Automatic driving parking lot path planning method based on fuzzy control and path optimization

Abstract

The invention discloses an automatic driving parking lot path planning method based on fuzzy control and path optimization. The method comprises the following steps: S1, collecting environmental data of a parking lot; s2, generating an environment state diagram; s3, establishing a fuzzy rule base; s4, dynamically adjusting path planning parameters; s5, constructing a path weight matrix for dynamic planning based on a fuzzy reasoning result; s6, calculating a global optimal path from the current position of the vehicle to the target parking space based on the weight matrix by using a dynamic planning algorithm; s7, updating the global optimal path by using a dynamic programming algorithm; and S8, guiding the vehicle to run to the target parking space according to the planned path. According to the invention, the real-time performance of path planning, the multi-objective optimization capability and the driving safety are obviously improved, and reliable technical guarantee is provided for safe and efficient operation of the automatic driving vehicle in the parking lot.

Description

Technical Field

[0001] The present invention relates to the field of autonomous driving technology, and in particular to an autonomous driving parking lot path planning method based on fuzzy control and path optimization. Background Art

[0002] With the rapid development of autonomous driving technology, parking lots, a typical semi-structured and dynamic environment, have become a key scenario for testing the adaptability and robustness of autonomous driving systems. As an important part of the operation of autonomous vehicles, parking lot path planning needs to solve the problems of dynamic environment perception, multi-objective path optimization, and complex control decision-making.

[0003] At present, the research on path planning of autonomous driving parking lots mainly focuses on two types of methods: path planning methods based on static maps and dynamic planning methods based on optimization algorithms. Path planning methods based on static maps usually rely on pre-built parking lot models and calculate the shortest path through traditional A* algorithms or Dijkstra algorithms. Such methods show good performance in static and regularized environments, but in dynamic parking lots, they cannot perceive and respond to changes in obstacles in real time. The existing technologies mainly have the following problems:

[0004] 1. Insufficient adaptability to dynamic environments: Static map planning methods cannot perceive changes in obstacles in real time, and it is difficult to dynamically adjust the planned path during vehicle driving, resulting in the path planning results being neither real-time nor adaptable.

[0005] 2. Inefficient path planning: Traditional optimization algorithms have low computational efficiency when dealing with narrow passages and complex obstacle distribution in parking lots. In scenarios of multi-objective optimization (path length, obstacle avoidance, and traffic efficiency), planning time is prone to being too long.

[0006] 3. Lack of driving safety: Existing technologies have limited control accuracy over vehicle driving status in narrow lanes or complex intersection scenarios. When faced with dynamic obstacles, the planned path cannot effectively avoid potential risks, which can easily lead to collisions or driving failures.

[0007] 4. Insufficient multivariable optimization capabilities: Existing path planning methods lack comprehensive analysis of the correlation between variables when dealing with multivariate factors in parking lots, making it difficult to simultaneously meet global optimality and local flexibility.

[0008] In summary, the existing technology has obvious deficiencies in dynamic environment adaptability, planning efficiency, driving safety and multivariable optimization capabilities. A new path planning method is urgently needed to effectively cope with the technical challenges in the complex scenario of parking lots. Summary of the invention

[0009] One purpose of the present invention is to propose an autonomous driving parking lot path planning method based on fuzzy control and path optimization. The present invention achieves significant improvements in the real-time performance, multi-objective optimization capability and driving safety of path planning, and provides reliable technical guarantee for the safe and efficient operation of autonomous driving vehicles in parking lots.

[0010] According to an embodiment of the present invention, a method for automatic driving parking lot path planning based on fuzzy control and path optimization includes the following steps:

[0011] The following steps are involved:

[0012] S1. Collect environmental data of the parking lot;

[0013] S2. Fusing and processing the collected environmental data, and generating an environmental status diagram based on the real-time dynamics of the parking lot;

[0014] S3. Define the fuzzy control input variables based on the environmental state diagram, define the fuzzy control output variables at the same time, and establish a fuzzy rule base based on the dynamic characteristics of the fuzzy control output variables;

[0015] S4. Use fuzzy inference algorithm to perform real-time calculation on environmental data set, output real-time driving parameters of the vehicle according to fuzzy rule base, and use the calculation results to dynamically adjust path planning parameters;

[0016] S5. construct a path weight matrix for dynamic planning based on the fuzzy reasoning results, and the weight value of the path weight matrix is ​​dynamically adjusted according to the relative distance between the vehicle and the obstacle, the congestion level of the path and the dynamic characteristics of the current environment;

[0017] S6. Calculate the global optimal path from the current position of the vehicle to the target parking space based on the weight matrix using a dynamic programming algorithm;

[0018] S7. When the obstacle distribution, other vehicle trajectories or the state of the target parking space in the parking lot environment changes, the environment state diagram is updated in real time, and the weight matrix is ​​re-adjusted based on the fuzzy reasoning results, and the global optimal path is updated using the dynamic programming algorithm;

[0019] S8. Generate vehicle control instructions based on the global optimal path generated by dynamic programming and the vehicle driving parameters output by fuzzy control to adjust the vehicle's steering angle, driving speed, acceleration and deceleration, and guide the vehicle to drive to the target parking space along the planned path.

[0020] Optionally, the S1 step specifically includes the following contents:

[0021] S11. Collect the spatial position data P of obstacles in the parking lot through the vehicle-mounted sensors obs ={(x i ,yi ,z i )}, where (x i ,y i ,z i ) represents the three-dimensional space coordinates of the i-th obstacle;

[0022] S12. Use the on-board camera and radar equipment to obtain the road layout data of the parking lot L road ;

[0023] S13. Use dynamic target tracking algorithm to collect the moving trajectory data of other vehicles T veh ;

[0024] S14. Obtain the distribution information of the target parking space through the parking lot management system or real-time detection slot ;

[0025] S15. Collect the spatial position data P of the obstacle obs , road layout data L road , the moving trajectory data of other vehicles T veh and the distribution information of the target parking space P slot Synchronize according to the unified timestamp to generate a consistent parking environment data set:

[0026] E park = {P obs ,L road ,T veh ,P slot}.

[0027] Optionally, the step S2 specifically includes the following contents:

[0028] S21. Parking lot environment data set E park Completeness verification is performed to ensure that there are no missing or redundant key parameters in each data subset;

[0029] S22. Obstacle spatial position data P obs The Kalman filter method is used to reduce noise interference in data collection and generate obstacle position data P′ after filtering optimization. obs ;

[0030] S23. Road layout data L road Smoothing is performed, and the path node coordinates are optimized by polynomial fitting method to generate the optimized road layout data L′ road ;

[0031] S24. Use trajectory prediction algorithm to predict the movement trajectory data of other vehicles T veh Perform time series analysis to predict the possible positions of other vehicles in the future and generate trajectory prediction data T′veh ;

[0032] S25. Based on the optimized obstacle position data P′ obs , road layout data L′ road , trajectory prediction data T′ veh and parking space distribution data P′ slot , the regional segmentation algorithm is used to generate the parking lot environment state map:

[0033] G env = {R obs ,R free ,R slot};

[0034] Among them, the obstacle area R obs The calculation is:

[0035]

[0036] Among them, N obs represents the number of obstacles, (x′ i ,y′ i ,z′ i ) represents the center coordinate of obstacle i, r obs,i ,i} is the influence radius of obstacle i;

[0037] Passable area R free The calculation is:

[0038]

[0039] Among them, N road represents the number of roads, L′ road,j represents the set of optimized path nodes for road j;

[0040] Parking area R slot The calculation is:

[0041]

[0042] Among them, N slot represents the number of parking spaces, P slot,k Represents the spatial coordinate set of parking space k.

[0043] Optionally, the S3 step specifically includes the following contents:

[0044] S31. Extract the relative distance variable D between the vehicle and the obstacle based on the parking lot environment state diagram rel :

[0045]

[0046] Among them, (x v ,y v ,z v ) represents the current position of the vehicle, R obs Indicates the obstacle area, D rel Used to measure the distance between the vehicle and the nearest obstacle;

[0047] Extracting vehicle speed variable V based on parking lot environment state diagram veh ,The vehicle speed is measured by the on-board speed sensor and divided into three fuzzy sets: low, medium and high;

[0048] Extracting vehicle steering angle variable A based on parking lot environment state diagram steer ,The current steering angle of the vehicle is obtained through the steering wheel angle sensor, which is defined as three fuzzy sets of negative steering angle, zero steering angle and positive steering angle;

[0049] Extracting the path congestion degree variable C based on the parking environment state diagram path , based on the traversable area R free and the predicted trajectory data T′ of other vehicles veh , calculate the congestion weight of the path node:

[0050]

[0051] in, Indicates whether the position (x, y) is in the predicted trajectory of other vehicles;

[0052] S32. Define fuzzy control output variables:

[0053] Optimized vehicle speed V′ veh , defined as three fuzzy sets of low speed, medium speed and high speed, which are used to dynamically adjust the speed of the vehicle in different scenarios;

[0054] Optimized steering angle A′ steer , defined as three fuzzy sets of negative adjustment angle, zero adjustment angle and positive adjustment angle, which are used to optimize the steering control of the vehicle;

[0055] Path priority P priority , defined as three fuzzy sets of low priority, medium priority and high priority, which are used to identify the degree of preference in path planning;

[0056] S33. Based on the parking lot scene characteristics and vehicle driving safety requirements, define a fuzzy rule base to associate the mapping relationship between fuzzy input variables and fuzzy output variables:

[0057] If D rel is "near", and C path is "high", then V' vehis "low speed", A' steer is the “positive adjustment angle”, P priority is "low priority";

[0058] If D rel is "far", and C path is "low", then V' veh A′ is “high speed” steer is the “zero adjustment angle”, P priority is "High Priority".

[0059] S34. Input the fuzzy control variable D rel ,V veh ,A steer ,C path and fuzzy control output variable V′ veh ,A′ steer ,P priority Applied to the Fuzzy Inference module.

[0060] Optionally, the S4 step specifically includes the following contents:

[0061] S41. Establish a multi-input and multi-output fuzzy reasoning model based on fuzzy control input variables and fuzzy control output variables:

[0062] F fuzzy :(D rel ,V veh ,A steer ,C path )→(V′ veh ,A′ steer ,P priority );

[0063] Among them, F fuzzy represents the fuzzy inference function, which is used to map input variables to output variables;

[0064] S42. Parking lot environment status diagram G env The extracted input variables are fuzzy processed and the relative distance variable D between the vehicle and the obstacle is converted into rel Fuzzy into near, medium and far fuzzy sets, for the vehicle speed variable V veh Apply the triangular membership function to divide the vehicle steering angle variable A into low-speed, medium-speed and high-speed fuzzy sets. steer The trapezoidal membership function is applied to divide the fuzzy sets into negative adjustment angle, zero adjustment angle and positive adjustment angle, and the path congestion degree variable C path Applying Gaussian membership function, it is divided into low congestion, medium congestion and high congestion fuzzy sets'

[0065] S43. Based on the fuzzy rule base, the fuzzified input variables are transformed into fuzzy reasoning model F fuzzy Perform calculations, use the minimum-maximum composite rule to map the membership of the input variable in combination with the rule base to the membership of the output variable, and output the fuzzy control variable;

[0066] S44. Fuzzy reasoning result V′ veh ,A′ steer ,P priority Perform defuzzification and use the centroid method to calculate the actual value of the output variable:

[0067] Defuzzified vehicle speed V′ veh :

[0068]

[0069] Among them, μ V (v) fuzzy membership function representing the speed variable;

[0070] Defuzzified vehicle steering angle A′ steer :

[0071]

[0072] Among them, μ A (a) Fuzzy membership function representing the steering angle variable;

[0073] Defuzzified path priority P priority :

[0074]

[0075] Among them, μ P (p) represents the fuzzy membership function of the path priority variable;

[0076] S45. Defuzzification result V′ veh ,A′ steer ,P priority As the input of the path planning module.

[0077] Optionally, the step S5 specifically includes the following contents:

[0078] S51. Based on parking environment state diagram G env And the fuzzy reasoning result (V′ veh ,A′ steer ,P priority ), initialize the dynamic path weight matrix

[0079]

[0080] Among them, i and j represent different nodes in the path network, Represents the initial weight value from node i to node j;

[0081] In order to adapt to the theme of autonomous driving parking lot path planning, the influence of fuzzy control output variables on path planning is fully considered during initialization. Let:

[0082]

[0083] Among them, φ(·) represents the initial weight generation function based on vehicle speed, steering angle and path priority, which is used to assign the relative importance of the connection between each node in the initial stage of path network construction;

[0084] S52. During the driving process of the vehicle, according to the relative distance D between the vehicle and the obstacle rel , path congestion degree C path And the dynamic characteristics of the current environment perform real-time scheduling of weights:

[0085]

[0086] Among them, w ij (t) represents the weight value from node i to node j at time t, Ψ(D rel ) represents the relative distance D between the vehicle and the obstacle rel The risk assessment function is used to amplify the weight increment when the distance is too close. α1 is the coefficient for adjusting the influence. Ω(C path ) represents the path congestion level C path The corresponding weighted function is used to reflect the changing trend of the parking lot road traffic volume, α2 is the coefficient for adjusting the impact, Γ(E dyn (t)) represents the correction function combining the dynamic characteristics of the parking environment, E dyn (t) reflects the traversable area R free The change information at time t is used to dynamically modify the weight value by real-time sensing of the addition of obstacles or changes in vehicle flow;

[0087] S53. When the environmental state detection and fuzzy reasoning results change, based on the weight w ij (t) The dynamic path weight matrix is ​​updated in real time, so that the weights between nodes are adjusted in time with the changes in the distance between the vehicle and the obstacle, the fluctuation of the path congestion level, and the dynamic characteristics of the parking environment.

[0088] Optionally, the step S6 specifically includes the following contents:

[0089] S61. Based on the dynamic path weight matrix W pathConstruct a path search space G = (V, E), where V = {v1, v2, ..., v n} represents the node set in the path network, including the vehicle's current location node v start 、Target parking space node v goal and other intermediate nodes, E={(v i ,v j )} represents node v i and v j The edge set between them, the corresponding edge weight is w ij Sure;

[0090] S62. is the current position v of the vehicle start To the target parking space v goal Generate the global optimal path P opt , define the dynamic programming objective function:

[0091]

[0092] Where P = {v start ,v1,v2,…,v goal} represents the candidate path, Cost(P) is the total weight of the path, which must satisfy:

[0093]

[0094] The objective function takes the minimum total weight of the path as the optimization goal, and combines the state transfer equation in dynamic programming to ensure the feasibility of the path:

[0095]

[0096] Among them, Cost(v j ) indicates reaching node v j The minimum cost, Pred(v j ) represents node v j The set of predecessor nodes;

[0097] S63. During the path planning process, constraints are added based on the characteristics of the parking lot scene, including avoiding obstacles so that each edge in the path satisfies Path congestion constraint, giving priority to paths with less congestion; path length constraint;

[0098] S64. Use the dynamic programming algorithm of the dynamic path weight matrix to search for the optimal path and initialize the current position of the vehicle v start The cost of the node is 0, and the cost of other nodes is infinite; from the starting node v start First, update the cost Cost(v j ) and the previous node; repeat the update until the target parking space node v is reachedgoal Or the search space traversal is completed; the global optimal path is generated by backtracking through the predecessor nodes:

[0099] P opt = {v start ,v1,v2,…,v goal}.

[0100] The beneficial effects of the present invention are:

[0101] (1) The present invention realizes the real-time response capability to the dynamic environment of the parking lot by deeply combining fuzzy control with dynamic programming algorithm. The fuzzy control module can adjust the dynamic path weight matrix in real time based on the relative distance between the vehicle and the obstacle, the path congestion level and the dynamic characteristics of the environment, so that the path planning can quickly adapt to the addition and movement of obstacles and the dynamic changes of the passage path. Compared with the traditional path planning method based on static maps, the path planning time of the present invention is significantly reduced, and the optimal path can be dynamically updated during the vehicle driving process, ensuring the real-time and adaptability of path planning in complex parking lot scenarios.

[0102] (2) The present invention adopts a dynamic programming algorithm combined with a multivariable optimization strategy, and comprehensively considers the multi-dimensional constraints of path length, obstacle distribution, and path congestion in path planning. By dynamically adjusting the weight matrix and introducing a multi-objective optimization function, the path planning not only meets the requirements of the global shortest path, but also effectively avoids obstacles and high congestion areas, ensuring the global optimality and local flexibility of path planning. This problem is effectively solved through the path weight calculation model and real-time optimization strategy, greatly improving the efficiency and quality of path planning.

[0103] (3) The present invention dynamically optimizes and controls the vehicle's driving speed, steering angle and path priority through fuzzy reasoning. Combined with the real-time updated path weight matrix, the present invention significantly improves the vehicle's driving safety in narrow lanes, complex intersections and dynamic obstacle environments. It can optimize the steering angle and speed control in narrow scenes according to the relative distance between the vehicle and the obstacle and the dynamic characteristics of the environment to reduce the risk of collision. In addition, the present invention optimizes the coupling design of path planning and vehicle control specifically for the dynamic and unstructured scene of a parking lot, making the vehicle more adaptable in complex environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0104] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0105] Figure 1 This is a flow chart of an automatic driving parking lot path planning method based on fuzzy control and path optimization proposed by the present invention. DETAILED DESCRIPTION

[0106] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, which only illustrate the basic structure of the present invention in a schematic manner, and therefore only show the components related to the present invention.

[0107] refer to Figure 1 , a method for automatic driving parking lot path planning based on fuzzy control and path optimization, comprising the following steps:

[0108] The following steps are involved:

[0109] S1. Collect environmental data of the parking lot;

[0110] S2. Fusing and processing the collected environmental data, and generating an environmental status diagram based on the real-time dynamics of the parking lot;

[0111] S3. Define the fuzzy control input variables based on the environmental state diagram, define the fuzzy control output variables at the same time, and establish a fuzzy rule base based on the dynamic characteristics of the fuzzy control output variables;

[0112] S4. Use fuzzy inference algorithm to perform real-time calculation on environmental data set, output real-time driving parameters of the vehicle according to fuzzy rule base, and use the calculation results to dynamically adjust path planning parameters;

[0113] S5. construct a path weight matrix for dynamic planning based on the fuzzy reasoning results, and the weight value of the path weight matrix is ​​dynamically adjusted according to the relative distance between the vehicle and the obstacle, the congestion level of the path and the dynamic characteristics of the current environment;

[0114] S6. Calculate the global optimal path from the current position of the vehicle to the target parking space based on the weight matrix using a dynamic programming algorithm;

[0115] S7. When the obstacle distribution, other vehicle trajectories or the state of the target parking space in the parking lot environment changes, the environment state diagram is updated in real time, and the weight matrix is ​​re-adjusted based on the fuzzy reasoning results, and the global optimal path is updated using the dynamic programming algorithm;

[0116] S8. Generate vehicle control instructions based on the global optimal path generated by dynamic programming and the vehicle driving parameters output by fuzzy control to adjust the vehicle's steering angle, driving speed, acceleration and deceleration, and guide the vehicle to drive to the target parking space along the planned path.

[0117] In this implementation, step S1 specifically includes the following contents:

[0118] S11. Collect the spatial position data P of obstacles in the parking lot through the vehicle-mounted sensors obs ={(xi ,y i ,z i )}, where (x i ,y i ,z i ) represents the three-dimensional space coordinates of the i-th obstacle;

[0119] S12. Use the on-board camera and radar equipment to obtain the road layout data of the parking lot L road ;

[0120] S13. Use dynamic target tracking algorithm to collect the moving trajectory data of other vehicles T veh ;

[0121] S14. Obtain the distribution information of the target parking space through the parking lot management system or real-time detection slot ;

[0122] S15. Collect the spatial position data P of the obstacle obs , road layout data L road , the moving trajectory data of other vehicles T veh and the distribution information of the target parking space P slot Synchronize according to the unified timestamp to generate a consistent parking environment data set:

[0123] E park = {P obs ,L road ,T veh ,P slot}.

[0124] In this implementation, step S2 specifically includes the following contents:

[0125] S21. Parking lot environment data set E park Completeness verification is performed to ensure that there are no missing or redundant key parameters in each data subset;

[0126] S22. Obstacle spatial position data P obs The Kalman filter method is used to reduce noise interference in data collection and generate obstacle position data P′ after filtering optimization. obs ;

[0127] S23. Road layout data L road Smoothing is performed, and the path node coordinates are optimized by polynomial fitting method to generate the optimized road layout data L′ road ;

[0128] S24. Use trajectory prediction algorithm to predict the movement trajectory data of other vehicles T vehPerform time series analysis to predict the possible positions of other vehicles in the future and generate trajectory prediction data T′ veh ;

[0129] S25. Based on the optimized obstacle position data P′ obs , road layout data L′ road , trajectory prediction data T′ veh and parking space distribution data P slot , the regional segmentation algorithm is used to generate the parking lot environment state map:

[0130] G env = {R obs ,R free ,R slot};

[0131] Among them, the obstacle area R obs The calculation is:

[0132]

[0133] Among them, N obs represents the number of obstacles, (x′ i ,y′ i ,z′ i ) represents the center coordinate of obstacle i, r obs,i ,i} is the influence radius of obstacle i;

[0134] Passable area R free The calculation is:

[0135]

[0136] Among them, N road represents the number of roads, L′ road,j represents the set of optimized path nodes for road j;

[0137] Parking area R slot The calculation is:

[0138]

[0139] Among them, N slot represents the number of parking spaces, P slot,k Represents the spatial coordinate set of parking space k.

[0140] In this implementation, step S3 specifically includes the following contents:

[0141] S31. Extract the relative distance variable D between the vehicle and the obstacle based on the parking lot environment state diagram rel :

[0142]

[0143] Among them, (x v ,y v ,z v ) represents the current position of the vehicle, R obs Indicates the obstacle area, D rel Used to measure the distance between the vehicle and the nearest obstacle;

[0144] Extracting vehicle speed variable V based on parking lot environment state diagram veh ,The vehicle speed is measured by the on-board speed sensor and divided into three fuzzy sets: low, medium and high;

[0145] Extracting vehicle steering angle variable A based on parking lot environment state diagram steer ,The current steering angle of the vehicle is obtained through the steering wheel angle sensor, which is defined as three fuzzy sets of negative steering angle, zero steering angle and positive steering angle;

[0146] Extracting the path congestion degree variable C based on the parking environment state diagram path , based on the traversable area R free and the predicted trajectory data T′ of other vehicles veh , calculate the congestion weight of the path node:

[0147]

[0148] in, Indicates whether the position (x, y) is in the predicted trajectory of other vehicles;

[0149] S32. Define fuzzy control output variables:

[0150] Optimized vehicle speed V′ veh , defined as three fuzzy sets of low speed, medium speed and high speed, which are used to dynamically adjust the speed of the vehicle in different scenarios;

[0151] Optimized steering angle A′ steer , defined as three fuzzy sets of negative adjustment angle, zero adjustment angle and positive adjustment angle, which are used to optimize the steering control of the vehicle;

[0152] Path priority P priority , defined as three fuzzy sets of low priority, medium priority and high priority, which are used to identify the degree of preference in path planning;

[0153] S33. Based on the parking lot scene characteristics and vehicle driving safety requirements, define a fuzzy rule base to associate the mapping relationship between fuzzy input variables and fuzzy output variables:

[0154] If D relis "near", and C path is "high", then V' veh is "low speed", A' steer is the “positive adjustment angle”, P priority is "low priority";

[0155] If D rel is "far", and C path is "low", then V' veh A′ is “high speed” steer is the “zero adjustment angle”, P priority is "High Priority".

[0156] S34. Input the fuzzy control variable D rel ,V veh ,A steer ,C path and fuzzy control output variable V′ veh ,A′ steer ,P priority Applied to the Fuzzy Inference module.

[0157] In this implementation, step S4 specifically includes the following contents:

[0158] S41. Establish a multi-input and multi-output fuzzy reasoning model based on fuzzy control input variables and fuzzy control output variables:

[0159] F fuzzy :(D rel ,V veh ,A steer ,C path )→(V′ veh ,A′ steer ,P priority );

[0160] Among them, F fuzzy represents the fuzzy inference function, which is used to map input variables to output variables;

[0161] S42. Parking lot environment status diagram G env The extracted input variables are fuzzy processed and the relative distance variable D between the vehicle and the obstacle is converted into rel Fuzzy into near, medium and far fuzzy sets, for the vehicle speed variable V veh Apply the triangular membership function to divide the vehicle steering angle variable A into low-speed, medium-speed and high-speed fuzzy sets. steer The trapezoidal membership function is applied to divide the fuzzy sets into negative adjustment angle, zero adjustment angle and positive adjustment angle, and the path congestion degree variable C path Applying Gaussian membership function, it is divided into low congestion, medium congestion and high congestion fuzzy sets'

[0162] S43. Based on the fuzzy rule base, the fuzzified input variables are transformed into fuzzy reasoning model F fuzzy Perform calculations, use the minimum-maximum composite rule to map the membership of the input variable in combination with the rule base to the membership of the output variable, and output the fuzzy control variable;

[0163] S44. Fuzzy reasoning result V′ veh ,A′ steer ,P priority Perform defuzzification and use the centroid method to calculate the actual value of the output variable:

[0164] Defuzzified vehicle speed V′ veh :

[0165]

[0166] Among them, μ V (v) fuzzy membership function representing the speed variable;

[0167] Defuzzified vehicle steering angle A′ steer :

[0168]

[0169] Among them, μ A (a) Fuzzy membership function representing the steering angle variable;

[0170] Defuzzified path priority P priority :

[0171]

[0172] Among them, μ P (p) represents the fuzzy membership function of the path priority variable;

[0173] S45. Defuzzification result V′ veh ,A′ steer ,P priority As the input of the path planning module.

[0174] In this implementation, step S5 specifically includes the following contents:

[0175] S51. Based on parking environment state diagram G env And the fuzzy reasoning result (V′ veh ,A′ steer ,P priority ), initialize the dynamic path weight matrix

[0176]

[0177] Among them, i and j represent different nodes in the path network, Represents the initial weight value from node i to node j;

[0178] In order to adapt to the theme of autonomous driving parking lot path planning, the influence of fuzzy control output variables on path planning is fully considered during initialization. Let:

[0179]

[0180] Among them, φ(·) represents the initial weight generation function based on vehicle speed, steering angle and path priority, which is used to assign the relative importance of the connection between each node in the initial stage of path network construction;

[0181] S52. During the driving process of the vehicle, according to the relative distance D between the vehicle and the obstacle rel , path congestion degree C path And the dynamic characteristics of the current environment perform real-time scheduling of weights:

[0182]

[0183] Among them, w ij (t) represents the weight value from node i to node j at time t, Ψ(D rel ) represents the relative distance D between the vehicle and the obstacle rel The risk assessment function is used to amplify the weight increment when the distance is too close. α1 is the coefficient for adjusting the influence. Ω(C path ) represents the path congestion level C path The corresponding weighted function is used to reflect the changing trend of the parking lot road traffic volume, α2 is the coefficient for adjusting the impact, Γ(E dyn (t)) represents the correction function combining the dynamic characteristics of the parking environment, E dyn (t) reflects the traversable area R free The change information at time t is used to dynamically modify the weight value by real-time sensing of the addition of obstacles or changes in vehicle flow;

[0184] S53. When the environmental state detection and fuzzy reasoning results change, based on the weight w ij (t) The dynamic path weight matrix is ​​updated in real time, so that the weights between nodes are adjusted in time with the changes in the distance between the vehicle and the obstacle, the fluctuation of the path congestion level, and the dynamic characteristics of the parking environment.

[0185] In this implementation, step S6 specifically includes the following contents:

[0186] S61. Based on the dynamic path weight matrix W pathConstruct a path search space G = (V, E), where V = {v1, v2, ..., v n} represents the node set in the path network, including the vehicle's current location node v start 、Target parking space node v goal and other intermediate nodes, E={(v i ,v j )} represents node v i and v j The edge set between them, the corresponding edge weight is w ij Sure;

[0187] S62. is the current position v of the vehicle start To the target parking space v goal Generate the global optimal path P opt , define the dynamic programming objective function:

[0188]

[0189] Where P = {v start ,v1,v2,…,v goal} represents the candidate path, Cost(P) is the total weight of the path, which must satisfy:

[0190]

[0191] The objective function takes the minimum total weight of the path as the optimization goal, and combines the state transfer equation in dynamic programming to ensure the feasibility of the path:

[0192]

[0193] Among them, Cost(v j ) indicates reaching node v j The minimum cost, Pred(v j ) represents node v j The set of predecessor nodes;

[0194] S63. During the path planning process, constraints are added based on the characteristics of the parking lot scene, including avoiding obstacles so that each edge in the path satisfies Path congestion constraint, giving priority to paths with less congestion; path length constraint;

[0195] S64. Use the dynamic programming algorithm of the dynamic path weight matrix to search for the optimal path and initialize the current position of the vehicle v start The cost of the node is 0, and the cost of other nodes is infinite; from the starting node v start First, update the cost Cost(v j ) and the previous node; repeat the update until the target parking space node v is reachedgoal Or the search space traversal is completed; the global optimal path is generated by backtracking through the predecessor nodes:

[0196] P opt = {v start ,v1,v2,…,v goal}.

[0197] Embodiment 1:

[0198] Embodiment: At 4:30 p.m. on December 15, 2024, an autonomous driving vehicle equipped with the system of the present invention enters an underground parking lot in an urban complex. The parking lot has a total of 320 parking spaces with a complex layout, a passage width of 3.5 meters, and multiple intersections and pillars distributed irregularly. The vehicle receives an instruction for the target parking space number "B12" at the entrance of the parking lot. The parking space is located approximately 350 meters from the entrance, and there are multiple dynamic obstacles (other vehicles) and static obstacles (parking pillars and dividers) along the way.

[0199] After entering the parking lot, the vehicle's sensor module collects parking lot environment data in real time, including:

[0200] 1. About 10 meters from the entrance, a white SUV (dynamic obstacle) was found slowly approaching, with its initial position at "(12.3,15.7)" meters.

[0201] A cylindrical static obstacle was detected on the right side of a channel 2.20 meters later. Its spatial coordinates were "(35.8,20.2)" meters and the diameter of the column was 0.5 meters.

[0202] 3. The location of the target parking space "B12" is marked as "(345.5,18.0)" meters.

[0203] Based on the above environmental data and the parking lot management system feedback, the vehicle initializes the dynamic path weight matrix and plans an initial path that avoids static obstacles and has a shortest path length of "352.7 meters".

[0204] When the vehicle traveled to 25 meters, it detected in real time that the white SUV had entered the main channel. Its current trajectory showed that it would stop at 35 meters, and the estimated time was 10 seconds. During the dynamic weight matrix update process, the system increased the path weight value near the white SUV to "1.5 times". The dynamic planning module adjusted the path in real time and replanned a detour path. The path length was extended to "365.3 meters". The vehicle slowed down to 5 kilometers per hour and smoothly bypassed the dynamic obstacle.

[0205] At the intersection 100 meters away, the vehicle detected a black car approaching quickly from the left at a speed of 15 km / h. The system activated the fuzzy reasoning module and calculated the optimal steering angle of "+20 degrees" and deceleration range of "3 km / h" based on the relative distance of the vehicles "5 meters", speed difference "10 km / h" and steering requirements. The vehicle successfully avoided the dynamic conflict at the intersection and entered the next channel.

[0206] When the vehicle drives to the vicinity of the target parking space (the distance is less than 5 meters), the parking lot environment status diagram shows that the width of the left boundary of the parking space from the obstacle is "3.2 meters". The fuzzy control module optimizes the steering angle and speed, and finally outputs the optimal reversing path. The vehicle slowly reverses into the target parking space at a speed of "1 meter per second". The entire parking process takes "12 seconds" and the final position deviation of the vehicle is "2 centimeters".

[0207] In order to verify the effect of the present invention, a comparative experiment was conducted using the traditional A* path planning algorithm in the same scenario. The specific data are as follows:

[0208] Test items Method of the present invention Traditional methods Initial path planning time 0.89 seconds 2.1 seconds Total driving time 6 minutes 15 seconds 8 minutes 20 seconds Dynamic obstacle avoidance times 3 times 2 times (1 failed) Static obstacle collision rate 0％ 5％ Total path length 368.7 m 362.5m

[0209] In a complex parking environment, the present invention effectively solves the problems of dynamic obstacle avoidance, precise navigation in narrow areas, and safe passage through complex intersections by dynamically adjusting the path weight matrix in real time, and an algorithm combining fuzzy control and dynamic programming. Compared with traditional methods, the present invention performs better in planning time, driving time, and safety, proving its feasibility and practicality in real scenarios.

[0210] The present invention realizes the real-time response capability to the dynamic environment of the parking lot by deeply combining fuzzy control with dynamic programming algorithm. The fuzzy control module can adjust the dynamic path weight matrix in real time based on the relative distance between the vehicle and the obstacle, the path congestion degree and the dynamic characteristics of the environment, so that the path planning can quickly adapt to the addition and movement of obstacles and the dynamic changes of the passage path. Compared with the traditional path planning method based on static maps, the path planning time of the present invention is significantly reduced, and the optimal path can be dynamically updated during the vehicle driving process, ensuring the real-time and adaptability of path planning in complex parking lot scenarios.

[0211] The present invention adopts a dynamic programming algorithm combined with a multivariable optimization strategy, and comprehensively considers the multi-dimensional constraints of path length, obstacle distribution, and path congestion in path planning. By dynamically adjusting the weight matrix and introducing a multi-objective optimization function to plan the path, it not only meets the requirements of the global shortest path, but also effectively avoids obstacles and high-congestion areas, ensuring the global optimality and local flexibility of path planning. This problem is effectively solved through the path weight calculation model and real-time optimization strategy, greatly improving the efficiency and quality of path planning.

[0212] The present invention dynamically optimizes and controls the vehicle's driving speed, steering angle and path priority through fuzzy reasoning, and combines a real-time updated path weight matrix to significantly improve the vehicle's driving safety in narrow lanes, complex intersections and dynamic obstacle environments. It can optimize the steering angle and speed control in narrow scenes according to the relative distance between the vehicle and the obstacle and the dynamic characteristics of the environment to reduce the risk of collision. In addition, the present invention optimizes the coupling design of path planning and vehicle control specifically for the dynamic and unstructured scene of a parking lot, making the vehicle more adaptable in complex environments.

[0213] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A method for automatic driving parking lot path planning based on fuzzy control and path optimization, characterized in that: The steps include: The following steps are involved: S1. Collect environmental data of the parking lot; S2. Fusing and processing the collected environmental data, and generating an environmental status diagram based on the real-time dynamics of the parking lot; S3. Define the fuzzy control input variables based on the environmental state diagram, define the fuzzy control output variables at the same time, and establish a fuzzy rule base based on the dynamic characteristics of the fuzzy control output variables; S4. Use fuzzy inference algorithm to perform real-time calculation on environmental data set, output real-time driving parameters of the vehicle according to fuzzy rule base, and use the calculation results to dynamically adjust path planning parameters; S5. construct a path weight matrix for dynamic planning based on the fuzzy reasoning results, and the weight value of the path weight matrix is ​​dynamically adjusted according to the relative distance between the vehicle and the obstacle, the congestion level of the path and the dynamic characteristics of the current environment; S6. Calculate the global optimal path from the current position of the vehicle to the target parking space based on the weight matrix using a dynamic programming algorithm; S7. When the obstacle distribution, other vehicle trajectories or the state of the target parking space in the parking lot environment changes, the environment state diagram is updated in real time, and the weight matrix is ​​re-adjusted based on the fuzzy reasoning results, and the global optimal path is updated using the dynamic programming algorithm; S8. Generate vehicle control instructions based on the global optimal path generated by dynamic programming and the vehicle driving parameters output by fuzzy control to adjust the vehicle's steering angle, driving speed, acceleration and deceleration, and guide the vehicle to drive to the target parking space along the planned path.

2. The method for automatic driving parking lot path planning based on fuzzy control and path optimization according to claim 1 is characterized in that: The S1 step specifically includes the following contents: S11. Collect the spatial position data P of obstacles in the parking lot through the vehicle-mounted sensors obs ={(x i ,y i ,z i )}, where (x i ,y i ,z i ) represents the three-dimensional space coordinates of the i-th obstacle; S12. Use the on-board camera and radar equipment to obtain the road layout data of the parking lot L road ; S13. Use dynamic target tracking algorithm to collect the moving trajectory data of other vehicles T veh ; S14. Obtain the distribution information of the target parking space through the parking lot management system or real-time detection slot ; S15. Collect the spatial position data P of the obstacle obs , road layout data L road , the moving trajectory data of other vehicles T veh and the distribution information of the target parking space P slot Synchronize according to the unified timestamp to generate a consistent parking environment data set: E park ={P obs ,L road ,T veh ,P slot }。 3. The method for automatic driving parking lot path planning based on fuzzy control and path optimization according to claim 1 is characterized in that: The S2 step specifically includes the following contents: S21. Parking lot environment data set E park Completeness verification is performed to ensure that there are no missing or redundant key parameters in each data subset; S22. Obstacle spatial position data P obs The Kalman filter method is used to reduce noise interference in data collection and generate obstacle position data P′ after filtering optimization. obs ; S23. Road layout data L road Smoothing is performed, and the path node coordinates are optimized by polynomial fitting method to generate the optimized road layout data L′ road ; S24. Use trajectory prediction algorithm to predict the movement trajectory data of other vehicles T veh Perform time series analysis to predict the possible positions of other vehicles in the future and generate trajectory prediction data T′ veh ; S25. Based on the optimized obstacle position data P′ obs , road layout data L′ road , trajectory prediction data T′ veh and parking space distribution data P slot , the regional segmentation algorithm is used to generate the parking lot environment state map: G env ={R obs ,R free ,R slot }; Among them, the obstacle area R obs The calculation is: Among them, N obs represents the number of obstacles, (x′ i ,y′ i ,z′ i ) represents the center coordinate of obstacle i, r obs,i ,i} is the influence radius of obstacle i; Passable area R free The calculation is: Among them, N road represents the number of roads, L′ road,j represents the set of optimized path nodes for road j; Parking area R slot The calculation is: Among them, N slot represents the number of parking spaces, P slot,k Represents the spatial coordinate set of parking space k.

4. The method for automatic driving parking lot path planning based on fuzzy control and path optimization according to claim 1 is characterized in that: The S3 step specifically includes the following contents: S31. Extract the relative distance variable D between the vehicle and the obstacle based on the parking lot environment state diagram rel : Among them, (x v ,y v ,z v ) represents the current position of the vehicle, R obs Indicates the obstacle area, D rel Used to measure the distance between the vehicle and the nearest obstacle; Extracting vehicle speed variable V based on parking lot environment state diagram veh ,The vehicle speed is measured by the on-board speed sensor and divided into three fuzzy sets: low, medium and high; Extracting vehicle steering angle variable A based on parking lot environment state diagram steer ,The current steering angle of the vehicle is obtained through the steering wheel angle sensor, which is defined as three fuzzy sets of negative steering angle, zero steering angle and positive steering angle; Extracting the path congestion degree variable C based on the parking environment state diagram path , based on the traversable area R free and the predicted trajectory data T′ of other vehicles veh , calculate the congestion weight of the path node: in, Indicates whether the position (x, y) is in the predicted trajectory of other vehicles; S32. Define fuzzy control output variables: Optimized vehicle speed V′ veh , defined as three fuzzy sets of low speed, medium speed and high speed, which are used to dynamically adjust the speed of the vehicle in different scenarios; Optimized steering angle A′ steer , defined as three fuzzy sets of negative adjustment angle, zero adjustment angle and positive adjustment angle, which are used to optimize the steering control of the vehicle; Path priority P priority , defined as three fuzzy sets of low priority, medium priority and high priority, used to identify the degree of preference in path planning; S33. Based on the parking lot scene characteristics and vehicle driving safety requirements, define a fuzzy rule base to associate the mapping relationship between fuzzy input variables and fuzzy output variables: If D rel is "near", and C path is "high", then V' veh is "low speed", A′ steer is the "positive adjustment angle", P priority is "low priority"; If D rel is "far", and C path is "low", then V' veh is "high speed", A′ steer is the "zero adjustment angle", P priority is "high priority". S34. Input the fuzzy control variable D rel ,V veh ,A steer ,C path and fuzzy control output variable V′ veh ,A′ steer ,P priority Applied to the Fuzzy Inference module.

5. The method for automatic driving parking lot path planning based on fuzzy control and path optimization according to claim 1 is characterized in that: The S4 step specifically includes the following contents: S41. Establish a multi-input and multi-output fuzzy reasoning model based on fuzzy control input variables and fuzzy control output variables: F fuzzy :(D rel ,V veh ,A steer ,C path )→(V′ veh ,A′ steer ,P priority ); Among them, F fuzzy represents the fuzzy inference function, which is used to map input variables to output variables; S42. Parking lot environment status diagram G env The extracted input variables are fuzzy processed and the relative distance variable D between the vehicle and the obstacle is converted into rel Fuzzy into near, medium and far fuzzy sets, for the vehicle speed variable V veh Apply the triangular membership function to divide the vehicle steering angle variable A into low-speed, medium-speed and high-speed fuzzy sets. steer The trapezoidal membership function is applied to divide the fuzzy sets into negative adjustment angle, zero adjustment angle and positive adjustment angle, and the path congestion degree variable C path Applying Gaussian membership function, it is divided into low congestion, medium congestion and high congestion fuzzy sets' S43. Based on the fuzzy rule base, the fuzzified input variables are transformed into fuzzy reasoning model F fuzzy Perform calculations, use the minimum-maximum composite rule to map the membership of the input variable in combination with the rule base to the membership of the output variable, and output the fuzzy control variable; S44. Fuzzy reasoning result V′ veh ,A′ steer ,P priority Perform defuzzification and use the centroid method to calculate the actual value of the output variable: Defuzzified vehicle speed V′ veh : Among them, μ V (v) fuzzy membership function representing the speed variable; Defuzzified vehicle steering angle A′ steer : Among them, μ A (a) Fuzzy membership function representing the steering angle variable; Defuzzified path priority P priority : Among them, μ P (p) represents the fuzzy membership function of the path priority variable; S45. Defuzzification result V′ veh ,A′ steer ,P priority As the input of the path planning module.

6. The method for automatic driving parking lot path planning based on fuzzy control and path optimization according to claim 1 is characterized in that: The S5 step specifically includes the following contents: S51. Based on parking environment state diagram G env And the fuzzy reasoning result (V′ veh ,A′ steer ,P priority ), initialize the dynamic path weight matrix Among them, i and j represent different nodes in the path network, Represents the initial weight value from node i to node j; In order to adapt to the theme of autonomous driving parking lot path planning, the influence of fuzzy control output variables on path planning is fully considered during initialization. Let: Among them, φ(·) represents the initial weight generation function based on vehicle speed, steering angle and path priority, which is used to assign the relative importance of the connection between each node in the initial stage of path network construction; S52. During the driving process of the vehicle, according to the relative distance D between the vehicle and the obstacle rel , path congestion degree C path And the dynamic characteristics of the current environment perform real-time scheduling of weights: Among them, w ij (t) represents the weight value from node i to node j at time t, Ψ(D rel ) represents the relative distance D between the vehicle and the obstacle rel The risk assessment function is used to amplify the weight increment when the distance is too close. α1 is the coefficient for adjusting the influence. Ω(C path ) represents the path congestion level C path The corresponding weighted function is used to reflect the changing trend of the parking lot road traffic volume, α2 is the coefficient for adjusting the impact, Γ(E dyn (t)) represents the correction function combining the dynamic characteristics of the parking environment, E dyn (t) reflects the traversable area R free The change information at time t is used to dynamically modify the weight value by real-time sensing of the addition of obstacles or changes in vehicle flow; S53. When the environmental state detection and fuzzy reasoning results change, based on the weight w ij (t) The dynamic path weight matrix is ​​updated in real time, so that the weights between nodes are adjusted in time with the changes in the distance between the vehicle and the obstacle, the fluctuation of the path congestion level, and the dynamic characteristics of the parking environment.

7. The method for automatic driving parking lot path planning based on fuzzy control and path optimization according to claim 1 is characterized in that: The S6 step specifically includes the following contents: S61. Based on the dynamic path weight matrix W path Construct a path search space G = (V, E), where V = {v1, v2, ..., v n } represents the node set in the path network, including the vehicle's current location node v start 、Target parking space node v goal and other intermediate nodes, E={(v i ,v j )} represents node v i and v j The edge set between them, the corresponding edge weight is w ij Sure; S62. is the current position v of the vehicle start To the target parking space v goal Generate the global optimal path P opt , define the dynamic programming objective function: Where P = {v start ,v1,v2,…,v goal } represents the candidate path, Cost(P) is the total weight of the path, which must satisfy: The objective function takes the minimum total weight of the path as the optimization goal, and combines the state transfer equation in dynamic programming to ensure the feasibility of the path: Among them, Cost(v j ) indicates reaching node v j The minimum cost, Pred(v j ) represents node v j The set of predecessor nodes; S63. During the path planning process, constraints are added based on the characteristics of the parking lot scene, including avoiding obstacles so that each edge in the path satisfies Path congestion constraint, giving priority to paths with less congestion; path length constraint; S64. Use the dynamic programming algorithm of the dynamic path weight matrix to search for the optimal path and initialize the current position of the vehicle v start The cost of the node is 0, and the cost of other nodes is infinite; from the starting node v start First, update the cost Cost (v j ) and the previous node; repeat the update until the target parking space node v is reached goal Or the search space traversal is completed; the global optimal path is generated by backtracking through the predecessor nodes: P opt ={v start ,v1,v2,…,v goal }。

Images