section; Use the actual cost function, the heuristic cost function, and the vehicle condition weight function to construct the objective function of the path planning algorithm, and the objective function is: f(n) = α2 * g(n) + (1 - α2) * h(n) + γ2 * K(n); In the formula, g(n) is the actual cost function, h(n) is the heuristic cost function, and K(n) is the vehicle condition weight function; α2 is a dynamic weight function, and γ 2 is the coefficient of the vehicle condition weight function; Construct a path planning algorithm, and obtain the optimal path of the vehicle from the current point to the target point according to the calculated value of the objective function.

5. The method for dual-vehicle cooperative obstacle avoidance and path planning based on adaptive weight and multi-sensor fusion according to claim 1, wherein Before performing the current transportation task according to the optimal transportation order and the optimal path, it also includes: Detect the remaining power of the vehicle battery. When the remaining power is not enough to support the subsequent journey, select the nearest charging pile location and evaluate the required charging amount, and re-sort the remaining task order at the charging location; When the charging amount of the vehicle battery is greater than the preset value each time, evaluate the vehicle battery efficiency, and when the vehicle battery efficiency is lower than the alarm threshold, alarm the host computer.

6. The method for dual-vehicle collaborative obstacle avoidance and path planning based on adaptive weight and multi-sensor fusion according to claim 1, wherein The determination of the affected obstacles on the path according to the expansion radii of the vehicle and the obstacles includes: Obtain the current passable distance according to the comparison result between the sum of the expansion radius preset by the vehicle itself and the expansion radius preset by the obstacle and the Euclidean distance between the vehicle and the obstacle; Determine the affected obstacles on the path according to the size of the current passable distance.

7. The method for dual-vehicle cooperative obstacle avoidance and path planning based on adaptive weight and multi-sensor fusion according to claim 6, wherein The determination of the affected obstacles on the path according to the size of the current passable distance includes: If the current passable distance is less than a predetermined value, the corresponding obstacle is the affected obstacle on the path; If the current passable distance is greater than or equal to the predetermined value, the corresponding obstacle does not affect passage, and after the vehicle bypasses the obstacle, it proceeds along the original path.

8. The method for dual-vehicle collaborative obstacle avoidance and path planning based on adaptive weights and multi-sensor fusion according to claim 1, wherein The adaptive obstacle avoidance according to the following vehicle following obstacle avoidance and the adaptive guiding vehicle strategy includes: In the case where the following vehicle follows the leading vehicle and encounters an obstacle and cannot pass, if it is on a one-way road, the following vehicle adaptively becomes a guiding vehicle, returns to the nearest passed marking point, and then performs path planning to the end point based on the path planning algorithm considering the adaptive weight; If it is at an intersection, judge whether the guiding vehicle passes through the intersection. If it does not pass through the intersection, the guiding vehicle remains unchanged and returns to the previous marking point to re-perform path planning; If it passes through the intersection, judge the size of the distance between the guiding vehicle and the intersection center and the distance between other vehicles and the intersection center. If the distance between the guiding vehicle and the intersection center is less than the distance between other vehicles and the intersection center, the guiding vehicle remains unchanged and returns to the previous marking point to re-perform path planning; If the distance between the guiding vehicle and the intersection center is greater than the distance between other vehicles and the intersection center, other vehicles adaptively become the guiding vehicle and return to the previous marking point to re-perform path planning.

9. The method for dual-vehicle collaborative obstacle avoidance and path planning based on adaptive weight and multi-sensor fusion according to claim 1, wherein The calculation of the safety distance threshold based on the vehicle dynamic reaction distance, the vehicle adaptive braking distance, and the road safety evaluation coefficient includes: Obtain the effective distance between adjacent vehicles, and calculate the dynamic vehicle condition weight factor according to the current road safety evaluation coefficient. The calculation formula for the dynamic vehicle condition weight factor is as follows: According to the dynamic vehicle condition weight factor, and in combination with the vehicle dynamic response distance and the vehicle adaptive braking distance, calculate the safety spacing threshold at the current speed. The calculation formula for the safety spacing threshold at the current speed is as follows: In the formula, v is the speed of the current leading vehicle, m is the total mass of the vehicle and the load of the current car, m0 is the standard unloaded mass of the current car, μ is the ground friction coefficient reference value of the current section, g is the gravitational acceleration value, τ is the sensor delay time of the following vehicle, and α3 is the default value of the mass attenuation index; w is the current road safety evaluation coefficient, γ is the mass response coefficient, β3 is the road condition mutation factor, and λ is the risk perception coefficient.

10. The method for dual-vehicle cooperative obstacle avoidance and path planning based on adaptive weights and multi-sensor fusion according to claim 1, wherein Controlling the speed of the following vehicle according to the comparison result between the effective distance between vehicles and the safety spacing threshold includes: If the effective distance between vehicles is greater than the safety spacing threshold, then on the premise that the following vehicle is less than the maximum speed of the preset vehicle, adjust to the same speed as the leading vehicle in a way of accelerating first and then decelerating; If the effective distance between vehicles is less than the safety spacing threshold, then on the premise that the following vehicle is less than the maximum speed of the preset vehicle, adjust to the same speed as the leading vehicle in a way of decelerating first and then accelerating.

Images