Automatic driving stop point selection method and device based on multi-modal multi-objective evolutionary algorithm
Optimizing the stop selection of autonomous vehicles through multimodal multi-objective evolution algorithms, solving the problem that the existing technology is difficult to take into account multiple target locations, and achieving lower fuel consumption and higher docking efficiency.
Patent Information
- Application Number
- CN202411071656.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-06
- Publication Date
- 2024-11-29
AI Technical Summary
Existing autonomous vehicles find it difficult to take into account multiple target locations when selecting stops, resulting in increased fuel consumption and poor docking results.
Using a multimodal multi-objective evolution algorithm, by obtaining the distance information between the to-selected stops and multiple target positions, using iterative processing of the main population and the auxiliary population, cross-optimization is used to determine the optimal stops to ensure the minimum distance and uniform distribution. .
It effectively reduces the increase in fuel consumption caused by unnecessary driving of autonomous vehicles, improves the diversity and rationality of stops, and reduces operating costs.
Smart Images

Figure CN119037467A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of autonomous driving technology, and in particular to an autonomous driving stop point selection method and device based on a multi-modal multi-objective evolutionary algorithm. Background Art
[0002] With the rapid development of artificial intelligence technology, autonomous vehicles have become an important part of the intelligent transportation system. Existing technology can automatically determine the stop location of unmanned vehicles to meet the needs of passengers to get on the bus nearby.
[0003] When faced with multiple target locations, this stop point determination method finds it difficult to find at least one optimal stop point that is evenly distributed to meet all the needs of multiple targets. In addition, the unreasonable selection of the stop point locations may lead to excessive unnecessary driving and idling of the vehicle, thereby increasing the fuel consumption of the autonomous driving vehicle. A better solution is needed. Summary of the invention
[0004] In view of this, the present invention provides an automatic driving stop point selection method and device based on a multi-modal multi-objective evolutionary algorithm to solve the problem of increased fuel consumption caused by the inability of automatic driving vehicles to simultaneously take into account multiple target locations when selecting a stop point.
[0005] On one hand, the present invention provides an automatic driving stop point selection method based on a multi-modal multi-objective evolutionary algorithm, the method comprising obtaining distance information between an individual stop point in a set of stop points for a vehicle and a plurality of target positions, and obtaining an individual distance information subset;
[0006] Utilizing the multimodal multi-objective evolutionary algorithm, taking the individual candidate stop points in the candidate stop point set and the distance information subset corresponding to the individual candidate stop points as individual parameter variables of the main population and the auxiliary population, taking the first objective function as the optimization target of the main population, taking the second objective function as the optimization target of the auxiliary population, performing iterative processing on the main population and the auxiliary population, and during the iterative processing, cross-optimizing the optimized parameter variable set of the main population and the optimized parameter variable set of the auxiliary population;
[0007] When the iterative processing meets the iteration stop condition, a target optimization parameter variable set is determined based on the optimization parameter variable set of the main population, and at least one optimal stopping point is determined based on the candidate stopping points in the target optimization parameter variable set; wherein the first objective function is determined based on the minimum distance information in the distance information subset of the optimization parameter variables; and the second objective function is determined based on the maximum distribution area of the candidate stopping points of the optimization parameter variables in the decision space.
[0008] On the other hand, the present invention also provides an automatic driving stop point selection device based on a multi-modal multi-objective evolutionary algorithm, the device comprising: a first determination module, for obtaining, for an individual stop point in a set of stop points for a vehicle, distance information between the individual stop point and a plurality of target positions, and obtaining an individual distance information subset; an algorithm iteration module, for using the multi-modal multi-objective evolutionary algorithm, using the individual stop points in the set of stop points for a vehicle and the distance information subsets corresponding to the individual stop points for a vehicle as individual parameter variables of a main population and an auxiliary population, using a first objective function as the optimization target of the main population, and using a second objective function as the optimization target of the auxiliary population, The main population and the auxiliary population are iteratively processed, and during the iterative processing, the optimization parameter variable set of the main population and the optimization parameter variable set of the auxiliary population are cross-optimized; a second determination module is used to determine the target optimization parameter variable set based on the optimization parameter variable set of the main population when the iterative processing meets the iteration stop condition, and determine at least one optimal stop point according to the candidate stop points in the target optimization parameter variable set; wherein the first objective function is determined based on the minimum distance information in the distance information subset of the optimization parameter variables; the second objective function is determined based on the maximum distribution area of the candidate stop points of the optimization parameter variables in the decision space.
[0009] On the other hand, the present invention also provides a computer device, which includes: a memory and a processor, the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor implements the above-mentioned automatic driving stop point selection method based on the multi-modal multi-objective evolutionary algorithm by executing the computer instructions.
[0010] On the other hand, the present invention also provides a computer-readable storage medium, on which computer instructions are stored, and the computer instructions are used to enable a computer to implement the above-mentioned automatic driving stop point selection method based on a multi-modal multi-objective evolutionary algorithm.
[0011] On the other hand, the present invention also provides a computer program product, including computer instructions, which are used to enable a computer to execute the automatic driving stop point selection method based on a multi-modal multi-objective evolutionary algorithm of the above-mentioned first aspect or any corresponding embodiment.
[0012] In this process, an individual distance information subset is determined according to the distance information from the individual candidate stop point to multiple target locations, and the individual candidate stop point and the corresponding distance information subset of the individual candidate stop point are used as individual parameter variables of the main population and the auxiliary population. The main population and the auxiliary population are cross-optimized, and the auxiliary population is set with the optimization goal of increasing the diversity of the candidate stop points. This allows the determination of the candidate stop points to not only meet the minimum distance to multiple target locations, but also increase the diversity of the candidate stop points, thereby reducing the fuel consumption problem of the autonomous driving vehicle due to unreasonable stop point selection and reducing the operating cost of the autonomous driving vehicle. BRIEF DESCRIPTION OF THE DRAWINGS
[0013] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the related technologies, the drawings required for use in the specific embodiments or the related technical descriptions will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0014] Figure 1 It is a flowchart of a method for selecting an automatic driving stop point based on a multi-modal multi-objective evolutionary algorithm provided by an embodiment of the present invention;
[0015] Figure 2 It is a schematic diagram of target position distribution on a physical map of an automatic driving stop point selection method based on a multi-modal multi-objective evolutionary algorithm provided by an embodiment of the present invention;
[0016] Figure 3 Schematic diagram of a multi-modal multi-objective problem of an automatic driving stop point selection method based on a multi-modal multi-objective evolutionary algorithm provided by an embodiment of the present invention;
[0017] Figure 4 It is a Pareto region schematic diagram of an automatic driving stop point selection method based on a multi-modal multi-objective evolutionary algorithm provided by an embodiment of the present invention;
[0018] Figure 5 It is a specific flow chart of a method for selecting an automatic driving stop point based on a multi-modal multi-objective evolutionary algorithm provided by an embodiment of the present invention;
[0019] FIG6( a ) is a schematic diagram of a stop point selection situation of a DN-NSGA-II algorithm of an automatic driving stop point selection method based on a multi-modal multi-objective evolutionary algorithm provided in an embodiment of the present invention;
[0020] FIG6( b ) is a schematic diagram of a stop point selection situation of an MMEA-WI algorithm of an automatic driving stop point selection method based on a multi-modal multi-objective evolutionary algorithm provided by an embodiment of the present invention;
[0021] FIG6( c ) is a schematic diagram of a stop point selection situation of a TADE_DDS algorithm of an automatic driving stop point selection method based on a multi-modal multi-objective evolutionary algorithm provided by an embodiment of the present invention;
[0022] FIG6( d ) is a schematic diagram of a stop point selection situation of an MMOEA / DC algorithm of an automatic driving stop point selection method based on a multi-modal multi-objective evolutionary algorithm provided by an embodiment of the present invention;
[0023] FIG6( e ) is a schematic diagram of a stop point selection situation of an MMOGA algorithm of an automatic driving stop point selection method based on a multi-modal multi-objective evolutionary algorithm provided by an embodiment of the present invention;
[0024] FIG6( f ) is a schematic diagram of a stop point selection situation of an automatic driving stop point selection method based on a multi-modal multi-objective evolutionary algorithm provided by an embodiment of the present invention;
[0025] FIG. 7( a ) is a partial schematic diagram of a stop point selection situation of a DN-NSGA-II algorithm of an autonomous driving stop point selection method based on a multi-modal multi-objective evolutionary algorithm provided by an embodiment of the present invention;
[0026] FIG7( b ) is a partial schematic diagram of a stop point selection situation of an MMEA-WI algorithm of an automatic driving stop point selection method based on a multi-modal multi-objective evolutionary algorithm provided by an embodiment of the present invention;
[0027] FIG7( c ) is a partial schematic diagram of a stop point selection situation of a TADE_DDS algorithm of an automatic driving stop point selection method based on a multi-modal multi-objective evolutionary algorithm provided by an embodiment of the present invention;
[0028] FIG7( d ) is a partial schematic diagram of a stop point selection situation of an MMOEA / DC algorithm of an automatic driving stop point selection method based on a multi-modal multi-objective evolutionary algorithm provided by an embodiment of the present invention;
[0029] FIG7( e ) is a partial schematic diagram of a stop point selection situation of an MMOGA algorithm of an automatic driving stop point selection method based on a multi-modal multi-objective evolutionary algorithm provided by an embodiment of the present invention;
[0030] FIG7( f ) is a partial schematic diagram of a stop point selection situation of an automatic driving stop point selection method based on a multi-modal multi-objective evolutionary algorithm provided by an embodiment of the present invention;
[0031] Figure 8 is a structural schematic diagram of an automatic driving stop point selection device based on a multi-modal multi-objective evolutionary algorithm provided by an embodiment of the present invention;
[0032] Fig. 9It is a structural schematic diagram of another automatic driving stop point selection device based on a multi-modal multi-objective evolutionary algorithm provided in an embodiment of the present invention. DETAILED DESCRIPTION
[0033] With the development of artificial intelligence technology, autonomous vehicles have become an important part of intelligent transportation systems. In the daily operation of autonomous vehicles, how to efficiently and safely select multiple equivalent stops to facilitate the allocation of vehicles to pick up passengers as usual in the face of emergencies and meet passenger requirements and preferences is a key issue in improving user experience and system performance.
[0034] In the prior art, based on the current position of the unmanned vehicle, a target stopping area is determined in the target area where the unmanned vehicle is located, and then a target stopping point is determined in the target stopping area, thereby obtaining a target driving route from the current position of the unmanned vehicle to the target stopping point, and controlling the unmanned vehicle to drive to the target stopping point based on the target driving route.
[0035] However, the above technology may have the following problems:
[0036] First, the existing technology only considers the needs of passengers to get off nearby, but does not consider the distance between the vehicle and the passengers at any place (especially popular places), which may result in excessive ineffective driving distance and increase vehicle fuel consumption;
[0037] Second, the selection of vehicle stops in the prior art can only consider the situation of one goal. When there are multiple goals, it is difficult to find at least one optimal stop that is evenly distributed to meet the needs of all goals. This may result in sacrificing the experience of other goals in order to meet a certain goal, thus affecting the overall vehicle stop effect.
[0038] Third, existing technologies may have limitations such as insufficient consideration of single-modal data and relatively simple target setting, which cannot fully meet the complex and changeable actual needs and may lead to insufficient passenger satisfaction.
[0039] In order to solve at least one of the above problems, various embodiments of the present invention provide an automatic driving stop point selection method based on a multi-modal multi-objective evolutionary algorithm, including: for an individual stop point in a set of stop points for a vehicle, obtaining distance information between the individual stop point and multiple target positions to obtain an individual distance information subset; using a multi-modal multi-objective evolutionary algorithm, using the individual stop point and the distance information subset corresponding to the individual stop point as individual parameter variables of a main population and an auxiliary population, using a first objective function as the optimization target of the main population, and using a second objective function as the optimization target of the auxiliary population, The main population and the auxiliary population are iteratively processed. During the iterative processing, the optimization parameter variable set of the main population and the optimization parameter variable set of the auxiliary population are cross-optimized; when the iterative processing meets the iteration stop condition, the target optimization parameter variable set is determined based on the optimization parameter variable set of the main population, and at least one optimal stop point is determined based on the candidate stop points in the target optimization parameter variable set; wherein the first objective function is determined based on the minimum distance information in the distance information subset of the optimization parameter variables; the second objective function is determined based on the maximum distribution area of the candidate stop points of the optimization parameter variables in the decision space.
[0040] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.
[0041] According to an embodiment of the present invention, a method for selecting an automatic driving stop point based on a multi-modal multi-objective evolutionary algorithm is provided. Figure 1 is a flow chart of a method for selecting an automatic driving stop point based on a multi-modal multi-objective evolutionary algorithm provided by an embodiment of the present invention, such as Figure 1 As shown, the process includes the following steps:
[0042] Step S101, for an individual stop point in the vehicle's stop point set, obtaining distance information between the individual stop point and multiple target locations to obtain an individual distance information subset;
[0043] Step S102, using a multi-modal multi-objective evolutionary algorithm, taking individual candidate stop points in the candidate stop point set and a distance information subset of the corresponding individual candidate stop points as individual parameter variables of the main population and the auxiliary population, taking the first objective function as the optimization target of the main population, taking the second objective function as the optimization target of the auxiliary population, performing iterative processing on the main population and the auxiliary population, and in the iterative processing process, cross-optimizing the optimization parameter variable set of the main population and the optimization parameter variable set of the auxiliary population;
[0044] Step S103, when the iterative processing meets the iteration stop condition, the target optimization parameter variable set is determined based on the optimization parameter variable set of the main population, and at least one optimal stop point is determined according to the candidate stop points in the target optimization parameter variable set; wherein the first objective function is determined based on the minimum distance information in the distance information subset of the optimization parameter variables; and the second objective function is determined based on the maximum distribution area of the candidate stop points of the optimization parameter variables in the decision space.
[0045] In one possible implementation, the vehicle may include but is not limited to an autonomous driving vehicle; the individual candidate stop point may refer to a coordinate point in the decision space to represent the stop point coordinates of the vehicle; the target position may refer to a preset position coordinate in the decision space to represent the actual position point in the physical map.
[0046] Here, the decision space may refer to a set describing all possible solutions or strategies, for example, a set of coordinates of all possible stop locations of a vehicle; the target location may refer to the location of a popular place in a physical map, which may include but is not limited to: elementary schools, junior high schools, convenience stores, and train stations in a physical map.
[0047] For example, Figure 2 As shown, Figure 2 This is a schematic diagram of the physical map target location distribution of an automatic driving stop point selection method based on a multi-modal multi-objective evolutionary algorithm provided by an embodiment of the present invention, where a primary school is represented by a triangular image, a junior high school is represented by a hexagon, a convenience store is represented by a diamond, and a railway station is represented by a five-pointed star surrounded by an outer frame, wherein:
[0048] The location coordinates of each primary school may include: (3, 37), (42, 96), (45, 60), (50, 25), (83, 72), (98, 38);
[0049] The location coordinates of each junior high school may include: (40, 20), (51, 60), (95, 51);
[0050] The location coordinates of each convenience store may include: (10, 55), (15, 15), (15, 78), (15, 88), (20, 23), (20, 70), (32, 42), (35, 60), (40, 76), (52, 78), (52, 96), (55, 33), (75, 27);
[0051] The location coordinates of each train station may include: (17.5, 82.5), (55.5, 82.5), (94.5, 6.5).
[0052] Furthermore, the distance information to the target location may refer to the Euclidean distance (or Euclidean distance), that is, the straight-line distance between two points in a two-dimensional or three-dimensional space.
[0053] In one possible implementation, in a multimodal multi-objective evolutionary algorithm, multi-objectives may refer to the distances to primary schools, junior high schools, convenience stores and train stations being multiple conflicting objectives, multi-modality may refer to the presence of multiple equivalent stops with the same distances to primary schools, junior high schools, convenience stores and train stations, and an evolutionary algorithm (EA) may refer to a random search algorithm based on biological evolution theory for solving optimization problems.
[0054] For example, the multi-objective evolutionary algorithm may include, but is not limited to: Non-dominated Sorting Genetic Algorithm II (NSGA-II) and Improved Strength Pareto Evolutionary Algorithm 2 (SPEA2).
[0055] For example, Figure 3 is a schematic diagram of a multimodal multi-objective problem of an automatic driving stop point selection method based on a multimodal multi-objective evolutionary algorithm provided by an embodiment of the present invention; wherein,
[0056] In the decision space graph on the left, the two triangles belong to different Pareto optimal solution sets and are far apart in the decision space, but in the target space graph on the right, the two triangles are mapped to the same target vector; similarly, in the decision space graph on the left, the five-pointed star and circle are also mapped to the same five-pointed star and circle in the target space graph on the right.
[0057] Here, multimodality can be characterized as the existence of at least two different Pareto optimal solution sets corresponding to the same Pareto optimal frontier; wherein, the Pareto optimal solution set can refer to the set of Pareto optimal solutions, which is used to characterize all optimal solutions or near-optimal solutions in the decision space; the Pareto frontier can refer to the curve or surface formed by all Pareto optimal solution sets in the target space.
[0058] For example, the multi-objective optimization problem is shown in the following formula:
[0059] Minimize F(x)={f 1 (x),f 2 (x),....,f m (x)},
[0060] stx=(x 1 ,x 2 ,...,x n )∈Ω
[0061] Where Ω can refer to the search space, m can refer to the number of objectives to be optimized, and x can refer to the number of n decision variables x. i The decision vector composed of Then represent x a Dominate x b , x a is a non-dominated solution, the set of all x represents the Pareto optimal solution, and the set of all f(x) represents the Pareto frontier.
[0062] Furthermore, the individual parameter variable may refer to the variable that determines how the individuals in the initial main population and auxiliary population are generated, thereby affecting the initial search space of the algorithm.
[0063] In a possible implementation, the first objective function is used as the optimization target of the main population, and the second objective function is used as the optimization target of the auxiliary population, including:
[0064] The first objective function is determined based on the minimum distance information in the distance information subset of the optimized parameter variable, and the second objective function is determined based on the maximum distribution area of the selected stop points in the decision space of the optimized parameter variable;
[0065] The first objective function is used as the optimization target of the main population so that the main population can focus on the convergence of the target optimization parameter variable set during the iteration process; the second objective function is used as the optimization target of the auxiliary population so that the auxiliary population can focus on the diversity of the target optimization parameter variable set of the main population during the iteration process.
[0066] Here, convergence can refer to the tendency of individuals in a population to gradually move towards the optimal solution. Convergence can characterize how quickly an evolutionary algorithm can find n optimal or satisfactory solutions. Diversity can refer to the degree of diversity of individuals in the main population, that is, the degree to which individuals in the population are distributed in the decision space, and the degree of difference in the attributes of individuals.
[0067] Among them, focusing on the convergence of the target optimization parameter variable set of the main population can reduce the distance from the candidate stop to the target position, and focusing on the diversity of the target optimization parameter variable set of the main population can improve the uniformity of the distribution of the optional candidate stop and avoid local optimal solutions.
[0068] Here, by taking into account the balance between convergence and diversity, while determining the nearest candidate stop points to multiple target locations, the distribution diversity and uniformity of the determined candidate stop points can be improved.
[0069] Furthermore, the first objective function may include but is not limited to: 1 (x), f 2 (x), f 3 (x) and f 4 (x); where f 1 (x) can represent the shortest distance to the primary school, f 2 (x) can represent the shortest distance to junior high school, f 3 (x) can represent the shortest distance to the convenience store, and f 4 (x) can represent the shortest distance to the train station.
[0070] In a possible implementation, during the iterative processing, cross-optimizing the optimization parameter variable set of the main population and the optimization parameter variable set of the auxiliary population may include:
[0071] During the iteration process, the optimization parameter variable set of the auxiliary population is used to exchange the optimization parameter variable set to influence the generation of the optimization parameter variable set of the main population; the optimization parameter variable set of the main population is used to exchange the optimization parameter variable set to influence the generation of the optimization parameter variable set of the auxiliary population.
[0072] Here, the solutions with rich diversity in the auxiliary population can be introduced into the main population to increase the search area of the main population to avoid the main population converging to the local optimal solution too early; the solutions with strong convergence in the main population can be introduced into the auxiliary population to improve the quality of the solutions in the auxiliary population.
[0073] In a possible implementation, determining the target optimization parameter variable set based on the optimization parameter variable set of the main population may also include:
[0074] Based on the optimization parameter variable set of the main population and the optimization parameter variable set of the auxiliary population, a target optimization parameter variable set is determined.
[0075] Here, the generation of the main population may be affected based on the optimized parameter variable set of the auxiliary population during the last iteration.
[0076] Through the above method, a multi-modal and multi-objective evolutionary algorithm is adopted to determine at least one optimal stop point based on iterative processing of the main population and the auxiliary population, which can meet the minimum distance to multiple target locations, thereby improving the diversity and rationality of the stop point selection of the autonomous driving vehicle and reducing the problem of increased fuel consumption caused by unnecessary driving of the vehicle.
[0077] In a specific embodiment, for an individual stop point in the set of stop points to be selected for the vehicle, distance information between the individual stop point and multiple target locations is obtained to obtain an individual distance information subset, which is determined based on the following steps:
[0078] Using a preset evaluation function, determine multiple target locations in the target location set that are closest to the individual candidate stop points;
[0079] The distance information between the individual candidate stop point and multiple target locations is obtained to obtain an individual distance information subset.
[0080] In a possible implementation, a preset evaluation function is used to determine multiple target positions in the target position set that are closest to the individual candidate stop points, based on the following steps:
[0081] The preset evaluation function is used to calculate the distance information between the individual candidate stop point and each target position in the target position set, and multiple target positions closest to the individual candidate stop point are determined.
[0082] The preset evaluation function can be used to calculate the distance information between the individual candidate stop point and the target location. The preset evaluation function here can refer to the Euclidean distance. The target location set can include Figure 2 Primary schools, junior high schools, convenience stores and train stations in the area.
[0083] Exemplarily, the distance between an individual candidate stop and each elementary school, junior high school, convenience store and railway station in the target location set is calculated, and the elementary school, junior high school, convenience store and railway station closest to the individual candidate stop are determined as the individual distance information subset.
[0084] Through the above method, only the target position closest to the individual candidate stop point is calculated, which can reduce the calculation complexity and improve the processing speed of stop point confirmation.
[0085] In a specific embodiment, the iterative processing of the primary population and the secondary population is implemented based on the following steps:
[0086] The N first to-be-selected stop points in the main population are randomly initialized so that the N first to-be-selected stop points are randomly distributed in the decision space, and the N second to-be-selected stop points in the auxiliary population are randomly initialized so that the N second to-be-selected stop points are randomly distributed in the decision space;
[0087] Based on the first selection criterion determined by the first objective function, a binary tournament selection strategy is used to select N first parents from the N first candidate stops in the primary population; based on the second selection criterion determined by the second objective function, a binary tournament selection strategy is used to select N second parents from the N second candidate stops in the secondary population;
[0088] Determine the N first offsprings generated by the N first parents and the N second offsprings generated by the N second parents, and determine the main population and the auxiliary population after iterative processing according to the N first parents, the N first offsprings, the N second parents and the N second offsprings; wherein the offsprings are generated by the parents through crossover and mutation operations.
[0089] In a possible implementation, a preset number of points may be selected in the decision space represented by the physical map as a set of to-be-selected stop points.
[0090] Among them, the size of the decision space is a rectangular area of [0, 100]*[0, 100], the optional range of the coordinates (x1, x2) of the stop points in the set of stop points to be selected can be x1=(0.00, 0.05, 0.10,…, 100.00), x2=(0.00, 0.05, 0.10,…, 100.00), and the preset number can be 2001*2001.
[0091] In a possible implementation, N first to-be-selected stop points in the main population are randomly initialized so that the N first to-be-selected stop points are randomly distributed in the decision space, and N second to-be-selected stop points in the auxiliary population are randomly initialized so that the N second to-be-selected stop points are randomly distributed in the decision space, including:
[0092] N first to-be-selected stop points are randomly generated in the decision space, and the N first to-be-selected stop points are randomly distributed in the decision space. N second to-be-selected stop points are randomly generated in the decision space, and the N second to-be-selected stop points are randomly distributed in the decision space; wherein the first to-be-selected stop points and the second to-be-selected stop points represent different potential stop point selection schemes.
[0093] Here, by determining N first candidate stop points and N second candidate stop points through random initialization, it can be ensured that the algorithm is adaptable and robust to different problem instances.
[0094] In one possible implementation, based on the first selection criterion determined by the first objective function, a binary tournament selection strategy is used to select N first parents from N first candidate stops in the main population, and based on the second selection criterion determined by the second objective function, a binary tournament selection strategy is used to select N second parents from N second candidate stops in the auxiliary population based on the following steps:
[0095] Determining a first selection criterion adopted by the main population based on the first objective function;
[0096] The binary tournament selection strategy under the first selection standard is adopted to screen the N first candidate stops in the main population and determine the N first candidate stops as the first parents of the main population;
[0097] Determining a second selection criterion adopted by the auxiliary population based on the second objective function;
[0098] The binary tournament selection strategy under the second selection criterion is adopted to screen the N second candidate stops in the auxiliary population, and determine the N second candidate stops as the second parents of the auxiliary population.
[0099] Here, the first selection criterion can be used to focus on the convergence of the N first candidate stops in the main population. For example, the first selection criterion can be the distance information from the first candidate stop to the target location; the second selection criterion can be used to focus on the diversity of the N second candidate stops in the auxiliary population. For example, the second selection criterion can be the distribution of the second candidate stops in the decision space.
[0100] Furthermore, the binary tournament selection strategy can refer to simulating the concept of a tournament in nature, randomly selecting two individuals for each selection for comparison, and selecting the individual with better performance as a candidate for the next generation.
[0101] For example, two individuals A and B are randomly selected from the main population, and the fitness of A and B under the first selection criterion are compared. If the fitness of A is better than the fitness of B, A is selected as the first parent. Among them, fitness can refer to a quantitative indicator of the ability of an individual to survive and reproduce in a specific environment. The greater the fitness, the more likely it is to be transmitted to the next generation.
[0102] In a possible implementation, determining N first children generated by N first parent generations and N second children generated by N second parent generations includes:
[0103] The first parent generation of the main population generates N first offspring by crossover and mutation, and the second parent generation of the auxiliary population generates N second offspring by crossover and mutation.
[0104] Here, crossover may include but is not limited to single-point crossover, multi-point crossover, uniform crossover and partial mapping crossover; wherein, single-point crossover may refer to selecting a cutting point between two parents and exchanging the data content after the cutting point. For example, the coordinates of stop point A are (1,8), and the coordinates of stop point B are (2,10). The offspring generated after the two parents cross over are (1,10) and (2,8); mutation may refer to random changes in the coordinates of the stop points. Mutation may include but is not limited to: bit mutation, exchange mutation, insertion mutation, inversion mutation and Gaussian mutation.
[0105] In a possible implementation, determining the main population and the auxiliary population after iterative processing according to the N first parents, the N first children, the N second parents, and the N second children may include one of the following:
[0106] According to the N first parents, the N first children and the N second parents, the main population and the auxiliary population after iterative processing are determined;
[0107] According to the N first parent generations, the N first offspring generations and the N second offspring generations, a main population and an auxiliary population after iterative processing are determined;
[0108] According to the N second parent generations, the N second child generations and the N first parent generations, the main population and the auxiliary population after iterative processing are determined;
[0109] According to the N second parent generations, the N second offspring generations and the N first offspring generations, a main population and an auxiliary population after iterative processing are determined.
[0110] Through the above method, iterative evolution based on the dual population structure can maintain population diversity, avoid falling into local optimality, and prevent the main population from converging too quickly; by using the binary tournament strategy when selecting the parent generation, better individuals can be effectively selected for crossover and mutation to increase the fitness of the offspring; under the influence of the auxiliary population, the main population can bring about a faster convergence speed based on additional diverse individuals.
[0111] In a specific embodiment, based on a first selection criterion determined by a first objective function, a binary tournament selection strategy is used to select N first parents from N first candidate stops in a primary population, and based on a second selection criterion determined by a second objective function, a binary tournament selection strategy is used to select N second parents from N second candidate stops in a secondary population, including:
[0112] Randomly select two first candidate stop points from the N first candidate stop points in the main population, and determine the candidate stop point with a smaller value among the two first candidate stop points as the first parent based on a first selection criterion; the first selection criterion is used to characterize the distance information from the first candidate stop points to the multiple target locations respectively;
[0113] Two second candidate stops are randomly selected from the N second candidate stops in the auxiliary population, and the candidate stop with a larger value among the two second candidate stops is determined as the second parent based on a second selection criterion; the second selection criterion is used to characterize the crowding distance between the second candidate stops.
[0114] In a possible implementation, two first candidate stop points are randomly selected from the N first candidate stop points in the main population, and the candidate stop point with a smaller value among the two first candidate stop points is determined as the first parent based on the first selection criterion, which is implemented based on the following steps:
[0115] Randomly select two first candidate stop points from the N first candidate stop points of the main population, calculate the distance information from the two first candidate stop points to multiple target locations, and select the first candidate stop point with a closer distance between the two first candidate stop points as the individual in the first parent generation;
[0116] Randomly select two second candidate stop points from the N second candidate stop points of the auxiliary population, calculate the crowding distances between the two second candidate stop points and other surrounding second candidate stop points, and select the second candidate stop point with a larger crowding distance among the two second candidate stop points as the individual in the second parent generation.
[0117] The crowding distance may refer to a dimensionless distance metric used to estimate the distribution of each individual and its neighboring individuals in the target space. The larger the crowding distance, the smaller the density of the individual in the decision space.
[0118] Here, based on the second selection criterion, the candidate stop point with a larger value among the two second candidate stop points is determined as the second parent, which can increase the diversity of individuals in the auxiliary population and make the candidate stop points in the auxiliary population have a larger distribution area in the decision space.
[0119] Furthermore, when two first candidate stop points are randomly selected from the N first candidate stop points of the main population, a smaller value or a closer distance may mean smaller or closer in a relative sense; for example, the distance between stop point A and the four target positions may be [3, 4, 8, 1], and the distance between stop point B and the four target positions may be [4, 8, 2, 1]. At this time, among the four position information of stop point A, 3<4 and 4<8 and 1=1, but 8>4, which can indicate that the distances from stop point A to the first position and the second position are respectively smaller than those of point B, and the distance from stop point A to the fourth position is equal to that of stop point B, but the distance from stop point A to the third position is larger than that of stop point B. At this time, there is no situation where stop points A and B have smaller values or closer distances, and AB are on par with each other; at this time, one can be randomly selected between stop points A and B as an individual in the first parent generation.
[0120] Through the above method, by adopting a dual population structure of a main population and an auxiliary population, combined with different objective functions and selection strategies, the ability to solve multimodal and multi-objective optimization problems can be significantly improved, avoiding premature convergence of the population due to lack of diversity during the search process.
[0121] In a specific embodiment, determining N first children generated by N first parent generations and N second children generated by N second parent generations, and determining a main population and an auxiliary population after iterative processing according to the N first parent generations, the N first children, the N second parent generations, and the N second children, includes:
[0122] Merge N first parents, N first children and N second parents to obtain a first population containing 3N candidate stops, and make the first population enter the environmental selection stage. According to the first selection standard, the 3N candidate stops in the first population are evaluated respectively to determine the first fitness corresponding to the 3N candidate stops respectively; the environmental selection stage is used to determine the individuals retained in the population; the first fitness is used to characterize the performance score of the 3N candidate stops corresponding to the first selection standard;
[0123] Sort the 3N candidate stop points based on the first fitness, obtain the top N candidate stop points as the first candidate stop points, and determine the main population after iteration;
[0124] Merge N second parents, N second children and N first parents to obtain a second population including 3N to-be-selected stops, and make the second population enter the environmental selection stage. According to the second selection criteria, the 3N to-be-selected stops in the second population are evaluated respectively to determine the second fitness corresponding to the 3N to-be-selected stops respectively; the second fitness is used to characterize the congestion degree corresponding to the 3N to-be-selected stops respectively;
[0125] The 3N candidate stop points are sorted based on the second fitness, the top N candidate stop points are obtained as the first candidate stop points, and the secondary population after iteration is determined.
[0126] Specifically, based on the first selection criterion, the first fitness corresponding to the 3N candidate stop points in the first population is determined, the 3N candidate stop points are sorted based on the first fitness, and the first N candidate stop points are obtained as the main population after iteration; based on the second selection criterion, the second fitness corresponding to the 3N candidate stop points in the second population is determined, the 3N candidate stop points are sorted based on the second fitness, and the first N candidate stop points are obtained as the auxiliary population after iteration.
[0127] The degree of crowding represented by the second fitness can be determined by a small habitat, and the second fitness is inversely proportional to the degree of crowding; the small habitat can refer to a diversity maintenance technology in the field of evolutionary algorithms.
[0128] The definition of the congestion degree of the waiting stop point in the group Pop can be shown as follows:
[0129]
[0130]
[0131] Among them, d(a,b) represents the Euclidean distance between individuals a and b, r can be the radius of the niche. When d(a,b)≤r, it means that the two individuals are in the niche, and the crowding degree increases. When d(a,b)>r, it means that the two individuals are not in the same niche, which does not affect the crowding degree. The niche radius r can be the average value of the distance between each individual and the nearest third vehicle.
[0132] Here, the range of crowding degree is [0,1]. When the crowding degree is smaller, the diversity of individuals is higher. When the crowding degree is 0, it means that there are no other individuals within the ecological niche of the individual.
[0133] In a specific embodiment, 3N candidate stop points are sorted based on the first fitness, the top N candidate stop points are obtained as the first candidate stop points, and the main population after iteration is determined, including:
[0134] A non-dominated sorting method is used to determine non-dominated solutions among 3N candidate stop points. When the number of non-dominated solutions is less than N, the dominated solutions among the 3N candidate stop points are sorted based on the first fitness. According to the sorting result, a first preset number of dominated solutions and the first candidate stop point corresponding to the non-dominated solutions are selected as the main population after iteration; the first preset number is the difference between N and the number of non-dominated solutions; the non-dominated solution is used to represent that it is not worse than other solutions in at least one objective, and the dominated solution is used to represent the solution dominated by the non-dominated solution;
[0135] When the number of non-dominated solutions is greater than or equal to N, the non-dominated solutions are sorted based on the first fitness, and the first candidate stop points corresponding to the top N non-dominated solutions are obtained as the main population after iteration;
[0136] The 3N candidate stop points are sorted based on the second fitness, the top N candidate stop points are obtained as the first candidate stop points, and the secondary population after iteration is determined, including:
[0137] Based on the second fitness, the 3N to-be-selected stop points are sorted in descending order, the to-be-selected stop point with the smallest second fitness is removed, and the sorting result and the number of the to-be-selected stop points are updated;
[0138] When the number of the updated to-be-selected stop points is equal to N, N to-be-selected stop points are obtained as the secondary population after iteration.
[0139] In a possible implementation, a non-dominated sorting method is used to determine non-dominated solutions among 3N candidate stop points. When the number of non-dominated solutions is less than N, the dominated solutions among the 3N candidate stop points are sorted based on the first fitness. According to the sorting result, a first preset number of dominated solutions and the first candidate stop point corresponding to the non-dominated solution are selected as the main population after iteration, which is implemented based on the following steps:
[0140] In the environmental selection stage, the non-dominated sorting method is used to determine the non-dominated solutions among the 3N candidate stops, and the number of non-dominated solutions is determined to be M. The M non-dominated solutions are used as individuals in the main population;
[0141] When M<N, the other dominant solutions among the 3N candidate stop points are sorted according to the numerical values of the first fitness, and the first NM dominant solutions in the sorting result are determined as the remaining individuals in the main population after iteration.
[0142] Here, by giving priority to non-dominated solutions in the main population, we can ensure that the selected stop points are not mutually superior in each optimization objective, thereby improving the overall quality of the main population; by combining non-dominated sorting and fitness sorting, we can ensure the diversity of individuals in the population, avoid premature convergence of the algorithm, and improve the diversity of the main population.
[0143] In a possible implementation, when the number of non-dominated solutions is greater than or equal to N, the non-dominated solutions are sorted based on the first fitness, and the first candidate stop points corresponding to the top N non-dominated solutions are obtained as the main population after iteration, including:
[0144] When the number of non-dominated solutions M ≥ N, determine the first fitness corresponding to each of the M non-dominated solutions, sort them according to the numerical value of the first fitness, and determine the first N non-dominated solutions in the sorting result as the individuals of the main population after iteration.
[0145] Here, by selecting non-dominated solutions with better fitness, the main population can be guided to evolve towards a better decision space direction, making the algorithm more directional in the optimization process and avoiding unnecessary searches; inferior solutions can be filtered out from the non-dominated solutions, making the individuals in the main population closer to the global optimum on the first objective function, thereby accelerating the overall convergence.
[0146] In a possible implementation, 3N candidate stop points are sorted in descending order based on the second fitness, the candidate stop point with the smallest second fitness is removed, the sorting result and the number of candidate stop points are updated, and when the updated number of candidate stop points is equal to N, N candidate stop points are obtained as the secondary population after iteration, which is implemented based on the following steps:
[0147] Obtain the congestion levels corresponding to the 3N to-be-selected stop points in the second population, and determine the second fitness levels corresponding to the 3N to-be-selected stop points respectively based on the congestion levels;
[0148] Based on the second fitness, remove the candidate stop point with the lowest second fitness, update the congestion level of the remaining candidate stop points and the corresponding second fitness, and remove the candidate stop point with the lowest second fitness in the current second population again, until the number of candidate stop points is equal to N, and obtain the remaining N candidate stop points as the iterated auxiliary population.
[0149] Here, by removing the most crowded candidate stops, the distribution diversity of the candidate stops in the auxiliary population is maintained to avoid the candidate stops being concentrated in a specific area; by continuously removing the most crowded candidate stops, the algorithm can conduct more extensive exploration in the entire decision space and increase the probability of finding the global optimal solution.
[0150] Through the above method, the degree of crowding of individuals is determined based on the microhabitat, and then the second fitness of the auxiliary population is determined. The iterated auxiliary population is determined based on the second fitness, which provides diversity support for the iteration of the main population, prevents the main population from converging to the local optimal solution too early, and increases the main population's ability to search for the global optimal solution.
[0151] In a specific embodiment, the method further comprises:
[0152] Obtain the optimization parameter variable set of the main population that meets the iteration stop condition, determine the number of to-be-selected stop points in the optimization parameter variable set, and when the number of to-be-selected stop points is greater than a preset number threshold, based on a preset population maintenance standard, eliminate the to-be-selected stop points that do not meet the preset population maintenance standard in the optimization parameter variable set; the preset number threshold is determined based on the iteration stop condition.
[0153] Specifically, all optimization parameter variable sets of the current main population are obtained, and the optimization parameter variable set represents all possible stop point locations currently determined; the number of to-be-selected stop points in the optimization parameter variable set is determined, and when the number is greater than a preset number threshold, unqualified to-be-selected stop points are eliminated based on a preset population maintenance standard.
[0154] Among them, the iteration stopping conditions may include but are not limited to: reaching the maximum number of iterations, reaching the expected fitness value, and reaching the number of individuals in the preset population; the preset population maintenance standards may include but are not limited to: fitness, diversity, and stability.
[0155] Furthermore, the preset quantity threshold can be determined according to the preset number of evaluations in the iteration stop condition. For example, the preset number of evaluations represents the expected number of 6000 evaluations, and the preset number of iterations is 30. Then, the maximum number of candidate stop points determined by the main population each time represented by the preset number threshold can be 6000 / 30=200; here, the number of evaluations can refer to the number of individuals evaluated by the objective function.
[0156] Furthermore, a target optimization parameter variable set is determined based on the optimization parameter variable set composed of the candidate stop points determined by the final main population, and a corresponding number of equivalent optimal stop points are determined according to the candidate stop points in the target optimization parameter variable set, and the distance between each optimal stop point and the four target positions is the minimum.
[0157] In one possible implementation, based on the location information corresponding to the multiple optimal stopping points determined, the vehicle is allocated to the area corresponding to the optimal stopping point; the vehicle reservation order determined by the passenger based on his or her own preferences is obtained, and based on the passenger's vehicle reservation order, the corresponding type of autonomous driving vehicle is determined so that the vehicle can reach the passenger's location in the shortest time and the vehicle meets the requirements in the passenger's vehicle reservation order.
[0158] Here, the vehicle reservation order determined by the passenger based on his / her own preferences may include: vehicle color, vehicle brand, and vehicle interior space.
[0159] In a possible implementation, a preset Pareto area corresponding to the physical map may be used as a judgment criterion to compare and determine whether the selection of the candidate stop points in the target optimization parameter variable set is reasonable.
[0160] For example, Figure 4 As shown, Figure 4 It is a Pareto region schematic diagram of an automatic driving stop point selection method based on a multi-modal multi-objective evolutionary algorithm provided by an embodiment of the present invention;
[0161] The Pareto region here can be identified and generated based on historical data using a multimodal and multi-objective optimization algorithm.
[0162] In a possible implementation, the method further includes:
[0163] Adding at least one constraint condition in the target space, the constraint condition including but not limited to: minimum passenger waiting time, minimum vehicle energy consumption, and maximum passenger satisfaction;
[0164] Based on different situations, the weight value of the first objective function is adjusted according to the constraints.
[0165] For example, during rush hour, the weight of minimum passenger waiting time can be increased; when energy is scarce, the weight of minimum vehicle energy consumption can be increased.
[0166] Here, in actual applications, the operating company of autonomous driving vehicles can adjust the weight of maximum passenger satisfaction based on the specific needs of passengers, so that passengers can book their favorite vehicles according to their preferences, thereby improving the applicability of application scenarios; they can also quickly switch to other vehicles to provide services to passengers in the event of a breakdown or other emergency based on specific road and traffic conditions.
[0167] Through the above method, introducing individuals from the auxiliary population in the generation process of the main population after iteration can increase the diversity of individuals in the main population and avoid the main population from converging to the local optimal solution too early. Introducing individuals from the main population in the generation process of the auxiliary population after iteration can increase the convergence of individuals in the auxiliary population and improve the iteration quality of the auxiliary population.
[0168] Through the above method, the weight of the objective function can be adjusted according to different traffic conditions and parking needs to cope with different situations; when a vehicle breaks down, it can quickly switch to a spare vehicle or call other vehicles to increase fault tolerance when picking up and dropping off passengers; it can provide passengers with personalized vehicle selection recommendations to meet the personalized needs of passengers; individuals that do not meet the preset population maintenance standards can be removed to avoid their negative impact on the quality of the final solution, thereby improving the accuracy and effectiveness of the selection of candidate stops, thereby reducing the driving distance of the vehicle to each target location.
[0169] Figure 5 is a specific flow chart of a method for selecting an automatic driving stop point based on a multi-modal multi-objective evolutionary algorithm provided by an embodiment of the present invention, such as Figure 5 As shown, the figure includes the following steps:
[0170] Step S501, randomly initializing individual stop points in the set of stop points to be selected in the map;
[0171] Here, it is equivalent to dividing 2001*2001 waiting stops in the physical map as a waiting stop point set;
[0172] Step S502, dividing into two populations with N to-be-selected stop points;
[0173] Here, two populations with N waiting stops are randomly divided, one as the main population and the other as the auxiliary population;
[0174] Step S503, the main population determines the first parent generation based on the first selection criterion;
[0175] Here, N first parents are selected from N individuals in the main population based on the binary tournament strategy under the first selection criterion;
[0176] Step S504, the auxiliary population determines a second parent generation based on a second selection criterion;
[0177] Here, N second parents are selected from the N individuals in the auxiliary population based on the binary tournament strategy under the second selection criterion;
[0178] Step S505, determining that the first population enters the environmental selection phase, and generating an iterated main population;
[0179] Here, the first population is generated, and the first fitness of 3N individuals in the first population is determined based on the first selection criterion in the environmental selection stage, and N individuals are selected to generate the main population after iteration;
[0180] Step S506, determining that the second population enters the environmental selection phase, and generating an iterated auxiliary population;
[0181] Here, a second population is generated, and in the environmental selection phase, the second fitness of 3N individuals in the second population is determined based on the second selection criterion, and N individuals are selected to generate the iterated auxiliary population;
[0182] Step S507, outputting the target optimization parameter variable set of the main population;
[0183] Here, the target optimization parameter variable set of the main population that satisfies the iteration stop condition is determined, and the to-be-selected stop point in the target optimization parameter variable set is taken as the optimal stop point;
[0184] Step S508, allocating vehicles to the selected stop points in the target optimization parameter variable set;
[0185] Here, based on the position information corresponding to the determined multiple optimal stop points, the vehicle is assigned to the area corresponding to the optimal stop point;
[0186] Step S509, obtaining the vehicle reservation order determined by the passenger based on his / her own needs;
[0187] Here, the vehicle reservation order determined by the passenger based on his / her own preferences is obtained;
[0188] Step S510, assigning a corresponding vehicle to the passenger's location;
[0189] Here, based on the passenger's vehicle reservation order, the corresponding type of autonomous driving vehicle is determined so that the vehicle can reach the passenger's location in the shortest time.
[0190] In one possible implementation, the hypervolume (HV) indicator (or HV value) can be used to measure the convergence and diversity of the solution set of the multimodal multi-objective evolutionary algorithm. The HV indicator can reflect the volume of the target space covered by the solution set. The larger the value of the HV indicator, the larger the area covered by the solution set in the target space, that is, the better the quality of the solution set.
[0191] Specifically, the average HV value corresponding to the solution set obtained by the multi-modal multi-objective evolutionary algorithm in this specific embodiment is compared with the average HV values of other algorithms, as shown in Table 1:
[0192]
[0193] Table 1
[0194] Among them, the HV value of Diversity-enhanced Non-dominated Sorting Genetic Algorithm II (DN-NSGA-II) is 4.5797e-1, and the standard deviation is 9.48e-3; the HV value of Multimodal Multi-Objective Evolutionary Algorithm with Weight-based Initialization (MMEA-WI) is 4.4255e-1, and the standard deviation is 8.05e-3; the HV value of Adaptive Differential Evolution withDynamic Diversity Strategy (TADE_DDS) is 4.5673e-1, and the standard deviation is 7.41e-3; the HV value of Multimodal Multi-Objective Evolutionary Algorithm with Dual Clustering, MMOEA / DC) has an HV value of 4.3853e-1, and a standard deviation of 1.76e-2; the HV value of MMOGA (Multimodal Multi-Objective Evolutionary Algorithm) is 4.3853e-1, and a standard deviation of 1.76e-2; the HV value of an embodiment of the present invention is 4.5828e-1, and the standard deviation is 6.51e-3.
[0195] It can be seen that the HV index value of the embodiment of the present invention is the largest, indicating that the embodiment of the present invention has found a better and more evenly distributed solution set in the target space for the problem of selecting a stop point for an autonomous driving vehicle; the standard deviation of the embodiment of the present invention is the smallest, indicating that the solution set has less fluctuation in multiple experiments and has better stability.
[0196] Exemplarily, the positions of the candidate stops determined by the above-mentioned multi-modal multi-objective optimization algorithms in the problem of selecting a stop point for autonomous driving are shown in FIG6 , which is a schematic diagram of the stop point selection of various algorithms of an autonomous driving stop point selection method based on a multi-modal multi-objective evolutionary algorithm provided by an embodiment of the present invention, wherein:
[0197] Figure 6(a) shows DN-NSGA-II, Figure 6(b) shows MMEA-WI, Figure 6(c) shows TADE_DDS, Figure 6(d) shows MMOEA / DC, Figure 6(e) shows MMOGA, and Figure 6(f) shows the embodiment of the present invention; all the dots in Figure 6 represent the selected stop positions. In terms of distribution uniformity, in other algorithms such as DN-NSGA-II, the stop positions found in some areas are relatively concentrated, while in some areas they are relatively sparse. For example, the stop positions found by TADE_DDS are all relatively sparse, while the stop positions in the embodiment of the present invention are relatively more evenly distributed; in terms of coverage, in other algorithms such as MMOGA, there is a relatively large available docking area left in the middle, while the stop positions of the present invention can cover the available area more fully.
[0198] It can be seen from the comparison results in Figure 6 that the stop positions determined by the stop position selection method in the embodiment of the present invention are evenly distributed, have a wide coverage range, and have good objective space convergence and diversity. They can distribute vehicles more reasonably, reduce vehicle congestion in some areas, and at the same time make full use of all available docking areas, improve the coverage range and response speed of services, and can provide a more convenient and efficient travel experience for passengers, and also help to improve the overall operation efficiency and service quality of the autonomous driving system.
[0199] In Figure 6, the solutions obtained around three points corresponding to the position coordinates (x1, x2) in the range of 30 < x1 < 52 and x2 = 60 in the map are selected. Among them, the rhombus represents a convenience store, the triangle represents a primary school, and the pentagram represents a railway station, as shown in Figure 7. Figure 7 is a partial schematic diagram of the stop position selection of multiple algorithms for an autonomous driving stop position selection method based on a multi-modal multi-objective evolutionary algorithm provided by the embodiment of the present invention. Among them,
[0200] Figure 7(a) shows DN-NSGA-II, Figure 7(b) shows MMEA-WI, Figure 7(c) shows TADE_DDS, Figure 7(d) shows MMOEA / DC, Figure 7(e) shows MMOGA, and Figure 7(f) shows the embodiment of the present invention; as shown in Figure 7(e), there is no solution for the stop position on the rhombus representing the convenience store in the MMOGA algorithm, while there are non-dominated solutions on the rhombus in other algorithms. The MMOGA algorithm does not have an extreme value solution corresponding to the convenience store optimization target, so the HV index of the MMOGA algorithm in Table 1 is relatively low; as shown in Figure 7(c), there is no solution for the stop position around the triangle representing the primary school in TADE_DDS, and TADE_DDS does not have an extreme value solution corresponding to the primary school optimization target; while there are non-dominated solutions around all three points in the stop position selection in the embodiment of the present invention.
[0201] Figure 8 It is a schematic structural diagram of an autonomous driving stop position selection device based on a multi-modal multi-objective evolutionary algorithm provided by the embodiment of the present invention, as Figure 8As shown, the device can be applied to intelligent electronic devices such as servers and computers; the device includes: a first determination module 801, an algorithm iteration module 802, and a second determination module 803;
[0202] The first determining module 801 is used to obtain distance information between an individual to-be-selected stop point in the to-be-selected stop point set of the vehicle and a plurality of target locations, and obtain an individual distance information subset;
[0203] The algorithm iteration module 802 is used to utilize a multi-modal multi-objective evolutionary algorithm, take the individual candidate stop points in the candidate stop point set and the distance information subset of the corresponding individual candidate stop points as individual parameter variables of the main population and the auxiliary population, take the first objective function as the optimization target of the main population, take the second objective function as the optimization target of the auxiliary population, perform iterative processing on the main population and the auxiliary population, and in the iterative processing process, perform cross-optimization on the optimization parameter variable set of the main population and the optimization parameter variable set of the auxiliary population;
[0204] The second determination module 803 is used to determine the target optimization parameter variable set based on the optimization parameter variable set of the main population when the iterative processing meets the iteration stop condition, and determine at least one optimal stop point according to the candidate stop points in the target optimization parameter variable set; wherein the first objective function is determined based on the minimum distance information in the distance information subset of the optimization parameter variables; and the second objective function is determined based on the maximum distribution area of the candidate stop points of the optimization parameter variables in the decision space.
[0205] In a specific embodiment, the first determination module 801 is used to determine, by using a preset evaluation function, a plurality of target locations in the target location set that are closest to the individual candidate stop points;
[0206] The distance information between the individual candidate stop point and multiple target locations is obtained to obtain an individual distance information subset.
[0207] In a specific embodiment, the algorithm iteration module 802 is used to randomly generate N first to-be-selected stop points in the main population based on the individual parameter variables of the main population, and to randomly generate N second to-be-selected stop points in the auxiliary population based on the individual parameter variables of the auxiliary population;
[0208] The N first to-be-selected stop points in the main population are randomly initialized so that the N first to-be-selected stop points are randomly distributed in the decision space, and the N second to-be-selected stop points in the auxiliary population are randomly initialized so that the N second to-be-selected stop points are randomly distributed in the decision space;
[0209] Based on the first selection criterion determined by the first objective function, a binary tournament selection strategy is used to select N first parents from the N first candidate stops in the primary population; based on the second selection criterion determined by the second objective function, a binary tournament selection strategy is used to select N second parents from the N second candidate stops in the secondary population;
[0210] Determine the N first offsprings generated by the N first parents and the N second offsprings generated by the N second parents, and determine the main population and the auxiliary population after iterative processing according to the N first parents, the N first offsprings, the N second parents and the N second offsprings; wherein the offsprings are generated by the parents through crossover and mutation operations.
[0211] In a specific embodiment, the algorithm iteration module 802 is used to randomly select two first candidate stop points from the N first candidate stop points in the main population, and determine the candidate stop point with a smaller value among the two first candidate stop points as the first parent based on a first selection criterion; the first selection criterion is used to characterize the distance information from the first candidate stop points to the multiple target locations respectively;
[0212] Two second candidate stops are randomly selected from the N second candidate stops in the auxiliary population, and the candidate stop with a larger value among the two second candidate stops is determined as the second parent based on a second selection criterion; the second selection criterion is used to characterize the crowding distance between the second candidate stops.
[0213] In a specific embodiment, the algorithm iteration module 802 is used to merge N first parents, N first children and N second parents to obtain a first population including 3N candidate stop points, so that the first population enters the environmental selection stage, and the 3N candidate stop points in the first population are evaluated according to the first selection standard to determine the first fitness corresponding to the 3N candidate stop points; the environmental selection stage is used to determine the individuals retained in the population; the first fitness is used to characterize the performance score of the 3N candidate stop points corresponding to the first selection standard;
[0214] Sort the 3N candidate stop points based on the first fitness, obtain the top N candidate stop points as the first candidate stop points, and determine the main population after iteration;
[0215] Merge N second parents, N second children and N first parents to obtain a second population including 3N to-be-selected stops, and make the second population enter the environmental selection stage. According to the second selection criteria, the 3N to-be-selected stops in the second population are evaluated respectively to determine the second fitness corresponding to the 3N to-be-selected stops respectively; the second fitness is used to characterize the congestion degree corresponding to the 3N to-be-selected stops respectively;
[0216] The 3N candidate stop points are sorted based on the second fitness, the top N candidate stop points are obtained as the first candidate stop points, and the secondary population after iteration is determined.
[0217] In a specific embodiment, the algorithm iteration module 802 is used to determine the non-dominated solutions among the 3N candidate stop points by using a non-dominated sorting method. When the number of non-dominated solutions is less than N, the dominated solutions among the 3N candidate stop points are sorted based on the first fitness. According to the sorting result, a first preset number of dominated solutions and the first candidate stop point corresponding to the non-dominated solution are selected as the main population after iteration; the first preset number is the difference between N and the number of non-dominated solutions; the non-dominated solution is used to represent that it is not worse than other solutions in at least one objective, and the dominated solution is used to represent the solution dominated by the non-dominated solution;
[0218] When the number of non-dominated solutions is greater than or equal to N, the non-dominated solutions are sorted based on the first fitness, and the first candidate stop points corresponding to the top N non-dominated solutions are obtained as the main population after iteration;
[0219] The 3N candidate stop points are sorted based on the second fitness, the top N candidate stop points are obtained as the first candidate stop points, and the secondary population after iteration is determined, including:
[0220] Based on the second fitness, the 3N to-be-selected stop points are sorted in descending order, the to-be-selected stop point with the smallest second fitness is removed, and the sorting result and the number of the to-be-selected stop points are updated;
[0221] When the number of the updated to-be-selected stop points is equal to N, N to-be-selected stop points are obtained as the secondary population after iteration.
[0222] In a specific embodiment, the device further includes: a elimination module 804; wherein,
[0223] The elimination module 804 is used to obtain the optimization parameter variable set of the main population when the iteration stop condition is met, determine the number of to-be-selected stop points in the optimization parameter variable set, and when the number of to-be-selected stop points is greater than a preset number threshold, based on the preset population maintenance standard, eliminate the to-be-selected stop points that do not meet the preset population maintenance standard in the optimization parameter variable set; the preset number threshold is determined based on the iteration stop condition.
[0224] It should be noted that: the automatic driving stop selection device based on multimodal multi-objective evolutionary algorithm provided in the above embodiment only uses the division of the above program modules as an example to illustrate when implementing the corresponding automatic driving stop selection method based on multimodal multi-objective evolutionary algorithm. In actual application, the above processing can be assigned to different program modules as needed, that is, the internal structure of the device can be divided into different program modules to complete all or part of the above-described processing. In addition, the device provided in the above embodiment and the corresponding Figure 1 The embodiments of the method shown belong to the same concept, and the specific implementation process is detailed in the method embodiments, which will not be repeated here.
[0225] The embodiment of the present invention also provides a computer device having the above Figure 8 The autonomous driving stop point selection device based on a multi-modal and multi-objective evolutionary algorithm is shown.
[0226] See also Fig. 9 , Fig. 9 is a schematic diagram of the structure of another automatic driving stop point selection device based on a multi-modal multi-objective evolutionary algorithm provided by an embodiment of the present invention, such as Fig. 9 As shown, the computer device includes: one or more processors 10, a memory 20, and interfaces for connecting various components, including high-speed interfaces and low-speed interfaces. Various components are connected to each other using different buses for communication, and can be installed on a common mainboard or installed in other ways as needed. The processor can process the instructions executed in the computer device, including instructions stored in or on the memory to display the graphical information of the GUI on an external input / output device (such as, a display device coupled to the interface). In some optional embodiments, if necessary, multiple processors and / or multiple buses can be used together with multiple memories and multiple memories. Similarly, multiple computer devices can be connected, and each device provides some necessary operations (for example, as a server array, a group of blade servers, or a multi-processor system). Fig. 9 A processor 10 is taken as an example.
[0227] The processor 10 may be a central processing unit, a network processor or a combination thereof. The processor 10 may further include a hardware chip. The hardware chip may be a dedicated integrated circuit, a programmable logic device or a combination thereof. The programmable logic device may be a complex programmable logic device, a field programmable gate array, a general purpose array logic or any combination thereof.
[0228] The memory 20 stores instructions executable by at least one processor 10, so that the at least one processor 10 executes the method shown in the above embodiment.
[0229] The memory 20 may include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required for at least one function; the data storage area may store data created according to the use of the computer device, etc. In addition, the memory 20 may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some optional embodiments, the memory 20 may optionally include a memory remotely arranged relative to the processor 10, and these remote memories may be connected to the computer device via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.
[0230] The memory 20 may include a volatile memory, such as a random access memory; the memory may also include a non-volatile memory, such as a flash memory, a hard disk or a solid state drive; the memory 20 may also include a combination of the above types of memory.
[0231] The computer device also includes an input device 30 and an output device 40. The processor 10, the memory 20, the input device 30 and the output device 40 may be connected via a bus or other means. Fig. 9 The example of connecting through bus is taken in the following.
[0232] The input device 30 can receive input digital or character information, and generate key signal input related to the user settings and function control of the computer device, such as a touch screen, a keypad, a mouse, a track pad, a touch pad, an indicator bar, one or more mouse buttons, a trackball, a joystick, etc. The output device 40 may include a display device, an auxiliary lighting device (e.g., an LED) and a tactile feedback device (e.g., a vibration motor), etc. The above-mentioned display device includes but is not limited to a liquid crystal display, a light emitting diode, a display and a plasma display. In some optional embodiments, the display device can be a touch screen.
[0233] The computer device further comprises a communication interface 30 for the computer device to communicate with other devices or a communication network.
[0234] The embodiment of the present invention also provides a computer-readable storage medium. The method according to the embodiment of the present invention can be implemented in hardware, firmware, or can be implemented as a computer code that can be recorded in a storage medium, or can be implemented as a computer code that is originally stored in a remote storage medium or a non-temporary machine-readable storage medium and will be stored in a local storage medium through a network download, so that the method described herein can be stored in such software processing on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. Among them, the storage medium can be a magnetic disk, an optical disk, a read-only storage memory, a random access memory, a flash memory, a hard disk or a solid-state hard disk, etc.; further, the storage medium can also include a combination of the above types of memories. It can be understood that a computer, a processor, a microprocessor controller, or programmable hardware includes a storage component that can store or receive software or computer code. When the software or computer code is accessed and executed by a computer, a processor, or hardware, the method shown in the above embodiment is implemented.
[0235] A part of the present invention may be applied as a computer program product, such as a computer program instruction, which, when executed by a computer, can call or provide the method and / or technical solution according to the present invention through the operation of the computer. Those skilled in the art should understand that the existence of the computer program instruction in a computer-readable medium includes, but is not limited to, a source file, an executable file, an installation package file, etc., and accordingly, the way in which the computer program instruction is executed by the computer includes, but is not limited to: the computer directly executes the instruction, or the computer compiles the instruction and then executes the corresponding compiled program, or the computer reads and executes the instruction, or the computer reads and installs the instruction and then executes the corresponding installed program. Here, the computer-readable medium may be any available computer-readable storage medium or communication medium accessible to the computer.
[0236] Although the embodiments of the present invention have been described in conjunction with the accompanying drawings, those skilled in the art may make various modifications and variations without departing from the spirit and scope of the present invention, and such modifications and variations are all within the scope defined by the appended claims.
Claims
1. A method for selecting stop points for autonomous driving based on a multi-modal and multi-objective evolutionary algorithm. It is characterized in that The method comprises: For an individual stop point in the set of stop points for the vehicle, distance information between the individual stop point and multiple target locations is obtained to obtain an individual distance information subset; Utilizing the multimodal multi-objective evolutionary algorithm, taking the individual candidate stop points in the candidate stop point set and the distance information subset corresponding to the individual candidate stop points as individual parameter variables of the main population and the auxiliary population, taking the first objective function as the optimization target of the main population, taking the second objective function as the optimization target of the auxiliary population, performing iterative processing on the main population and the auxiliary population, and during the iterative processing, cross-optimizing the optimized parameter variable set of the main population and the optimized parameter variable set of the auxiliary population; When the iterative processing meets the iteration stop condition, a target optimization parameter variable set is determined based on the optimization parameter variable set of the main population, and at least one optimal stopping point is determined based on the candidate stopping points in the target optimization parameter variable set; wherein the first objective function is determined based on the minimum distance information in the distance information subset of the optimization parameter variables; and the second objective function is determined based on the maximum distribution area of the candidate stopping points of the optimization parameter variables in the decision space.
2. The method according to claim 1, It is characterized in that For an individual stop point in the vehicle's stop point set, distance information between the individual stop point and multiple target locations is obtained to obtain an individual distance information subset, which is determined based on the following steps: Using a preset evaluation function, determine multiple target positions in the target position set that are closest to the individual candidate stop points; The distance information of the individual to-be-selected stop point and the multiple target locations is obtained to obtain the individual distance information subset.
3. The method according to claim 2, It is characterized in that The main population and the auxiliary population are iteratively processed based on the following steps: Based on the individual parameter variables of the main population, randomly generating N first to-be-selected stop points in the main population, and based on the individual parameter variables of the auxiliary population, randomly generating N second to-be-selected stop points in the auxiliary population; Performing random initialization processing on the N first to-be-selected stop points in the main population so that the N first to-be-selected stop points are randomly distributed in the decision space, and performing the random initialization processing on the N second to-be-selected stop points in the auxiliary population so that the N second to-be-selected stop points are randomly distributed in the decision space; Based on a first selection criterion determined by a first objective function, a binary tournament selection strategy is used to select N first parents from the N first waiting-for-selection stops in the primary population, and based on a second selection criterion determined by a second objective function, the binary tournament selection strategy is used to select N second parents from the N second waiting-for-selection stops in the secondary population; Determine the N first offspring generations generated by the N first parent generations and the N second offspring generations generated by the N second parent generations, and determine the main population and the auxiliary population after iterative processing according to the N first parent generations, the N first offspring generations, the N second parent generations and the N second offspring generations; wherein the offspring generations are generated by the parent generations through crossover and mutation operations.
4. The method according to claim 3, It is characterized in that The first selection criterion determined based on the first objective function, using a binary tournament selection strategy to select N first parents from the N first waiting stops in the main population, and the second selection criterion determined based on the second objective function, using the binary tournament selection strategy to select N second parents from the N second waiting stops in the auxiliary population, include: Randomly selecting two first to-be-selected stop points from the N first to-be-selected stop points in the main population, and determining the to-be-selected stop point with a smaller value among the two first to-be-selected stop points as the first parent based on the first selection criterion; the first selection criterion is used to characterize the distance information from the first to-be-selected stop points to a plurality of target locations respectively; Two second to-be-selected stop points are randomly selected from the N second to-be-selected stop points in the secondary population, and the stop point with a larger value among the two second to-be-selected stop points is determined as the second parent based on the second selection criterion; the second selection criterion is used to characterize the crowding distance between the second to-be-selected stop points.
5. The method according to claim 4, It is characterized in that The determining of the N first offsprings generated by the N first parent generations and the N second offsprings generated by the N second parent generations, and determining the main population and the auxiliary population after iterative processing according to the N first parent generations, the N first offsprings, the N second parent generations, and the N second offsprings, includes: The N first parents, the N first children and the N second parents are combined to obtain a first population including 3N candidate stop points, and the first population is put into an environmental selection stage. The 3N candidate stop points in the first population are evaluated according to a first selection criterion to determine first fitness values corresponding to the 3N candidate stop points. The environmental selection stage is used to determine the individuals retained in the population. The first fitness value is used to characterize the performance scores of the 3N candidate stop points corresponding to the first selection criterion. Sorting the 3N candidate stop points based on the first fitness, obtaining the top N candidate stop points as the first candidate stop points, and determining the main population after iteration; Merging the N second parents, the N second children and the N first parents to obtain a second population including 3N to-be-selected stops, allowing the second population to enter the environment selection stage, and evaluating the 3N to-be-selected stops in the second population according to a second selection criterion to determine second fitnesses corresponding to the 3N to-be-selected stops respectively; the second fitnesses are used to characterize the congestion levels corresponding to the 3N to-be-selected stops respectively; The 3N candidate stop points are sorted based on the second fitness, the top N candidate stop points are obtained as the first candidate stop points, and the secondary population after iteration is determined.
6. The method according to claim 5, It is characterized in that The step of sorting the 3N candidate stop points based on the first fitness, obtaining the top N candidate stop points as first candidate stop points, and determining the main population after iteration includes: A non-dominated sorting method is used to determine non-dominated solutions among the 3N candidate stop points. When the number of the non-dominated solutions is less than N, the dominated solutions among the 3N candidate stop points are sorted based on the first fitness. According to the sorting result, a first preset number of the dominated solutions and the first candidate stop point corresponding to the non-dominated solutions are selected as the main population after iteration; the first preset number is the difference between N and the number of the non-dominated solutions; the non-dominated solutions are used to represent solutions that are not worse than other solutions in at least one objective, and the dominated solutions are used to represent solutions dominated by non-dominated solutions; When the number of the non-dominated solutions is greater than or equal to N, the non-dominated solutions are sorted based on the first fitness, and the first candidate stop points corresponding to the top N non-dominated solutions are obtained as the main population after the iteration; The step of sorting the 3N candidate stop points based on the second fitness, obtaining the top N candidate stop points as the first candidate stop points, and determining the secondary population after iteration includes: Based on the second fitness, the 3N to-be-selected stop points are sorted in descending order, the to-be-selected stop point with the smallest second fitness is removed, and the sorting result and the number of the to-be-selected stop points are updated; When the updated number of the to-be-selected stop points is equal to N, the N to-be-selected stop points are obtained as the iterated auxiliary population.
7. The method according to claim 6, It is characterized in that The method further comprises: Obtain an optimized parameter variable set of the main population when the iteration stop condition is met, determine the number of to-be-selected stop points in the optimized parameter variable set, and when the number of to-be-selected stop points is greater than a preset number threshold, based on a preset population maintenance standard, eliminate the to-be-selected stop points in the optimized parameter variable set that do not meet the preset population maintenance standard; the preset number threshold is determined based on the iteration stop condition.
8. An automatic driving stop point selection device based on multi-modal multi-objective evolutionary algorithm, It is characterized in that The device comprises: A first determination module is used to obtain distance information between an individual to-be-selected stop point in the to-be-selected stop point set of the vehicle and a plurality of target locations, and obtain an individual distance information subset; an algorithm iteration module, for utilizing the multi-modal multi-objective evolutionary algorithm, taking individual candidate stop points in the candidate stop point set and a distance information subset corresponding to the individual candidate stop points as individual parameter variables of a main population and an auxiliary population, taking a first objective function as an optimization target of the main population, taking a second objective function as an optimization target of the auxiliary population, performing iterative processing on the main population and the auxiliary population, and during the iterative processing, performing cross-optimization on the optimized parameter variable set of the main population and the optimized parameter variable set of the auxiliary population; The second determination module is used to determine the target optimization parameter variable set based on the optimization parameter variable set of the main population when the iterative processing meets the iteration stop condition, and determine at least one optimal stopping point according to the candidate stopping points in the target optimization parameter variable set; wherein the first objective function is determined based on the minimum distance information in the distance information subset of the optimization parameter variables; and the second objective function is determined based on the maximum distribution area of the candidate stopping points of the optimization parameter variables in the decision space.
9. A computer device, It is characterized in that include: A memory and a processor, wherein the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the automatic driving stop point selection method based on a multi-modal multi-objective evolutionary algorithm as described in any one of claims 1 to 7 by executing the computer instructions.
10. A computer-readable storage medium, It is characterized in that The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a computer to execute the automatic driving stop point selection method based on a multi-modal multi-objective evolutionary algorithm as described in any one of claims 1 to 7.
Citation Information
Cited By
Regional agricultural planting structure optimization method and device based on constrained multi-objective evolutionary algorithm, equipment and medium
CN119991330A
