Unmanned vehicle path planning method and device, electronic equipment and readable storage medium

Abstract

This application relates to a method, apparatus, electronic device, and readable storage medium for path planning of unmanned vehicles, including: initializing a population through a preset chaotic mapping method; adjusting the exploration probability based on diversity and convergence indices; determining a preset iteration strategy for iterating the population based on the adjusted exploration probabilities; iterating each individual in the population based on the preset iteration strategy to obtain an iterative population; determining the fitness value of each individual in the iterative population; updating the global historical optimal path based on the candidate paths corresponding to the fitness values; repeatedly executing the steps of monitoring the diversity and convergence indices of the population and adjusting the exploration probability based on the diversity and convergence indices until the iteration termination condition is met; and using the current global historical optimal path as the target planning path so that the unmanned vehicle can drive based on the target planning path. The method provided in this application outputs a target planning path that balances path length, smoothness, and obstacle avoidance requirements.

Description

Technical Field

[0001] This application relates to the field of path planning technology, and in particular to an unmanned vehicle path planning method, device, electronic device and readable storage medium. Background Technology

[0002] In terms of optimization algorithms and intelligent control, metaheuristic algorithms such as the Eagle Optimizer are used to solve nonlinear optimization problems. They can generate a population randomly in the solution space using a pseudo-random number generator, but this may lead to uneven population distribution and an inability to systematically cover the entire solution space. Especially in high-dimensional problems, this can easily lead to a lack of global representativeness in the early stages of the algorithm, affecting the efficiency of subsequent global exploration, resulting in slow convergence or premature entrapment in local optima.

[0003] In existing technologies, for the exploration-exploitation balance problem, stage transitions can be controlled through preset fixed parameters or linear time-varying strategies. However, these strategies lack awareness of the actual running state of the algorithm and cannot be dynamically adjusted according to population diversity and convergence trends, easily leading to premature convergence or inefficient exploration. For local optima problems, random perturbation strategies such as Gaussian perturbation are often used to disturb elite individuals. However, these perturbations are not very directional and have a tendency to be blind, which may lead the solution to a worse region, resulting in inefficiency in escaping local optima.

[0004] However, existing technologies still suffer from shortcomings such as insufficient traversal of population initialization, rigid exploration and development conversion mechanisms, and lack of directional guidance for local perturbations. These shortcomings limit the accuracy, speed, and robustness of the algorithms when solving complex and dynamic optimization problems. Therefore, it is necessary to address the problems of low global exploration efficiency, premature convergence, and poor adaptability to dynamic environments in existing technologies. Summary of the Invention

[0005] In view of this, embodiments of this application provide an unmanned vehicle path planning method, device, electronic device, and readable storage medium to solve the problems of low global exploration efficiency, premature convergence, and poor adaptability to dynamic environments in the prior art.

[0006] A first aspect of this application provides an unmanned vehicle path planning method, including:

[0007] The system acquires the starting point, target point, and environmental map information of the autonomous vehicle. Based on these information, it initializes a population using a pre-defined chaotic mapping method. The environmental map information includes the locations of road boundaries and known fixed obstacles. Each individual in the population represents a candidate path from the starting point to the target point. The system monitors the diversity and convergence indices of the population and adjusts the exploration probability based on these indices.

[0008] Based on the adjusted exploration probability, a preset iteration strategy is determined for iterating the population. Each individual in the population is iterated upon according to this preset strategy to obtain an iterative population. The preset iteration strategy includes the Levy flight search strategy and a locally exploited weighted combined search strategy. The fitness value of each individual in the iterative population is determined, and the global historical optimal path is updated based on the candidate paths corresponding to the fitness values. The steps of monitoring the diversity and convergence indices of the population and adjusting the exploration probability based on these indices are repeated until the iteration termination condition is met. The iteration termination condition includes reaching a preset maximum number of iterations or the absolute value of the change in the fitness value corresponding to the global historical optimal path in consecutive iterations being less than a preset change. The current global historical optimal path is used as the target planning path so that the autonomous vehicle can travel based on the target planning path.

[0009] A second aspect of this application provides an unmanned vehicle path planning device, comprising:

[0010] The initialization module acquires the starting point, target point, and environmental map information of the autonomous vehicle. Based on these information, it initializes the population using a pre-defined chaotic mapping method. The environmental map information includes the locations of road boundaries and known fixed obstacles. Each individual in the population represents a candidate path from the starting point to the target point. The adjustment module monitors the diversity and convergence indices of the population and adjusts the exploration probability based on these indices. The processing module determines a pre-defined iteration strategy for iterating over the population based on the adjusted exploration probability. It then iterates over each individual in the population using this pre-defined strategy to obtain an iterative population. The iteration strategy includes the Levy flight search strategy and the local development weighted combination search strategy; the optimization module is used to determine the fitness value of each individual in the iterative population and update the global historical optimal path based on the candidate paths corresponding to the fitness values; the termination module is used to repeatedly execute the steps of monitoring the diversity and convergence indicators of the population and adjusting the exploration probability based on the diversity and convergence indicators until the iteration termination condition is met, wherein the iteration termination condition includes reaching a preset maximum number of iterations or the absolute value of the change in the fitness value corresponding to the global historical optimal path in consecutive iterations being less than a preset change amount; the control module is used to use the current global historical optimal path as the target planning path so that the unmanned vehicle travels based on the target planning path.

[0011] A third aspect of this application provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the above-described method.

[0012] A fourth aspect of this application provides a readable storage medium storing a computer program that, when executed by a processor, implements the steps of the above-described method.

[0013] Compared with the prior art, the beneficial effects of this application embodiment are as follows: by initializing the population through chaotic mapping, the candidate paths are more evenly distributed in the solution space, avoiding the individual clustering that may be caused by random initialization. During the iteration process, the exploration probability is dynamically adjusted by monitoring the diversity and convergence indicators of the population in real time. It can adaptively balance global search and local development according to the population state, avoiding the defects of unstable performance of fixed parameter strategies at different problem stages. The preset iteration strategy for iterating the population is determined by the adjusted exploration probability, and iteration is carried out according to the preset iteration strategy, which enhances the directionality and stability of the iteration. Thus, it can continuously optimize candidate paths in complex environment maps and output target planning paths that take into account path length, smoothness and obstacle avoidance requirements. This improves the target path planning quality and navigation safety of unmanned vehicles in environmental map information, and solves the problems of insufficient traversal of population initialization, rigid balance between exploration and development and weak directionality of local perturbations in the prior art.

[0014] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and do not limit this application. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of this application, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 This is a schematic diagram illustrating an application scenario of an embodiment of this application;

[0017] Figure 2 This is a flowchart illustrating an unmanned vehicle path planning method provided in an embodiment of this application;

[0018] Figures 3(a) and 3(b) are schematic diagrams of a path planning method for bypassing obstacle areas provided in an embodiment of this application;

[0019] Figures 4(a), 4(b), and 4(c) are schematic diagrams of a process for avoiding sudden fixed obstacles provided by an embodiment of this application;

[0020] Figures 5(a), 5(b), and 5(c) are schematic diagrams illustrating a process of avoiding oncoming moving obstacles provided by an embodiment of this application;

[0021] Figures 6(a), 6(b), and 6(c) are schematic diagrams of a process for avoiding moving obstacles when nodes meet, provided in an embodiment of this application.

[0022] Figures 7(a), 7(b), 7(c) and 7(d) are schematic diagrams of a process of avoiding a moving obstacle that is overtaking in the same direction, provided by an embodiment of this application;

[0023] Figures 8(a), 8(b), and 8(c) are schematic diagrams of a process for avoiding multiple moving obstacles in complex conflicts provided by an embodiment of this application;

[0024] Figure 9 This is a schematic diagram of another population iterative update process for unmanned vehicle path planning provided in an embodiment of this application;

[0025] Figure 10 This is a schematic diagram of the structure of an unmanned vehicle path planning device provided in an embodiment of this application;

[0026] Figure 11 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application.

[0027] 101: First terminal device; 102: Second terminal device; 103: Third terminal device; 104: Server; 105: Network.

[0028] 1010: Initialization module, 1020: Adjustment module, 1030: Processing module, 1040: Optimization module, 1050: Termination module, 1060: Control module;

[0029] 11: Electronic device; 1101: Processor; 1102: Memory; 1103: Computer program. Detailed Implementation

[0030] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application may also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods have been omitted so as not to obscure the description of this application with unnecessary detail.

[0031] The following will describe in detail, with reference to the accompanying drawings, an unmanned vehicle path planning method and apparatus according to embodiments of this application.

[0032] Figure 1This is a schematic diagram illustrating an application scenario according to an embodiment of this application. The application scenario may include a first terminal device 101, a second terminal device 102, a third terminal device 103, a server 104, and a network 105.

[0033] The first terminal device 101, the second terminal device 102, and the third terminal device 103 can be hardware or software. When the first terminal device 101, the second terminal device 102, and the third terminal device 103 are hardware, they can be various electronic devices with displays and supporting communication with the server 104, including but not limited to smartphones, tablets, laptops, and desktop computers. When the first terminal device 101, the second terminal device 102, and the third terminal device 103 are software, they can be installed on the aforementioned electronic devices. The first terminal device 101, the second terminal device 102, and the third terminal device 103 can be implemented as multiple software programs or software modules, or as a single software program or software module; this application embodiment does not impose any limitations on this. Furthermore, various applications can be installed on the first terminal device 101, the second terminal device 102, and the third terminal device 103, such as data processing applications, instant messaging tools, social platform software, search applications, shopping applications, etc.

[0034] Server 104 can be a server that provides various services, such as a backend server that receives requests sent by terminal devices with which it has established communication connections. This backend server can receive and analyze the requests sent by the terminal devices and generate processing results. Server 104 can be a single server, a server cluster consisting of several servers, or a cloud computing service center. This application embodiment does not limit this.

[0035] It should be noted that the server 104 can be either hardware or software. When the server 104 is hardware, it can be various electronic devices that provide various services to the first terminal device 101, the second terminal device 102, and the third terminal device 103. When the server 104 is software, it can be multiple software programs or software modules that provide various services to the first terminal device 101, the second terminal device 102, and the third terminal device 103, or it can be a single software program or software module that provides various services to the first terminal device 101, the second terminal device 102, and the third terminal device 103. This application embodiment does not impose any limitations on this.

[0036] Network 105 can be a wired network using coaxial cable, twisted pair, and fiber optic connection, or it can be a wireless network that enables interconnection of various communication devices without wiring, such as Bluetooth, Near Field Communication (NFC), and Infrared. This application embodiment does not limit this.

[0037] Users can establish a communication connection with the server 104 via the network 105 through the first terminal device 101, the second terminal device 102 and the third terminal device 103 to receive or send information, etc.

[0038] It should be noted that the specific types, quantities and combinations of the first terminal device 101, the second terminal device 102, the third terminal device 103, the server 104 and the network 105 can be adjusted according to the actual needs of the application scenario, and this application embodiment does not limit this.

[0039] It should be noted that the acquisition, storage, use, and processing of data in this application embodiment all comply with the relevant provisions of national laws and regulations.

[0040] Furthermore, it should be noted that in the embodiments of this application, certain software, components, models, and other existing solutions in the industry may be mentioned. These should be considered as exemplary, and their purpose is only to illustrate the feasibility of implementing the technical solution of this application. However, it does not mean that the applicant has used or necessarily used the solution.

[0041] Figure 2 This is a flowchart illustrating an unmanned vehicle path planning method provided in an embodiment of this application. Figure 2 Autonomous vehicle path planning methods can be derived from Figure 1 The terminal device executes the command. For example... Figure 2 As shown, the autonomous vehicle path planning method includes:

[0042] Step S201: Obtain the starting point location, target point location, and environmental map information of the unmanned vehicle. Based on the starting point location, target point location, and environmental map information, initialize the population through a preset chaotic mapping method.

[0043] The environmental map information includes the locations of road boundaries and known fixed obstacles, and each individual in the population represents a candidate path from the starting point to the target point.

[0044] Specifically, the system acquires the autonomous vehicle's starting point, target point, and environmental map information. The starting point refers to the vehicle's spatial coordinates at the start of path planning, defining the starting point. The target point refers to the spatial coordinates of the vehicle's destination, defining the ending point. Environmental map information refers to relatively unchanging environmental data acquired through environmental perception modules (such as LiDAR, cameras, and high-precision maps), including the locations of road boundaries and known fixed obstacles. Road boundaries define the legal drivable area, and the locations of known fixed obstacles are used to pre-plan and avoid these immovable objects during path planning.

[0045] Furthermore, the starting point coordinates and target point coordinates can be received from the positioning module via the vehicle bus or communication interface. At the same time, the processed environmental map information can be received from the data fusion unit of the environmental perception module. The environmental perception module fuses, segments, and identifies the raw sensor data, extracts lane lines, curbs, guardrails, etc. as road boundary information, and identifies buildings, stationary vehicles, traffic facilities, etc. as known fixed obstacles. This information is then converted into the starting point position and target point position in vector or raster form with geographic coordinates.

[0046] Furthermore, the starting point, target point, and environmental map information can be encoded into a data structure that the optimization algorithm can process. For example, the map information can be converted into a cost map, where the cost of each grid or path segment reflects its traversal difficulty. The cost of areas where known fixed obstacles are located is set to extremely high, and areas outside the road boundaries are also set to impassable.

[0047] Furthermore, based on the starting point location, target point location, and environmental map information, the population is initialized using a preset chaotic mapping method. The initialization process refers to the process of generating the first generation of the population before the algorithm iterative optimization process begins. The preset chaotic mapping method is a mathematical method that generates a similar random, ergodic sequence through deterministic equations. It is used to generate population individuals with high ergodicity and uniformity to cover different regions of the solution space. The preset chaotic mapping method can be calculated iteratively using the Circle mapping formula.

[0048] This application embodiment obtains precise starting point location, target point location, and environmental map information, making the search space of subsequent optimization algorithms clear and conforming to actual road rules. By initializing the population through a chaotic mapping method based on a preset chaotic mapping iteration formula, the individuals in the generated population are more widely and evenly distributed in the solution space. This helps the autonomous vehicle's path planning algorithm to conduct a more comprehensive global exploration in the early stages, avoiding search blind spots caused by the population gathering in a certain local area. This lays a good foundation for finding globally better candidate paths and improves the overall quality and robustness of path planning.

[0049] Step S202: Monitor the diversity and convergence indices of the population, and adjust the exploration probability based on the diversity and convergence indices.

[0050] Specifically, diversity and convergence indices are monitored for the population. This monitoring and probability adjustment operation is performed after each population update. The diversity index is used to characterize the global exploration ability of the population. The larger the value, the greater the difference in individual paths within the population and the larger the exploration range of the population. The convergence index is used to characterize the local development ability of the population. The smaller the convergence index, the better the convergence effect of the population towards the optimal individual.

[0051] In addition, the exploration probability is adaptively adjusted based on the monitoring results of diversity and convergence indicators. As the number of iterations increases, the exploration probability shows an overall decreasing trend, making the later stages of iteration biased towards local development.

[0052] This application embodiment can capture the optimization state of the population during the iteration process by monitoring the diversity and convergence indices of the population, so that the adjustment of the exploration probability has a real population state basis. The exploration probability is adaptively adjusted according to the diversity and convergence indices, realizing a dynamic balance between exploration capability and development capability, and improving the overall optimization efficiency.

[0053] Step S203: Determine a preset iteration strategy for iterating the population based on the adjusted exploration probability, and iterate each individual in the population based on the preset iteration strategy to obtain an iterative population.

[0054] Among them, the preset iterative strategies include the Levy flight search strategy and the local development weighted combination search strategy.

[0055] Specifically, the magnitude of the exploration probability reflects the degree of inclination to conduct a global exploration in the current iteration. The larger the value, the more individuals are inclined to perform a large-scale search with a certain probability. The smaller the value, the more individuals are inclined to make fine adjustments in the vicinity of the current region.

[0056] After determining the adjusted exploration probability, a random number with a value between [0, 1] is generated for each individual in the population. This random number is compared with the adjusted exploration probability, and the corresponding preset iteration strategy is determined based on the comparison result to obtain the iterative population.

[0057] The preset iterative strategies include the Lévy flight strategy and the local development weighted combined search strategy. The Lévy flight strategy's occasional long step size helps to escape local extrema and avoid premature convergence. The local development weighted combined search strategy's multi-information fusion avoids the directional bias of a single information source and enhances the stability and directionality of the local search. Thus, it can take into account both the breadth of global exploration and the accuracy of local development throughout the entire iteration process, resulting in better convergence performance and the quality of candidate paths.

[0058] Among them, Levy flight is a random walk with a step size that follows a heavy-tailed distribution. Its significant feature is that the vast majority of the step sizes are very short (which is beneficial for local search), but occasionally there will be very long jump step sizes (which is beneficial for global exploration). That is, it introduces the possibility of long-distance jumps in global exploration and introduces deterministic local search in local exploration.

[0059] Local refinement uses smaller, more deterministic, and more focused step sizes to search the vicinity of the current individual. It provides local development capabilities, and once the algorithm lands in a promising region, the refined local search can accurately find the optimal individual.

[0060] In this embodiment, different preset iteration strategies are selected based on the adjusted exploration probability. The preset iteration strategies are highly adapted to the current population optimization state. The Levy flight strategy achieves extensive coverage of the search space, reducing the possibility of missing the optimal path. The introduction of the local development weighted combination search strategy realizes the mining of high-quality search areas and improves the accuracy of the obtained global historical optimal path.

[0061] Step S204: Determine the fitness value of each individual in the iterative population, and update the global historical best path based on the candidate paths corresponding to the fitness values.

[0062] Specifically, the fitness value of each individual in the iterative population can be determined by a preset cost function. The fitness value is negatively correlated with the corresponding candidate path. That is, the smaller the fitness value, the shorter the candidate path, the smoother the turn, and the less risk of obstacle collision.

[0063] The candidate path corresponding to the individual with the minimum fitness value in the iterative population is compared with the currently recorded global historical best path. If the fitness value of the best individual path in the iterative population is smaller, the global historical best path is updated to the best individual path in the iterative population. If the fitness value of the best individual path in the iterative population is greater than or equal to the fitness value of the current global historical best path, the global historical best path remains unchanged, thus completing the global historical best path update for a single iteration.

[0064] This application embodiment determines the fitness value of an individual and updates it by comparing the candidate paths corresponding to the individuals in the iterative population with the global historical optimal path. This can retain the optimal individuals found during the iteration process, provide accurate optimization guidance for subsequent iterations, and approach the global historical optimal path.

[0065] Step S205: Repeat the steps of monitoring the diversity and convergence indices of the population, and adjusting the exploration probability based on the diversity and convergence indices, until the iteration termination condition is met.

[0066] The iteration termination conditions include reaching the preset maximum number of iterations or the absolute value of the change in the fitness value corresponding to the global historical best path in multiple consecutive iterations being less than the preset change amount.

[0067] Specifically, the steps of S202 are repeated, namely, monitoring the diversity and convergence indices of the current iterative population, readjusting the exploration probability based on the new monitoring results, and performing population updates, fitness value calculations, and global historical best path updates according to the adjusted exploration probability, so as to continuously iterate the algorithm until the iteration termination condition is met.

[0068] The iteration termination condition includes two parallel conditions, and the iteration stops when either one is met. The first iteration termination condition is reaching the preset maximum number of iterations. The second iteration termination condition is that the absolute value of the change in the fitness value corresponding to the global historical optimal path in consecutive iterations is less than the preset change amount, which is used to indicate that the algorithm has converged. The change amount is the difference between the fitness value corresponding to the global historical optimal path in consecutive iterations and its previous fitness value.

[0069] This application embodiment iterates through the steps of monitoring diversity and convergence indicators, probabilistic exploration adjustment, and population update, allowing the algorithm to gradually approach the global optimal path through multiple optimizations. At the same time, each population iteration readjusts the exploration probability based on the current population state, enabling the population to maintain a dynamic optimization strategy during the iteration process, thereby improving the overall convergence performance.

[0070] Step S206: Use the current global historical best path as the target planning path so that the autonomous vehicle can drive based on the target planning path.

[0071] Specifically, the currently recorded global historical best path is determined as the target planning path for this autonomous vehicle path planning, and the coordinate information of the target planning path is sent to the autonomous vehicle's driving control module. Based on the coordinate information of the target planning path, the module controls the autonomous vehicle's driving parameters such as driving direction and speed, so that the autonomous vehicle can start from the starting point and drive smoothly along the target planning path to the target point.

[0072] The embodiments of this application use the global historical optimal path at the end of the iteration as the target planning path, which satisfies the multiple constraints of the unmanned vehicle on path length, turning angle and obstacle avoidance in the environment. This can ensure the efficiency and safety of the unmanned vehicle's driving, directly guide the actual driving of the unmanned vehicle, and improve the practicality of the path planning method.

[0073] This application embodiment achieves uniform initialization of the population through a preset chaotic mapping method, ensuring the diversity and effectiveness of the initial population. By real-time monitoring of population diversity and convergence indices, adaptive adjustment of the exploration probability is achieved, resolving the core contradiction of the difficulty in balancing exploration and development in traditional algorithms. By determining the corresponding preset iteration strategy based on the adjusted exploration probability, dynamic adaptation between global exploration and local development is achieved, improving the optimization efficiency and path accuracy of the algorithm. Through multiple iterations of optimization and setting reasonable iteration termination conditions, it ensures that the global optimal path can be approximated, making the target planned path meet the driving requirements of the autonomous vehicle. This allows the autonomous vehicle to travel a shorter, smoother, and safer path from the starting point to the target point, effectively improving the path planning capability and autonomous driving safety of the autonomous vehicle in environmental map information.

[0074] In some embodiments, the population is initialized using a preset chaotic mapping method, including:

[0075] Determine the dimensions and population size corresponding to the path planning;

[0076] A multidimensional vector is randomly generated based on the dimension as the initial chaotic variable;

[0077] Based on the preset chaotic mapping method, the chaotic sequence is iteratively calculated with the initial chaotic variable as the starting value and the population size as the length of the chaotic sequence.

[0078] The chaotic sequence is linearly transformed to the actual range of values ​​corresponding to the path solution space to obtain the population.

[0079] Specifically, the path planning dimension and population size are determined. The path planning dimension can be determined by the spatial characteristics of the autonomous vehicle's driving environment. That is, when the autonomous vehicle is driving in a two-dimensional planar road environment, the path planning dimension is twice the number of path points. When the autonomous vehicle's driving environment involves three-dimensional spatial scenarios such as ramps and overpasses, the path planning dimension is three times the number of path points. The population size is the total number of candidate path individuals in the population. Its value needs to take into account both the algorithm's optimization efficiency and computational cost.

[0080] For example, the population size is preset to 40 in a two-dimensional path planning scenario and to 60 in a three-dimensional path planning scenario. This range of values ​​ensures that the population has sufficient diversity while avoiding a surge in initial computation due to an excessively large population size.

[0081] In addition, a multidimensional vector is randomly generated based on the path planning dimension as the initial chaotic variable. The initial chaotic variable is the starting value for initiating the chaotic mapping iteration, and its dimension is consistent with the path planning dimension. During the generation process, it must be ensured that the value of each element in the vector falls within the effective input interval of the chaotic mapping ([0,1] in this embodiment).

[0082] Taking a two-dimensional path planning scenario as an example, the randomly generated initial chaotic variable can be [0.23, 0.78]. The two elements of this vector correspond to the initial chaotic values ​​in the x-axis and y-axis directions, respectively, and all element values ​​satisfy the input requirements of the chaotic mapping, thus avoiding the failure of chaotic sequence generation due to the initial value exceeding the range.

[0083] In addition, based on the preset chaotic mapping method, the chaotic sequence is iteratively calculated with the initial chaotic variable as the starting value and the population size as the length of the chaotic sequence.

[0084] In the iterative calculation process, the preset chaotic mapping formula is applied to each element of the initial chaotic variable in each dimension, and a new chaotic variable is generated after each iteration until the total number of generated chaotic variables reaches the population size.

[0085] In this context, the state of each autonomous vehicle (an individual in the population) is randomly generated within the search space. The process of ensuring diversity in the initial state can be represented as follows:

[0086] ;in, Representing the individual, The first representing the state One dimension, The representative indicated the first The first individual One dimension, It is the first The individual Chaotic variables of dimension.

[0087] The initial state matrix of the entire autonomous vehicle swarm is represented as:

[0088] .

[0089] Population initialization can be performed using the Circle chaotic mapping formula, which ensures a more uniform spatial distribution of the population and achieves better optimization results than random generation. The expression of the Circle chaotic mapping formula is as follows:

[0090] ;in, Indicates the generated first The modulo operation, taking the current value of the chaotic sequence, ensures that the result always falls within the interval. Inside.

[0091] Furthermore, the chaotic sequence is linearly transformed to the actual value range corresponding to the path solution space to obtain the population. The path solution space is the actual feasible area for autonomous vehicle path planning, and its value range is determined by the road boundary in the environmental map. After completing the transformation of all chaotic variables, each transformed actual coordinate sequence constitutes a candidate path from the starting point to the target point, and the set of all candidate paths is the population.

[0092] In this embodiment, the population is initialized by a preset chaotic mapping method. By utilizing the inherent ergodicity of the chaotic sequence and its sensitivity to initial conditions, the generated population is more evenly and widely distributed in the solution space, covering the diversity of the entire path solution space. At the same time, it ensures that each candidate path falls within the road boundary range, effectively improving the quality of the population and reducing the amount of invalid computation for optimization in the iteration process.

[0093] In some embodiments, the exploration probability is adjusted based on diversity and convergence metrics, including:

[0094] Determine the iteration decay term, diversity weight, and convergence weight; the iteration decay term decreases as the number of iterations increases.

[0095] The impact value of diversity is determined based on diversity indicators and diversity weights, and the impact value of convergence is determined based on convergence indicators and convergence weights.

[0096] The sum of the iterative decay term, the diversity influence value, and the convergence influence value is determined as the exploration probability.

[0097] Specifically, the iteration decay term, diversity weight, and convergence weight are determined. The iteration decay term controls the overall decay trend of the exploration probability, showing a monotonically decreasing trend with the number of iterations. The diversity weight and convergence weight are used to quantify the coefficients of the influence of diversity and convergence indicators on the exploration probability. They need to be combined with the feature preset of the path planning scenario, and the sum of the two needs to be controlled within a reasonable range to avoid the excessive dominance of a single factor.

[0098] In addition, the diversity impact value and the convergence impact value are calculated separately. That is, the diversity impact value is determined based on the diversity index and diversity weight, and the convergence impact value is determined based on the convergence index and convergence weight.

[0099] Among them, a low population diversity index indicates that individuals in the population tend to cluster together, while a low convergence index indicates that the population is gradually approaching the optimal individual.

[0100] Furthermore, the exploration probability of the current iteration round is obtained by summing the iteration decay term, the diversity influence value, and the convergence influence value. The expression for the exploration probability is:

[0101] ;in, It is an iterative decay term. This represents the total number of iterations. Diversity weight (usually 0.3-0.5). It is the convergence weight (usually 0.2-0.4). , is a normalized measure of diversity and convergence.

[0102] Diversity ) and convergence ( ) can be represented as: ;in, It is the population size. No. The state of an individual. Average population status. It is the vector magnitude.

[0103] ;in, The optimal fitness value in the current iteration. The previous generation's optimal fitness value, Initial optimal fitness value.

[0104] This application embodiment provides a global iterative trend constraint on the exploration probability through an iterative decay term, avoiding over-exploration in the later stages of the iteration. By determining the exploration probability through summation, it achieves coordinated adjustment of multiple factors, allowing the exploration probability to respond to the real-time state of the population (diversity, convergence) and follow the overall law of algorithm iteration (iterative decay). This enables the algorithm to maintain an appropriate preset iterative strategy throughout the entire iteration cycle, avoiding the situation where exploration and development are difficult to balance.

[0105] In some embodiments, a preset iteration strategy for iterating over individuals in the population is determined based on an adjusted exploration probability, and the population is iterated over based on the preset iteration strategy to obtain an iterative population, including:

[0106] Generate random numbers within a preset numerical range;

[0107] For each individual, if the random number is less than the adjusted exploration probability, the individual is iterated based on the Levy flight search strategy to obtain an iterative population;

[0108] Provided that the random number is not less than the adjusted exploration probability, individuals are iterated through a locally developed weighted combinatorial search strategy to obtain an iterative population.

[0109] Specifically, the generated random number is compared with the adjusted exploration probability, which has been calculated in the previous steps based on the population diversity and convergence indices. Its value also falls within the range of [0, 1], reflecting the degree of tendency to conduct global exploration in the current iteration cycle.

[0110] For each individual in the current population, if the random number is less than the adjusted exploration probability, a Lévy flight search strategy is selected for that individual. Lévy flight is a random walk mechanism with a step size following a heavy-tailed distribution. To enhance search directionality, it is often combined with the optimal individual. The process can be represented as follows:

[0111] ;

[0112] in, Individual In the The state of the next iteration. This is the step size scaling factor. This represents element-wise multiplication (Hadamard product). It is a random step-size vector following a Lévy distribution, which can be implemented using the Mantegna algorithm. Furthermore, for each dimension... There exists the following expression: ;in, . , usually take .

[0113] and The process of determining can be expressed as: ;in It is the Gamma function. β is the characteristic exponent of the Lévy stable distribution, and its value range is: This is to ensure that the generated random step size follows a heavy-tailed distribution.

[0114] The β value directly affects the algorithm's exploration ability, as shown in Table 1.

[0115] Table 1. Impact of Exploration Ability

[0116]

[0117] The essence of iterative search using the Lévy flight strategy is to superimpose a random perturbation following a Lévy distribution under the guidance of the current global optimal direction. A large number of short step-size components help to conduct a refined search near the current optimal region, while occasional long step-size components enable individuals to traverse larger distances and jump to more distant regions in the solution space for exploration.

[0118] When the random number is not less than the adjusted exploration probability, a local development weighted combination search strategy is selected for the current individual. Its essence is to conduct a deep search in the most promising area. The weighted combination strategy guides the search direction by integrating information from multiple "guides" to avoid the limitations of a single information source.

[0119] ;

[0120] in, This is the current state; maintain the continuity and stability of the search. It is the globally historically optimal path, guiding the entire population to converge toward the globally historically optimal path. It is the individual's historical best, maintaining the continuity of the individual's search experience. It is the optimal individual among other individuals, achieving a balance between exploration and development through local collaboration, information sharing, and mutual assistance.

[0121] Weight parameters The method for determining is as follows:

[0122] The inertia decreased from 0.5 to 0.2; Global bootloader performance was upgraded from 0.2 to 0.5; Individual history is fixed at 0.2; Random exploration decreased from 0.1 to 0, where, .

[0123] This application embodiment iterates each individual in the population using different preset iteration strategies to obtain the iterative population corresponding to the current population. The iterative population can be used as input for subsequent steps to calculate fitness values, update the global historical optimal path, and re-monitor diversity and convergence indices, thereby improving the diversity and convergence of the population.

[0124] In some embodiments, the fitness value of each individual in the iterative population is determined, and the global historical best path is updated based on the candidate paths corresponding to the fitness values, including:

[0125] Based on the fitness values ​​of each individual in the iterative population, elite individuals are determined from the iterative population according to a preset proportion range;

[0126] For each elite individual, a corresponding Cauchy random vector is generated based on the standard Cauchy distribution;

[0127] Determine the difference vector between the current global best individual and the elite individual, and calculate the weighted sum of the Cauchy random vector and the difference vector to obtain the guiding perturbation vector of the elite individual;

[0128] The perturbation vector is superimposed onto the corresponding elite individual to determine the elite fitness value of the perturbed elite individual.

[0129] The candidate path corresponding to the minimum elite fitness value before and after the perturbation is determined as the current global historical best path.

[0130] Specifically, based on the fitness values ​​of each individual in the iterative population, elite individuals are determined from the iterative population according to a preset proportion range. The preset proportion range is used to control the percentage of elite individuals selected from the current population relative to the total population, in order to balance the algorithm's development capability and computational overhead.

[0131] In addition, for each selected elite individual, a Cauchy random vector that follows a standard Cauchy distribution is generated. The standard Cauchy distribution is a probability distribution, which means that the random numbers generated have a higher probability of appearing as large values ​​far from the mean. The resulting Cauchy random vector provides the basic random perturbation component for the update of elite individuals, so that the individuals can escape local optima.

[0132] Furthermore, the difference vector between the current global best individual and the elite individual, where the current global best individual is the individual with the smallest fitness value in all iterations, represents the direction from the current elite individual to the global best individual, providing guidance information for perturbation towards a better region.

[0133] Furthermore, the generated Cauchy random vector can be weighted and synthesized with the calculated difference vector to obtain a guiding perturbation vector. This guiding perturbation vector is then applied to elite individuals to perturb them. The direction of this perturbation vector is guided by the difference vector, while its magnitude and randomness are influenced by the Cauchy random vector.

[0134] The elite fitness value corresponding to the perturbed elite individual is determined according to the preset cost function. The elite fitness values ​​before and after the perturbation are compared, and the candidate path corresponding to the minimum value is determined as the current global historical optimal path, thereby realizing the update of the global historical optimal path.

[0135] In autonomous vehicle path planning, each individual corresponds to a parameterized representation of a candidate path. Adding the guiding perturbation vector to the elite individual is a vector addition operation. Its function is to superimpose a perturbation with a specific direction and amplitude on the elite individual, thereby generating a new candidate path. Completing the perturbation update means that the elite individual has been adjusted. The new elite individual may have a smaller fitness value, or it may be used to maintain the diversity of the population.

[0136] This application's embodiments generate random perturbation vectors using a standard Cauchy distribution with heavy-tailed characteristics, giving the perturbations stronger mutation capabilities and exploratory nature. This helps the algorithm escape local optima traps. The difference vector between the current global best individual and elite individuals is used as guiding information, directing the perturbation direction towards the currently known better region, enhancing the targeting and efficiency of local search. By combining random perturbations with directional guidance through weighted synthesis, the generated guiding perturbation vector can enhance local fine-grained search while adding random perturbations to individuals, increasing the exploration space range. This achieves a better balance between improving local development capabilities and avoiding premature convergence, thus improving the accuracy of the global historical best path determined by the optimization algorithm.

[0137] In some embodiments, it also includes:

[0138] Monitor real-time environmental information while the autonomous vehicle is driving along the planned path;

[0139] Based on environmental map information and real-time environmental information, and given the dynamic obstacles that appear ahead of the target planned path, the collision risk area where the dynamic obstacles and the autonomous vehicle collide is predicted.

[0140] The collision risk area is marked as a temporary restricted area. Taking the current position of the autonomous vehicle as the starting point, the starting point position, target point position, and environmental map information of the autonomous vehicle are obtained. Based on the starting point position, target point position, and environmental map information, the updated target planning path is obtained. The updated target planning path is an alternative path to avoid the temporary restricted area.

[0141] Specifically, as the autonomous vehicle travels along the planned path, real-time environmental information is acquired and analyzed. Combined with environmental map information, this information is used to predict the potential collision location of newly appearing dynamic obstacles ahead of the planned path. Dynamic obstacles refer to obstacles that newly appear or change position during the autonomous vehicle's movement, and the collision risk area refers to the spatial range predicted based on the dynamic obstacle's trajectory and the autonomous vehicle's current path, indicating a possible collision.

[0142] In addition, the predicted collision risk areas are marked as temporary restricted areas as hard constraints in the replanning process of the target planned path, and the replanning process is triggered to recalculate the optimized process of the safe path from the current position to the target point.

[0143] For example, when an autonomous vehicle is traveling along a planned path towards the loading area in a warehouse environment, the environmental perception module detects a forklift suddenly emerging from behind a shelf and crossing the planned path. Analyzing the forklift's speed and direction, it predicts that the forklift will collide with the autonomous vehicle at an intersection about ten meters ahead of the path. The module then marks the area around the intersection as a temporary restricted zone and, using the autonomous vehicle's current position as the new starting point, re-executes the entire process from chaotic mapping initialization to iterative optimization, generating a new path that bypasses the temporary restricted zone and guides the autonomous vehicle safely to the target point.

[0144] Furthermore, during the replanning process, the autonomous vehicle is subject to the following constraints: travel constraints, turning constraints, and temporary restricted areas. The cost function of path planning can be constructed through these constraints, thereby constraining the replanning process and ensuring the safety of the autonomous vehicle.

[0145] Due to limited energy (battery power), the autonomous vehicle faces a maximum travel constraint, meaning the travel path length must be limited by the maximum distance. To ensure successful mission completion and avoid energy depletion, we assume the maximum safe travel distance is... .

[0146] The total driving distance must meet the following constraints: Where node represents the number of travel segments. Indicates the first The distance traveled in a given segment of the journey should be chosen, and the shortest route should be selected as much as possible during the journey. ;in, This indicates that the path length is used as the standard, and the first... The optimal search path for each individual.

[0147] Turning angle constraints: When turning, autonomous vehicles need to adjust their heading angle to change course. The minimum turning angle is affected by the vehicle's wheelbase and speed. Design the centripetal acceleration (lateral acceleration) of the autonomous vehicle. ,radius The required lateral acceleration for turning is: .

[0148] There is an upper limit to the maximum lateral acceleration that a tire can provide: Where μ is the tire-road friction coefficient (approximately 0.8-1.0 for dry asphalt and approximately 0.3-0.5 for wet and slippery surfaces), and g is the acceleration due to gravity (9.8 m / s²).

[0149] Speed-related turning radius limits: To turn safely without skidding, the following must be met. ,Right now: .

[0150] This yields the minimum stable turning radius under dynamic constraints: For example, on dry road surfaces. At a speed of 60 km / h (16.67 m / s), the theoretical minimum stable turning radius required is at least 35.4 meters. If the turning radius is smaller than this value, the vehicle will lose control and skid.

[0151] The planned path in this embodiment can be obtained by using the cosine formula to obtain the turning angle. Let the nodes of the previous path be... , The autonomous vehicle is expected to turn and drive to the path node. The cosine of the angle between the two paths satisfies: .

[0152] Temporary restricted areas: During path planning, autonomous vehicles may encounter various obstacles, such as fixed stationary obstacles, moving obstacles, and suddenly appearing obstacles. When simulating obstacles, they can be represented by movable grids under a two-dimensional grid map. Figures 3(a) and 3(b) are schematic diagrams of path planning to bypass obstacle areas provided by an embodiment of this application. As shown in Figure 3(a), assuming that the autonomous vehicle is driving, it is expected to... arrive ,Although and It is not in the obstacle zone, but its driving path may enter the obstacle zone. Figure 3(b) shows that it can be accessed through the obstacle zone. and Add nodes between This allows autonomous vehicles to bypass obstacle areas to avoid collisions.

[0153] Cost function: In the process of autonomous vehicle path planning, the cost function is used to measure the merits of different paths. This cost function is also an evaluation function for improving traffic congestion optimization algorithms. A well-designed cost function can guide the algorithm to generate safe and efficient paths in complex environments, avoiding obstacles and threat areas. Based on a comprehensive consideration of factors such as maximum travel distance, turning angle constraints, and obstacle / threat area avoidance, this application constructs the following autonomous vehicle path cost function:

[0154] ;

[0155] ; ;

[0156] ;

[0157] ; ;

[0158] in, It is the total number of waypoints in the journey. Indicates distance cost. Indicates the cost of turning angle. Indicates the cost of obstacles and threats. , , These are the weighting coefficients corresponding to each cost item. Indicates the first The coordinates of the path points It is a path point The cosine value of the angle of rotation, and This represents the coordinates of the path points adjacent to it. Representing path points and waypoints The minimum distance between the formed path and the obstacle threat.

[0159] This application embodiment continuously monitors the environment and detects dynamic obstacles in real time, which can promptly discover new threats on the planned path, providing clear spatial constraints for path planning. By triggering a replanning process starting from the current position, it ensures the timeliness and continuity of the target driving path update, enabling the unmanned vehicle to actively avoid newly emerging obstacles in a dynamically changing environment in real time, thereby improving the safety of path planning and its adaptability to dynamic environments.

[0160] In some embodiments, when dynamic obstacles appearing ahead of the target planned path are identified based on environmental map information and real-time environmental information, the collision risk area where a collision between the dynamic obstacle and the autonomous vehicle is predicted includes:

[0161] When a dynamic obstacle and an autonomous vehicle are traveling in opposite directions and there is a path intersection point, based on the speed of the dynamic obstacle and the speed of the autonomous vehicle, predict the collision risk area at the path intersection point; and / or,

[0162] In the case where a dynamic obstacle intersects with the autonomous vehicle's driving direction and there is a crossroads, predict the collision risk area where both vehicles reach the crossroads within the same preset time window and a collision occurs; and / or

[0163] When a dynamic obstacle and an autonomous vehicle are traveling in the same direction and the autonomous vehicle's speed is greater than the obstacle's speed, predict the collision risk area where the autonomous vehicle will collide with the obstacle during its pursuit; and / or,

[0164] In the case where multiple dynamic obstacles coexist with the autonomous vehicle in the same environment, the collision risk areas between each dynamic obstacle and the autonomous vehicle are predicted, as well as the new collision risk areas between the autonomous vehicle and the next dynamic obstacle introduced by avoiding the previous dynamic obstacle.

[0165] Specifically, collision risk areas may include at least one of the following:

[0166] In the first scenario, the dynamic obstacle and the autonomous vehicle travel in opposite directions and have a path intersection point. "Opposite directions" means their movement directions are roughly opposite, and the path intersection point is the area where their paths overlap spatially. For example, on a straight section of a two-way single-lane road, vehicles traveling in opposite directions will inevitably overlap in the middle of the lanes. Based on the speed of the dynamic obstacle and the speed of the autonomous vehicle, the time it takes for both to reach the path intersection point can be calculated. If their arrival times are similar, the path intersection point is determined to be a collision risk area.

[0167] In the second scenario, the dynamic obstacle intersects with the autonomous vehicle's travel direction, creating a crossroads. Intersecting travel directions mean that their trajectories intersect on a planar projection; for example, the autonomous vehicle travels north-south while the dynamic obstacle travels east-west. A crossroads is the point in space where two paths traveling in different directions intersect, such as the center of an intersection. The arrival time of both vehicles at the crossroads is predicted within the same preset time window. If both vehicles arrive at the crossroads within this time window, the crossroads is identified as a collision risk area.

[0168] In the third scenario, the dynamic obstacle and the autonomous vehicle are traveling in the same direction, and the autonomous vehicle's speed is greater than the obstacle's speed. "Traveling in the same direction" means that both are moving in roughly the same direction, for example, both the autonomous vehicle and the vehicle in front are traveling from south to north. Due to the speed difference, the autonomous vehicle will gradually catch up with the dynamic obstacle. The risk area for collision during this catch-up is predicted. This risk area is typically located where the relative distance between the autonomous vehicle and the dynamic obstacle is less than a preset safety distance. Because the speed difference persists, multiple potential collision points may arise during the catch-up process.

[0169] In the fourth scenario, multiple dynamic obstacles coexist with the autonomous vehicle in the same environment. This increases the complexity of the scenario. It is necessary to consider not only the possible collisions with each dynamic obstacle individually, but also the new collision risks between the autonomous vehicle and other obstacles introduced by avoiding one obstacle. Therefore, it is necessary to predict the collision risk areas between each dynamic obstacle and the autonomous vehicle separately, and at the same time identify the new collision risk areas generated by the obstacle avoidance operation. Through this multi-round identification mechanism, comprehensive collision risk prediction can be achieved in complex scenarios with multiple obstacles.

[0170] In one embodiment, to ensure that the autonomous vehicle can effectively avoid obstacles and safely reach the target point, the path planning process is based on map rasterization to model the threat area. The experiment sets up 5 map models with different threat states to simulate autonomous vehicle driving road conditions of different complexity, which are discussed in the following cases: avoiding sudden fixed obstacles, avoiding moving obstacles that meet in the opposite direction, avoiding moving obstacles that meet at nodes, avoiding moving obstacles that overtake in the same direction, and avoiding multiple moving obstacles with complex conflicts.

[0171] Avoiding sudden, fixed obstacles: When faced with sudden, fixed obstacles, the obstacle avoidance strategy can respond quickly and take measures to ensure the safe operation of the autonomous vehicle, including accurate distance measurement, effective collision detection, timely path replanning, and continuous environmental monitoring. In this way, the algorithm can dynamically adapt to changes in complex environments, ensuring the safety and efficiency of autonomous vehicle navigation.

[0172] Figures 4(a), 4(b), and 4(c) are schematic diagrams illustrating a process of avoiding sudden fixed obstacles according to an embodiment of this application. An autonomous vehicle travels from the starting point to the target point along a straight path at a constant speed of 1 unit grid square per second (Figure 4(a)). At point A on the predetermined path, it encounters a fixed, sudden obstacle (Figure 4(b)). At this point, based on the aforementioned obstacle avoidance strategy mechanism, the obstacle is identified as affecting the normal movement of the autonomous vehicle, and point A is marked as an obstacle to be avoided, thus triggering a path replanning process, namely, the Improved Traffic Jam Optimizer (ITJO) algorithm obstacle avoidance path design, to avoid the obstacle and ensure the continuous movement of the autonomous vehicle. Subsequently, the autonomous vehicle encounters another static obstacle at point B during its journey. Through continuous monitoring, point B is identified as a new obstacle in real time, and the same ITJO algorithm is applied to design an obstacle avoidance path to avoid collisions with the obstacle at point B (Figure 4(c)).

[0173] Experimental results show that the proposed algorithm can effectively perform distance assessment and path replanning operations when faced with sudden static obstacles on the path, enabling the autonomous vehicle to successfully bypass the obstacles and safely reach its destination. This process not only verifies the adaptability and effectiveness of the ITJO algorithm in dynamic environments, but also demonstrates its potential to improve the navigation safety of autonomous vehicles.

[0174] Avoiding oncoming moving obstacles: In autonomous obstacle avoidance scenarios for self-driving vehicles (Vehicles), handling moving obstacles is a complex and critical issue. The changing positions of moving obstacles over time present additional challenges to the path planning of Vehicles. These obstacles not only require the Vehicle to accurately predict their future positions but also necessitate real-time adjustments to its route to avoid collisions. The key to handling moving obstacles lies in the ability to quickly and accurately predict their future positions and adjust the Vehicle's path planning accordingly, ensuring safe and efficient navigation in complex and ever-changing environments. This process requires not only efficient algorithms but also robust real-time processing capabilities to handle various unforeseen circumstances.

[0175] An autonomous vehicle and an obstacle moving in opposite directions encounter a direct head-on collision in a narrow passage or straight path, requiring dynamic path adjustment within a limited time window. Figures 5(a), 5(b), and 5(c) are schematic diagrams illustrating a process for avoiding an oncoming moving obstacle according to an embodiment of this application. As shown in 5(a), a moving autonomous vehicle encounters an oncoming obstacle on its path. Both move at a speed of 1 unit length / second, and a 3-second prediction time window is set for collision avoidance. To avoid a potential collision, the autonomous vehicle must complete its dynamic path adjustment within this limited time window.

[0176] Based on the obstacle avoidance strategy proposed above, the algorithm first calculates the possible positions of the autonomous vehicle and the moving obstacle within the next 3 seconds. Since both have a speed of 1 unit length / second, analysis reveals a potential collision point C. To ensure safe navigation, the potential collision point C is marked as a temporary restricted area (Figure 5(b)). Subsequently, based on the information from this temporary restricted area, the ITJO algorithm is used to plan a new path, redesigning a safe path to avoid the moving obstacle. The autonomous vehicle successfully avoids the moving obstacle according to the redesigned safe path (Figure 5(c)).

[0177] Through the above mechanism, potential collision risk points can be effectively identified and marked, and the autonomous vehicle can respond quickly in a short time to adjust its route, thereby ensuring that the autonomous vehicle can successfully bypass obstacles and continue to move towards the target location.

[0178] Avoiding moving obstacles at intersections: At path intersections, the simultaneous arrival of an autonomous vehicle and an obstacle in the intersection area can cause spatial contention. Figures 6(a), 6(b), and 6(c) are schematic diagrams illustrating a process for avoiding moving obstacles at intersections according to an embodiment of this application. As shown in Figure 6(a), there is a risk of collision between the mobile autonomous vehicle and a dynamic obstacle at their path intersection. Both vehicles travel at a speed of 1 unit length / second, and a 3-second prediction time window is set for pre-collision detection. The autonomous vehicle can predict the potential collision with the dynamic obstacle within the 3-second timeframe.

[0179] If the distance between the autonomous vehicle and the obstacle is less than 1 unit length at any time within the 3-second time window, a potential collision risk point is identified, as shown by points D and E (Figure 6(b)). The autonomous vehicle travels along its original path to point E, while the obstacle travels to point D at the same time. The relative distance between the two points begins to fall below the preset safe distance, therefore points D and E are designated as collision risk points, representing the two endpoints of a temporary restricted area. To address potential collisions, the identified collision point D is marked as a temporary restricted area, and the ITJO algorithm is used to replan the path to avoid these marked temporary restricted areas. The autonomous vehicle successfully avoids the moving obstacle using the redesigned safe path. This process is repeated until a safe and unobstructed path is determined (Figure 6(c)).

[0180] When an autonomous vehicle and an obstacle are about to meet at an intersection, the vehicle can proactively take evasive action, replanning its path using the ITJO algorithm to avoid collision risks. The vehicle will then choose to bypass the intersection and continue its journey. This mechanism ensures the vehicle's safe navigation capabilities in complex and dynamic environments, demonstrating a high degree of autonomy and adaptability.

[0181] Avoiding moving obstacles that are overtaking in the same direction: When an autonomous vehicle (RV) moves in the same direction as other vehicles moving at low speeds, the risk of a rear-end collision due to the difference in relative speed needs to be addressed by solving the problems of overtaking path planning and safe distance control. Figures 7(a), 7(b), 7(c), and 7(d) are schematic diagrams of a process for avoiding moving obstacles that are overtaking in the same direction, provided by an embodiment of this application. The moving RV moves at a constant speed of 2 units / second, while the dynamic obstacle moves at a constant speed of 1 unit / second (Figure 7(a)). Due to the difference in speed, there is a possibility that the RV may catch up with the dynamic obstacle and collide with it. To deal with this situation, a 3-second prediction time window is set to detect potential collision risks in advance. The ITJO algorithm identifies potential collision locations, namely points F and G. The autonomous vehicle travels along the original path to point F, and at the same time, the obstacle travels to point G. The relative distance between the two begins to be less than the preset safe distance. Therefore, points F and G are set as collision risk points. F and G are the two endpoints of the temporary restricted area. These locations are marked as temporary restricted areas, and the path replanning process is triggered to avoid these potential risk points (Figure 7(b)).

[0182] During the pursuit of dynamic obstacles by an autonomous vehicle, there may be one or more situations requiring path adjustments. Due to the speed difference between the two, new potential collision points may still exist after the initial avoidance, as shown by point H in Figure 7(c)). Faced with potential collisions caused by continuous pursuit, the algorithm continuously monitors new potential collision points and marks them one by one as temporary no-entry zones. Subsequently, by repeatedly performing path replanning operations, a safe path that completely avoids all known potential collision risks is found (Figure 7(d)).

[0183] This process demonstrates that even with speed adjustments, the location-based obstacle avoidance strategy and algorithm can still ensure the safe navigation capability of the autonomous vehicle in dynamic environments, showcasing the flexibility and robustness of the strategy and algorithm. Furthermore, by continuously repeating this detection and obstacle avoidance process until a collision-free safe path is found, the autonomous navigation performance of the autonomous vehicle in complex and changing environments is further improved.

[0184] Avoiding complex and conflicting moving obstacles: When multiple moving obstacles exist in the environment, the complexity of obstacle avoidance for autonomous vehicles increases significantly. It requires not only considering the trajectory of each obstacle but also comprehensively analyzing their interactions and the impact of these factors on the planned path. In other words, the autonomous vehicle must not only handle each obstacle individually but also consider the interactions of all obstacles and their overall impact. At this point, efficient path planning algorithms and real-time response mechanisms are crucial.

[0185] Figures 8(a), 8(b), and 8(c) are schematic diagrams of a process for avoiding complex conflicts involving multiple moving obstacles provided in an embodiment of this application. As shown in Figure 8(a), the mobile unmanned vehicle interacts with three dynamically moving obstacles moving in opposite directions. , and They coexist in the same environment.

[0186] First, the driverless car attempts to avoid moving obstacles. At this point, it first marks the predicted potential collision point J as a temporary restricted area and then performs secondary path planning based on this. However, in the current situation, proceeding according to this secondary planned path may cause the autonomous vehicle to enter a collision zone. or The collision path could lead to new potential collision risks (Figure 8(b)).

[0187] Based on the above, to effectively avoid all dynamic obstacles, the algorithm further identifies these newly emerging potential collision points I and K, and marks them as temporary restricted areas as well. Subsequently, considering all identified temporary restricted areas (including points I, J, and K), the algorithm performs a third path planning operation. This process ensures that the autonomous vehicle can successfully bypass the obstacles. , and Find a completely safe and unobstructed path to the target location for all dynamic obstacles (Figure 8(c)).

[0188] The multi-stage path planning strategy not only demonstrates the algorithm's adaptability in complex and dynamic environments but also showcases its effectiveness in handling continuously occurring obstacle avoidance challenges. By continuously updating temporary restricted areas and repeating the path planning process, the autonomous vehicle can intelligently select the optimal path while avoiding collisions with any obstacles, proving the potential of the ITJO algorithm to improve the safety and efficiency of autonomous vehicle navigation.

[0189] In the present application's embodiments, in oncoming encounter scenarios, it can identify conflict points and avoid them in advance; in node encounter scenarios, it can identify conflicts at intersecting nodes; in same-direction overtaking scenarios, it can continuously monitor multiple collision points caused by speed differences; and in complex scenarios with multiple obstacles, it can identify new collision risks introduced by avoidance actions and coordinate their avoidance. It transforms the dynamic interaction problem into the problem of identifying spatially overlapping areas. Through classification and prediction, it covers the main conflict patterns of autonomous vehicles interacting with dynamic obstacles in urban traffic environments. The identification of new collision risk areas ensures obstacle avoidance safety in multi-obstacle scenarios, avoiding new collisions caused by neglecting one aspect, thereby improving the autonomous navigation capability and driving safety of autonomous vehicles in complex dynamic environments.

[0190] Figure 9 This is a schematic diagram of another population iterative update process for path planning provided in an embodiment of this application, as shown below. Figure 9 As shown, the basic Traffic Jam Optimizer (TJO) algorithm uses random generation when initializing the population, which cannot cover the entire search space and results in unsatisfactory optimization. The Circle chaotic mapping method is introduced for population initialization. The chaotic mapping has good traversal and uniformity, which can ensure that the population is more evenly distributed in space and achieve better optimization results than random generation.

[0191] Dynamic Exploration Probability: As a metaheuristic algorithm, the core challenge of the TJO algorithm lies in the contradiction between exploration and exploitation. Exploration aims to broadly search the solution space, discover new regions, and avoid premature convergence. Exploration, on the other hand, involves deep searching within promising regions to improve convergence accuracy. This application's embodiments do not use fixed parameters for exploration and exploitation decisions; instead, they employ adaptive decision-making based on population diversity and convergence awareness.

[0192] Lévy Flight Global Exploration Step Size: Lévy Flight is a random walk with a heavy-tailed step size distribution. Its key characteristic is that the vast majority of step sizes are very short (beneficial for local search), but occasionally very long jump step sizes appear (beneficial for global exploration). That is, it introduces the possibility of long-distance jumps in global exploration and deterministic local search in local exploration. Local refinement uses smaller, more deterministic, and more concentrated step sizes to search in the vicinity of the current individual. It provides local exploration capabilities. Once the algorithm lands in a promising region, refined local search can precisely find the optimal individual.

[0193] During exploration, Levi's flight is used, and to enhance the directionality of the search, it is often combined with the optimal individual.

[0194] Local development weighted combination step size: The essence of the local development phase is to conduct a deep search in the most promising areas. The weighted combination strategy guides the search direction by integrating information from multiple "guides" to avoid the limitations of a single information source.

[0195] The elite individual-oriented Cauchy perturbation stage, also known as the elite individual-oriented Cauchy perturbation, is a method that applies Cauchy distribution perturbation to elite individuals, aiming to enhance the algorithm's ability to exploit local optima and escape local optima.

[0196] This application proposes a directional Cauchy perturbation, which is a directed exploration toward the currently known optimal region. For elite individuals... The Cauchy perturbation is updated as follows: ;

[0197] in, This represents the current state of elite individuals (good fitness, but possibly not the global optimum). C represents the globally optimal individual state found in the current population history, C(0,1) represents a random number that follows a standard Cauchy distribution (position parameter 0, scale parameter 1), and η represents the perturbation intensity coefficient (a small positive number). This represents the direction vector pointing to the global optimum.

[0198] The probability density function of the standard Cauchy distribution C(0,1) is: Guided Cauchy perturbation can enhance local search capabilities. When the perturbation step size is small, it performs a fine search around elite individuals; when the perturbation step size is large, it occasionally jumps out of the current region to explore potential neighboring individuals.

[0199] At the same time, such as Figure 9 As shown, the above steps are optimized based on the following steps, but they still follow the following traffic congestion algorithm, where the algorithm parameters are determined by the state of each individual in the group. Individual historical optimal state Group historical optimal state Population size Number of iterations It consists of, etc.

[0200] During the parameter initialization phase, the states of the autonomous vehicle swarm members are initialized, and the swarm members are randomly initialized within the search space. The state of each autonomous vehicle (individual) is randomly generated within the search space to ensure diversity of initial states:

[0201] ;in, Representing the individual, The first representing the state One dimension, The representative indicated the first The first individual Each dimension Represents a chaotic sequence value.

[0202] Set the objective function as The fitness value of each individual is expressed as: ,in, Indicates will Introducing the fitness function The fitness value. The shorter and smoother the path an autonomous vehicle takes through congested roads, the better its corresponding... The smaller the fitness value, the longer and more winding the path through congested sections. The larger.

[0203] Global exploration mode entering the inner loop phase: This phase requires extensive exploration of the state solution space to avoid prematurely getting trapped in local optima. Its mathematical description is as follows:

[0204] ;in, For the first The driver in The position updated in the autonomous phase during the next iteration. For the first The fusion of iteration results, i.e., the weighted result of individual historical best and group historical best, can be expressed as: ; .

[0205] in, As the iteration progresses, the number increases from 0 to 1, realizing the transition from individual experience to group experience. The violation degree coefficient simulates the behavior of individuals randomly deviating from the optimal direction, ensuring search diversity.

[0206] ;in, It is a random number between 0 and 1.

[0207] Autonomous stage, early iteration Smaller More dependent on individual historical best Individual exploration is highly independent, and the solution space covers a wide range; as iterations progress, As the population grows, the influence of group experience gradually becomes apparent, laying the foundation for subsequent convergence.

[0208] Transition and Local Adjustment Modes in the Inner Circulation Phase: Inner circulation driving easily leads to traffic congestion, meaning some individuals become trapped in local optima. At this point, drivers can autonomously adjust their driving strategies by observing the surrounding vehicle conditions, i.e., other individuals within the group. This can be mathematically described as follows:

[0209] ;

[0210] For the first Individual in the first The state is updated during the autonomous adjustment phase in the next iteration; For the first Individual in the first The state updated in the autonomous stage during the next iteration is the state of other randomly selected individuals, simulating the "driver's" observation and imitation of surrounding vehicles; For the first The results of each iteration are fused, resulting in a weighted state vector. It is a random number between 0 and 1.

[0211] It is a real-time generated random number between 0 and 1. Individuals move to the current position of another random individual (simulating "following others"), which enhances the algorithm's exploratory capabilities. By learning from random peers, individuals can explore new areas of the search space, maintain population diversity, and avoid premature convergence.

[0212] like The individual moves towards a randomly selected "historically optimal direction" (simulating "referencing historical experience"), which enhances the algorithm's development capabilities. By learning from historical experience, the individual is guided to conduct a more refined search in known promising areas, accelerating convergence.

[0213] This is an adjustment parameter that gradually decreases with iteration, causing the self-adjustment range to decrease from large to small. It maintains flexibility in the early stages of exploration while strengthening local development in the later stages. Its expression is: ;

[0214] This phase promotes information sharing within the group by introducing the influence of random neighbor individuals, while retaining the random direction switching mechanism to avoid excessive dispersion of individuals and prevent premature aggregation, thus balancing the intensity of exploration and development.

[0215] Entering the forced loop phase: Autonomous adjustment cannot alleviate the "congestion" (the group convergence speed slows down or gets stuck in a local optimum), so "traffic police" are needed to guide the forced optimization direction (to the globally optimal individual). This phase focuses on convergence towards the globally optimal region, improving the accuracy and convergence efficiency of the state individuals. The process can be represented as:

[0216] ;

[0217] in, This is the updated state after the enforcement phase; For the first The results of the next iteration are fused, and the weighted state vector is used, as described above. As the iterations increase, the guiding effect of the globally optimal direction is strengthened, causing individuals to quickly gather in high-quality areas, which can be represented as: ;

[0218] In the early stages of iteration With a smaller step size (around 0.5), individuals may still be relatively scattered. A smaller step size helps to smoothly begin a local search, avoiding missing fine areas due to blindly moving large steps.

[0219] Later stages of iteration A larger step size (around 1.5) indicates that the population has concentrated in the most promising region. A larger step size allows individuals to oscillate and explore more extensively around the optimal individual, helping to escape the "trap" of local optima, perform intensive exploration within a local range, and ensure the discovery of the true global optimum.

[0220] Boundary constraints and population updates:

[0221] For states that exceed the upper and lower bounds after the update, a truncation strategy is used to ensure the feasibility of the individual: ;

[0222] ;in, It is the individual state after the boundary constraint phase is updated, which is the individual's state. No. A new state in each dimension.

[0223] By comparing fitness values, the better individuals (those with smaller fitness values) are retained, thus achieving population evolution.

[0224] ;in, For the first After the nth iteration The state of an individual.

[0225] Individual historical best :like Then update for Group historical best This indicates that the process iterates through all individuals, and if there exists a first individual... Individual satisfaction Then update .

[0226] This means that, for an individual, It is a constant. For a group, .

[0227] Table 2 is a schematic table of key parameters of the TJO algorithm, which is an example of TJO algorithm parameters.

[0228] Table 2. Schematic diagram of key parameters of the TJO algorithm

[0229]

[0230] All of the above-mentioned optional technical solutions can be combined in any way to form the optional embodiments of this application, and will not be described in detail here.

[0231] The following are embodiments of the apparatus described in this application, which can be used to execute the embodiments of the method described in this application. For details not disclosed in the apparatus embodiments of this application, please refer to the embodiments of the method described in this application.

[0232] Figure 10 This is a schematic diagram of an unmanned vehicle path planning device provided in an embodiment of this application. Figure 10 As shown, the unmanned vehicle path planning device includes: an initialization module 1010, an adjustment module 1020, a processing module 1030, an optimization module 1040, a termination module 1050, and a control module 1060.

[0233] The initialization module 1010 is used to obtain the starting point position, target point position, and environmental map information of the unmanned vehicle. Based on the starting point position, target point position, and environmental map information, the population is initialized through a preset chaotic mapping method. The environmental map information includes the positions of road boundaries and known fixed obstacles. Each individual in the population represents a candidate path from the starting point position to the target point position.

[0234] The adjustment module 1020 is used to monitor the diversity and convergence indicators of the population and adjust the exploration probability based on the diversity and convergence indicators.

[0235] The processing module 1030 is used to determine a preset iteration strategy for iterating the population based on the adjusted exploration probability, and to iterate each individual in the population based on the preset iteration strategy to obtain an iterative population. The preset iteration strategy includes the Levy flight search strategy and the local development weighted combination search strategy.

[0236] The optimization module 1040 is used to determine the fitness value of each individual in the iterative population and update the global historical best path based on the candidate path corresponding to the fitness value.

[0237] The termination module 1050 is used to repeatedly execute the steps of monitoring the diversity index and convergence index of the population, and adjusting the exploration probability based on the diversity index and convergence index, until the iteration termination condition is met. The iteration termination condition includes reaching the preset maximum number of iterations or the absolute value of the change in the fitness value corresponding to the global historical best path in multiple consecutive iterations being less than the preset change.

[0238] The control module 1060 is used to use the current global historical best path as the target planning path so that the unmanned vehicle can drive based on the target planning path.

[0239] In some embodiments, the unmanned vehicle path planning device is used to: determine the dimension and population size corresponding to the path planning; randomly generate a multidimensional vector based on the dimension as an initial chaotic variable; iteratively calculate the chaotic sequence according to a preset chaotic mapping method, with the initial chaotic variable as the starting value and the population size as the length of the chaotic sequence; and linearly transform the chaotic sequence to the actual value range corresponding to the path solution space to obtain the population.

[0240] In some embodiments, the autonomous vehicle path planning device is used to: determine an iterative decay term, a diversity weight, and a convergence weight, wherein the iterative decay term decreases as the number of iterations increases; determine a diversity impact value based on a diversity index and a diversity weight, and determine a convergence impact value based on a convergence index and a convergence weight; and determine the sum of the iterative decay term, the diversity impact value, and the convergence impact value as the exploration probability.

[0241] In some embodiments, the unmanned vehicle path planning device is used to: determine a preset iteration strategy for iterating over each individual in the population, and iterate over the population based on the preset iteration strategy to obtain an iterative population, including: generating random numbers within a preset numerical range; for each individual, if the random number is less than the adjusted exploration probability, iterating over the individual based on the Levy flight search strategy to obtain an iterative population; and if the random number is not less than the adjusted exploration probability, iterating over the individual based on a local development weighted combination search strategy to obtain an iterative population.

[0242] In some embodiments, the unmanned vehicle path planning device is further configured to: determine elite individuals from the iterative population according to a preset ratio range based on the fitness values ​​of each individual in the iterative population; generate a corresponding Cauchy random vector for each elite individual based on the standard Cauchy distribution; determine the difference vector between the current global best individual and the elite individual; perform a weighted calculation on the Cauchy random vector and the difference vector to obtain the guidance perturbation vector of the elite individual; superimpose the guidance perturbation vector onto the corresponding elite individual to determine the elite fitness value corresponding to the perturbed elite individual; and determine the candidate path corresponding to the minimum value of the elite fitness values ​​before and after the perturbation as the current global historical best path.

[0243] In some embodiments, the autonomous vehicle path planning device is further configured to: monitor real-time environmental information while the autonomous vehicle is traveling along the target planned path; predict the collision risk area where a dynamic obstacle and the autonomous vehicle will collide, based on environmental map information and real-time environmental information and in the event of a dynamic obstacle appearing ahead of the target planned path; mark the collision risk area as a temporary restricted area; and, taking the current position of the autonomous vehicle as the starting point, perform the steps of obtaining the starting point position, target point position, and environmental map information of the autonomous vehicle, and obtaining an updated target planned path based on the starting point position, target point position, and environmental map information, wherein the updated target planned path is an alternative path that avoids the temporary restricted area.

[0244] In some embodiments, the autonomous vehicle path planning device is used to: predict a collision risk area at the path intersection point when a dynamic obstacle and an autonomous vehicle are traveling in opposite directions and there is a path intersection point, based on the movement speed of the dynamic obstacle and the driving speed of the autonomous vehicle; and / or, predict a collision risk area where the dynamic obstacle and the autonomous vehicle's driving directions intersect and there is a cross node, when the dynamic obstacle and the autonomous vehicle's driving directions intersect and there is a cross node, and / or, predict a collision risk area where the autonomous vehicle collides with the dynamic obstacle while chasing it, when the dynamic obstacle and the autonomous vehicle are traveling in the same direction and the driving speed of the autonomous vehicle is greater than the movement speed of the dynamic obstacle; and / or, predict the collision risk area between each dynamic obstacle and the autonomous vehicle, as well as the newly added collision risk area between the autonomous vehicle and the subsequent dynamic obstacle introduced by avoiding the previous dynamic obstacle, when multiple dynamic obstacles and the autonomous vehicle coexist in the same environment.

[0245] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0246] Figure 11 This is a schematic diagram of the electronic device 11 provided in an embodiment of this application. Figure 11 As shown, the electronic device 11 of this embodiment includes: a processor 1101, a memory 1102, and a computer program 1103 stored in the memory 1102 and executable on the processor 1101. When the processor 1101 executes the computer program 1103, it implements the steps in the various method embodiments described above. Alternatively, when the processor 1101 executes the computer program 1103, it implements the functions of each module / unit in the various device embodiments described above.

[0247] Electronic device 11 may be a desktop computer, laptop, handheld computer, cloud server, or other electronic device. Electronic device 11 may include, but is not limited to, processor 1101 and memory 1102. Those skilled in the art will understand that... Figure 11 This is merely an example of electronic device 11 and does not constitute a limitation on electronic device 11. It may include more or fewer components than shown, or different components.

[0248] The processor 1101 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc.

[0249] The memory 1102 can be an internal storage unit of the electronic device 11, such as a hard disk or RAM of the electronic device 11. The memory 1102 can also be an external storage device of the electronic device 11, such as a plug-in hard disk, Smart Media Card (SMC), Secure Digital (SD) card, FlashCard, etc., equipped on the electronic device 11. The memory 1102 can also include both internal and external storage units of the electronic device 11. The memory 1102 is used to store computer programs and other programs and data required by the electronic device.

[0250] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is merely an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiments can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0251] If integrated modules / units are implemented as software functional units and sold or used as independent products, they can be stored in a readable storage medium (e.g., a computer-readable storage medium). Based on this understanding, all or part of the processes in the methods of the above embodiments can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program may include computer program code, which may be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable storage medium may include: any entity or device capable of carrying computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc.

[0252] The above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. A method for unmanned vehicle path planning, characterized in that, include: The starting point position, target point position, and environmental map information of the unmanned vehicle are obtained. Based on the starting point position, target point position, and environmental map information, a population is initialized through a preset chaotic mapping method. The environmental map information includes the positions of road boundaries and known fixed obstacles. Each individual in the population represents a candidate path from the starting point position to the target point position. The diversity and convergence indices of the population are monitored to determine the iteration decay term, diversity weight, and convergence weight. The iteration decay term decreases as the number of iterations increases. The diversity impact value is determined based on the diversity index and the diversity weight, and the convergence impact value is determined based on the convergence index and the convergence weight. The sum of the iterative decay term, the diversity influence value, and the convergence influence value is determined as the exploration probability; Based on the exploration probability, a preset iteration strategy is determined to iterate the population. Based on the preset iteration strategy, each individual in the population is iterated to obtain an iterative population. The preset iteration strategy includes the Levy flight search strategy and the local development weighted combination search strategy. Determine the fitness value of each individual in the iterative population, and update the global historical optimal path based on the candidate path corresponding to the fitness value; Repeat the steps of monitoring the diversity and convergence indices of the population, determining the iteration decay term, diversity weight, and convergence weight, until the iteration termination condition is met. The iteration termination condition includes reaching a preset maximum number of iterations or the absolute value of the change in the fitness value corresponding to the global historical optimal path in consecutive iterations being less than a preset change. The current global historical best path is used as the target planning path so that the autonomous vehicle can travel based on the target planning path.

2. The method according to claim 1, characterized in that, The initialization of the population through a preset chaotic mapping method includes: Determine the dimensions and population size corresponding to the path planning; A multidimensional vector is randomly generated based on the aforementioned dimensions as the initial chaotic variable; According to the preset chaotic mapping method, the chaotic sequence is iteratively calculated with the initial chaotic variable as the starting value and the population size as the length of the chaotic sequence. The chaotic sequence is linearly transformed to the actual value range corresponding to the path solution space to obtain the population.

3. The method according to claim 1, characterized in that, The step of determining a preset iteration strategy for iterating over each individual in the population based on the adjusted exploration probability, and iterating over the population based on the preset iteration strategy to obtain an iterative population, includes: Generate random numbers within a preset numerical range; For each individual, if the random number is less than the adjusted exploration probability, the individual is iterated based on the Levy flight search strategy to obtain an iterative population; If the random number is not less than the adjusted exploration probability, the individual is iterated based on the local development weighted combinatorial search strategy to obtain an iterative population.

4. The method according to claim 1, characterized in that, The step of determining the fitness value of each individual in the iterative population and updating the global historical best path based on the candidate path corresponding to the fitness value includes: Based on the fitness value of each individual in the iterative population, elite individuals are determined from the iterative population according to a preset ratio range; For each of the elite individuals, a corresponding Cauchy random vector is generated based on the standard Cauchy distribution; Determine the difference vector between the current global optimal individual and the elite individual, and perform a weighted calculation on the Cauchy random vector and the difference vector to obtain the guidance perturbation vector of the elite individual; The guiding perturbation vector is superimposed on the corresponding elite individual to determine the elite fitness value corresponding to the perturbated elite individual; The candidate path corresponding to the minimum value of the elite fitness values ​​before and after the perturbation is determined as the current global historical optimal path.

5. The method according to claim 1, characterized in that, Also includes: During the process of the unmanned vehicle traveling along the target planned path, real-time environmental information is monitored; Based on the environmental map information and the real-time environmental information, when dynamic obstacles appear in front of the target planned path are identified, the collision risk area where the dynamic obstacles and the unmanned vehicle will collide is predicted. The collision risk area is marked as a temporary restricted area. Taking the current position of the unmanned vehicle as the starting point, the steps of obtaining the starting point position, target point position, and environmental map information of the unmanned vehicle are performed. Based on the starting point position, target point position, and environmental map information, the updated target planning path is obtained, wherein the updated target planning path is an alternative path to avoid the temporary restricted area.

6. The method according to claim 5, characterized in that, When dynamic obstacles appearing ahead of the target planned path are determined based on the environmental map information and the real-time environmental information, the collision risk area where the dynamic obstacle and the unmanned vehicle will collide is predicted, including: When the dynamic obstacle and the autonomous vehicle are traveling in opposite directions and there is a path intersection point, based on the speed of the dynamic obstacle and the speed of the autonomous vehicle, predict the collision risk area at the path intersection point; and / or, When the dynamic obstacle intersects with the autonomous vehicle's direction of travel and there is a crossing point, predict the collision risk area where both vehicles reach the crossing point within the same preset time window and a collision occurs; and / or, When the dynamic obstacle and the unmanned vehicle are traveling in the same direction and the unmanned vehicle's speed is greater than the obstacle's speed, predict the collision risk area where the unmanned vehicle will collide with the obstacle during its pursuit; and / or, In the case where multiple dynamic obstacles coexist with the unmanned vehicle in the same environment, the collision risk area between each dynamic obstacle and the unmanned vehicle is predicted, as well as the newly added collision risk area between the unmanned vehicle and the next dynamic obstacle introduced by avoiding the previous dynamic obstacle.

7. An unmanned vehicle path planning device, characterized in that, include: An initialization module is used to acquire the starting point position, target point position, and environmental map information of the unmanned vehicle. Based on the starting point position, target point position, and environmental map information, a population is initialized through a preset chaotic mapping method. The environmental map information includes the positions of road boundaries and known fixed obstacles. Each individual in the population represents a candidate path from the starting point position to the target point position. An adjustment module is used to monitor the diversity and convergence indices of the population, and determine the iteration decay term, diversity weight, and convergence weight, wherein the iteration decay term decreases as the number of iterations increases; The diversity impact value is determined based on the diversity index and the diversity weight, and the convergence impact value is determined based on the convergence index and the convergence weight. The sum of the iterative decay term, the diversity influence value, and the convergence influence value is determined as the exploration probability; The processing module is used to determine a preset iteration strategy for iterating the population based on the exploration probability, and to iterate each individual in the population based on the preset iteration strategy to obtain an iterative population, wherein the preset iteration strategy includes the Levy flight search strategy and the local development weighted combination search strategy. The optimization module is used to determine the fitness value of each individual in the iterative population and update the global historical optimal path based on the candidate path corresponding to the fitness value. The termination module is used to repeatedly execute the steps of monitoring the diversity index and convergence index of the population, determining the iteration decay term, diversity weight and convergence weight, until the iteration termination condition is met. The iteration termination condition includes reaching a preset maximum number of iterations or the absolute value of the change in the fitness value corresponding to the global historical optimal path in consecutive iterations being less than a preset change. The control module is used to take the current global historical best path as the target planning path, so that the unmanned vehicle can drive based on the target planning path.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the method as described in any one of claims 1 to 6.

9. A readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method as described in any one of claims 1 to 6.

Images