An automatic driving decision control method, system and related device

Abstract

The application discloses an automatic driving decision control method and system and related equipment, and belongs to the technical field of automatic driving. The method comprises the following steps: generating an initial behavior decision tree by using a Markov algorithm according to a vehicle state; pruning and reconstructing the initial behavior decision tree according to human driving characteristics to obtain a pre-screened behavior decision tree; pruning and removing unsafe behavior decision sequences by using an RSS model, and iteratively generating vehicle candidate trajectories in combination with a vehicle kinematics model and the behavior decision tree; generating trajectories for the vehicle candidate trajectory sequence to obtain predicted trajectories; performing safety speed checking and collision checking on the predicted trajectories, removing behavior sequences that do not meet the requirements, and obtaining preselected trajectories; and calculating the overall cost of each preselected trajectory, taking the behavior sequence with the minimum overall cost as the optimal vehicle decision result and outputting the same. The application improves the safety and rationality of automatic driving decisions by screening and optimizing the decision sequence at multiple levels.

Description

Technical Field

[0001] This invention relates to the field of autonomous driving technology, specifically to an autonomous driving decision-making and control method, system, and related equipment. Background Technology

[0002] With the rapid development of autonomous driving technology, the decision-making control system of autonomous vehicles, as a core component, directly affects the safety and reliability of autonomous driving. Currently, the autonomous driving decision-making control system is mainly responsible for making appropriate driving decisions, such as lane keeping, lane changing, acceleration, and deceleration, based on environmental information obtained by the perception system and the vehicle's status.

[0003] Decision-making algorithms for autonomous driving are crucial for ensuring driving safety and efficiency. They typically comprise three layers: global path planning, behavioral decision-making, and motion planning. Current decision-making algorithms face challenges when dealing with complex dynamic traffic scenarios, especially when considering multiple safety mechanisms.

[0004] Global path planning is the first step in the decision-making process, generating a global path based on map information and the destination. This step forms the foundation for subsequent behavioral decisions and motion planning. Autonomous driving systems need to make specific behavioral decisions based on perceived environmental information (including other vehicles, pedestrians, obstacles, and traffic rules), such as deciding whether to change lanes to overtake or follow the vehicle in front. Behavioral decisions need to consider multi-agent interactions and uncertainties; partially observable Markov decision processes can be used to handle such uncertainties. Based on the behavioral decisions, the motion planning layer is responsible for generating a specific trajectory that satisfies vehicle dynamics constraints, avoids collisions, and considers passenger comfort. Motion planning typically involves path planning and velocity planning, and can employ sampling-based methods, graph search methods, or numerical optimization methods. Autonomous driving systems need to be able to assess driving risks, including comparative analysis of time-based, acceleration-based, and distance-based metrics. Probabilistic assessment methods use probabilistic models to describe the motion of traffic participants and thus assess risk. Furthermore, potential field theory-based assessment methods are also used to describe collision risk. Current decision-making algorithms may not adequately consider multiple safety mechanisms, which could lead to safety risks during the planning process.

[0005] Safety challenges in autonomous driving include uncertainties in perception and localization systems, complexity in decision-making and planning systems, and reliability of navigation and control systems. Future research directions may include improving the ability to identify and handle complex environments, enhancing collaboration with human drivers, and constructing a holistic, collaborative, and integrated safety framework. The decision-making algorithm for autonomous driving is a complex system that needs continuous evolution to adapt to constantly changing traffic environments and ensure safe and efficient decision-making under various conditions.

[0006] In the field of autonomous driving decision-making and control, Markov decision processes are a commonly used decision-making model. However, existing autonomous driving decision-making and control technologies still have some problems: First, existing decision-making algorithms often lack multiple safety mechanisms, posing safety risks to planning. Although some methods use safety distance models for safety assessment, they lack accurate calculation of safety distances in different scenarios and multi-level safety check mechanisms, leading to potential safety hazards in complex traffic environments. Second, decision-making algorithms are inefficient and fail to effectively eliminate unreasonable decision results. Existing decision tree construction methods often lack consideration for human driving characteristics, resulting in decision trees containing a large number of decision sequences that do not conform to actual driving habits, increasing computational complexity and potentially leading to unreasonable decision results. Finally, existing decision-making algorithms fail to fully simulate the thinking patterns of human drivers, resulting in unnatural and unsafe decision results. Although some methods consider vehicle interaction behavior and driving style, they lack systematic modeling and application of human driving characteristics, causing the behavior of autonomous vehicles to differ from that of human drivers, affecting the acceptability and safety of autonomous driving systems. Summary of the Invention

[0007] To address the shortcomings of existing autonomous driving decision-making algorithms, such as the lack of multiple safety mechanisms, low decision-making efficiency, and failure to simulate human driver thinking, and to improve decision-making safety, algorithm efficiency, and driving experience, this invention provides an autonomous driving decision-making control method, system, and related equipment. This method comprehensively considers safety, efficiency, and human driving characteristics to enhance the safety, reliability, and naturalness of the autonomous driving system.

[0008] The technical solution adopted by this invention to solve its technical problem is: In a first aspect, the present invention provides an autonomous driving decision-making and control method, comprising: Based on the vehicle status input from the positioning, an initial behavior decision tree is generated using a Markov algorithm; The sequence of the initial behavior decision tree is pruned and reconstructed according to human driving characteristics to obtain a pre-selected behavior decision tree; Using the RSS model, unsafe behavior decision sequences in the pre-screened behavior decision tree are pruned and removed. Then, combined with the vehicle kinematics model and behavior decision tree, the vehicle candidate trajectory corresponding to each behavior sequence in the vehicle decision tree is iteratively generated. The sequence of candidate vehicle trajectories is used to generate a predicted trajectory. Perform safety speed checks and collision checks on the predicted trajectory, and eliminate behavioral sequences that do not meet the safety speed checks and collision checks to obtain the pre-selected trajectory; Calculate the total cost of each pre-selected trajectory, and output the behavior sequence with the minimum total cost as the optimal vehicle decision result.

[0009] As a further improvement of the present invention, the step of generating an initial behavior decision tree using a Markov algorithm based on the vehicle state input from the positioning input includes: Obtain the vehicle's status; Based on the vehicle state, an initial behavior decision tree is generated using the Markov method, and the action set of the vehicle decision process is constructed as A={LLC,LK,RLC}, where LLC represents changing lanes to the left, LK represents keeping the original lane, and RLC represents changing lanes to the right.

[0010] As a further improvement of the present invention, the step of pruning and reconstructing the sequence of the initial behavior decision tree according to human driving characteristics to obtain a pre-selected behavior decision tree includes: The behavior sequence of vehicle change behavior is pruned based on human driving characteristics, and all unreasonable behavior sequences in the overall behavior decision tree are pruned; unreasonable behaviors include sequences that simultaneously involve left and right lane changes and sequences that involve multiple vehicle change behaviors. After running the optimal decision sequence for one decision cycle, the behavior decision tree is reconstructed as the pre-selected behavior decision tree; Among these characteristics are: Heuristic rule 1: No two behavioral changes are allowed within a single decision-making time domain; Heuristic rule 2: In a single decision-making time domain, lane changes from left to right or from right to left are not allowed.

[0011] As a further improvement to the present invention, the RSS model is divided into longitudinal safety distance and lateral safety distance; Longitudinal safety distance d min The distance at which a collision can still be avoided even in the worst-case scenario is as follows:

[0012] in, v r It's the speed of the car behind. ρ It is the reaction time. a max It is the maximum acceleration. β min It's a comfortable deceleration. v f It is the speed of the vehicle in front, β max Maximum deceleration; Lateral safety distance: Between cars c1 and c2 traveling at lateral velocities v1 and v2, if both cars apply maximum lateral acceleration towards each other within the time interval [0, ρ], and then both cars apply minimum lateral deceleration until their relative lateral velocities are zero, a collision can be avoided under these conditions. The lateral safety distance is specifically defined as follows:

[0013] Where v1 is the speed of the car on the right, v2 is the speed of the car on the left, ρ is the time interval, and β is the velocity. 1,lat,min β is the minimum lateral deceleration of the vehicle on the right. 2,lat,min The minimum lateral deceleration of the vehicle on the left. a 1,max This represents the maximum lateral acceleration of the vehicle on the right. a 2,max The maximum lateral acceleration of the vehicle on the left. µ This represents the lateral distance between the last two cars.

[0014] As a further improvement of the present invention, the step of generating a predicted trajectory from the sequence of candidate vehicle trajectories includes: Based on the sequence of vehicle candidate trajectories, a pure tracking model, a vehicle kinematics model, and an intelligent driver IDM model are used to generate predicted trajectories in a rolling manner.

[0015] As a further improvement of the present invention, the step of performing a safety speed check on the predicted trajectory includes: The upper speed limit of the vehicle is determined based on the safe distance between the vehicle and the vehicle in front; the lower speed limit of the vehicle is determined based on the safe distance between the vehicle and the vehicle behind. If the vehicle speed exceeds the safe speed limit, adjust the vehicle speed to within the safe speed range and reduce the speed of each trajectory point in the candidate trajectory corresponding to the behavior sequence. If the vehicle speed is below the safe speed limit, adjust the vehicle speed to within the safe speed range and increase the speed of each trajectory point in the candidate trajectory corresponding to the behavior sequence.

[0016] As a further improvement of the present invention, the collision check includes: After obtaining the trajectory predictions of important surrounding vehicles and candidate trajectories of the vehicles based on the input of the prediction module, a collision safety check is performed by determining whether there is a collision risk between the vehicle and the important surrounding vehicles at each time in the decision time domain. Behavioral sequences that fail the collision safety check are marked as infeasible behavioral sequences and removed from the behavioral decision tree.

[0017] As a further improvement to the present invention, the function of the total cost is expressed as:

[0018] Je (τ) represents the vehicle's need to follow the globally optimal path as much as possible. When the vehicle needs to change lanes under special circumstances, there is also a lane-changing cost. Js (τ) represents high decision consistency, which encourages the current behavior sequence to be similar to the behavior sequence in the previous decision time domain. Jn (τ) represents the safety of vehicle driving as the most basic and important requirement. λ1, λ2, and λ3 are weight coefficients.

[0019] As a further improvement of the present invention, the step of taking the behavioral sequence with the minimum overall cost as the optimal vehicle decision result and outputting it includes: By calculating the total cost of each behavior sequence, the behavior sequence with the minimum total cost is taken as the optimal vehicle behavior sequence:

[0020] In the formula, J total (τ) represents the total cost.

[0021] In a second aspect, the present invention provides an autonomous driving decision control system, comprising: The decision tree generation module is used to generate an initial behavior decision tree based on the vehicle state input from the positioning input, using a Markov algorithm. The sequence pruning module is used to prune and reconstruct the sequence of the initial behavior decision tree according to human driving characteristics, so as to obtain a pre-selected behavior decision tree; The sequence elimination module is used to prune and eliminate unsafe behavior decision sequences in the pre-screened behavior decision tree using the RSS model. Then, combined with the vehicle kinematics model and behavior decision tree, iteratively generates the vehicle candidate trajectory corresponding to each behavior sequence in the vehicle decision tree. The trajectory generation module is used to generate a predicted trajectory from a sequence of candidate vehicle trajectories. The sequence elimination module is used to perform safety speed checks and collision checks on the predicted trajectory, and eliminate behavioral sequences that do not meet the safety speed checks and collision checks to obtain the pre-selected trajectory; The decision output module is used to calculate the overall cost of each pre-selected trajectory, and outputs the behavior sequence with the minimum overall cost as the optimal vehicle decision result.

[0022] Thirdly, the present invention provides an electronic device including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the autonomous driving decision control method.

[0023] Fourthly, the present invention provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the autonomous driving decision control method.

[0024] Fifthly, the present invention provides a computer program product, the computer program product including computer instructions, the computer instructions instructing a computer to execute the autonomous driving decision control method.

[0025] The beneficial effects of this invention are as follows: by introducing human-like thinking for decision tree pruning, unreasonable sequences are reduced, greatly improving algorithm efficiency; by introducing an RSS safety assessment model and performing speed and collision checks on the results, the safety and reliability of decisions are improved; through multiple safety check mechanisms (RSS model assessment, speed check, and collision check), the safety and reliability of autonomous driving decisions are ensured; by simulating the thinking patterns of human drivers, the decision results are made more natural, improving the driving experience; and through the design of the cost function, global path driving, decision consistency, and driving safety are balanced, making decisions more reasonable. Compared with existing technologies, the autonomous driving decision control method provided by this invention can effectively solve the problems of existing decision algorithms lacking multiple safety mechanisms, low decision efficiency, and failing to simulate the thinking patterns of human drivers, achieving safer, more efficient, and more natural autonomous driving decision control. Attached Figure Description

[0026] Figure 1 This is a flowchart of the autonomous driving decision control method in the embodiments of this application; Figure 2 This is a flowchart of the MDP decision-making process in the embodiments of this application; Figure 3 This is a flowchart of the pruning and reconstruction process in the embodiments of this application; Figure 4 This is an overall decision-making flowchart in the embodiments of this application; Figure 5 The present invention provides an autonomous driving decision control system; Figure 6 This is a schematic diagram of an electronic device provided by the present invention. Detailed Implementation

[0027] The technical solution of the present invention will now be clearly and completely described with reference to the accompanying drawings and embodiments. Obviously, the described embodiments are merely some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0028] Example 1 like Figure 1 As shown, an embodiment of the present invention provides an autonomous driving decision control method, including: S101, Based on the vehicle status input from the positioning, an initial behavior decision tree is generated using a Markov algorithm; Based on Markov decision processes, this method abstracts vehicle states (position, speed, environmental perception, etc.) into a state space. It then generates all possible driving behavior sequences (such as lane changing, acceleration, deceleration, etc.) through state transition probabilities, forming an initial decision tree. This method systematically exhaustively enumerates driving possibilities, avoiding omissions of key scenarios and providing a complete decision space for subsequent optimization.

[0029] Among them, the Markov Decision Process (MDP) is a core mathematical framework in reinforcement learning and dynamic decision-making problems, used to model the optimization problem in which an agent achieves its goal through sequential decision-making in an uncertain environment.

[0030] S102, the sequence of the initial behavior decision tree is pruned and reconstructed according to the characteristics of human driving to obtain a pre-selected behavior decision tree; By pruning and reconstructing, behavioral sequences that do not conform to human driving habits (such as frequent sudden braking and unreasonable lane changes) are removed, while decision-making paths that conform to traffic rules and common driving sense are retained. This step improves the humanization of decision-making and reduces the risks caused by machine decisions not conforming to human expectations (such as rear-end collisions or pedestrian misunderstandings).

[0031] S103 uses the RSS model to prune and remove unsafe behavior decision sequences in the pre-selected behavior decision tree, and then combines the vehicle kinematics model and behavior decision tree to iteratively generate vehicle candidate trajectories corresponding to each behavior sequence in the vehicle decision tree. RSS (Responsibility) - The Sensitive Safety (RSS) model is used to evaluate vehicle driving safety. It defines the safe distances that a vehicle should maintain under various conditions. The RSS model is divided into longitudinal safety distance and lateral safety distance.

[0032] RSS safety model: Defines safe distance thresholds (e.g., longitudinal / lateral safe distances) and eliminates behavioral sequences that may lead to collisions. Vehicle kinematics model: Combines vehicle dynamics parameters (e.g., acceleration limits, turning radius) to generate physically feasible candidate trajectories. This dual-safety approach avoids logically dangerous decisions while ensuring the trajectory conforms to the vehicle's physical characteristics.

[0033] S104, Generate a predicted trajectory from the sequence of candidate vehicle trajectories; S105, Perform safety speed checks and collision checks on the predicted trajectory, and eliminate behavioral sequences that do not meet the safety speed checks and collision checks to obtain the pre-selected trajectory; Multiple predicted trajectories (such as different speed curves and path planning) are generated based on candidate behavior sequences. Pre-selected trajectories are filtered through speed checks (such as speeding detection) and collision checks (such as time collision risk assessment). Redundancy schemes can be provided to enhance system fault tolerance and adapt to complex dynamic environments.

[0034] S106, calculate the total cost of each pre-selected trajectory, and output the behavior sequence with the minimum total cost as the optimal vehicle decision result.

[0035] Define a multi-objective cost function (such as time efficiency, energy consumption, comfort, and safety weights), calculate the overall cost of each pre-selected trajectory, and select the trajectory with the minimum cost. This can achieve a balance between multiple objectives and avoid suboptimal decisions caused by optimizing a single indicator (such as excessively pursuing speed at the expense of safety).

[0036] Therefore, the method of this invention significantly reduces accident risk through dual filtering of RSS model and collision check. It avoids generating physically infeasible dangerous trajectories by combining kinematic models. Pruning and reconstruction make decisions conform to human driving habits, improving passenger trust and road cooperation. A closed loop is formed from decision tree generation to optimal output, covering the entire driving logic chain. The cost function flexibly adapts to different scenario requirements (e.g., prioritizing efficiency on highways, prioritizing safety in urban areas). The multi-trajectory generation mechanism enhances the ability to respond to unexpected situations (e.g., the sudden appearance of obstacles). Through multi-level screening and optimization of decision sequences, the safety and rationality of autonomous driving decisions are improved.

[0037] Therefore, this invention's method, through a hierarchical optimization mechanism, balances the rational decision-making of autonomous driving systems with the intuitive cognition of human drivers, providing technical support for safe and efficient driving in complex scenarios. Examples of applicable scenarios include: In congested urban areas: The system prioritizes smooth following trajectories to avoid passenger discomfort caused by frequent starts and stops.

[0038] Overtaking on highways: The overtaking distance is ensured by using an RSS model, and a smooth lane change trajectory is generated by combining a kinematic model.

[0039] Emergency obstacle avoidance: Quickly generate multiple avoidance trajectories and select the shortest and safest path through a cost function.

[0040] Specifically, such as Figure 4 As shown, each step of the present invention will be described in detail.

[0041] S101, Based on the vehicle state input from the positioning, an initial behavior decision tree is generated using a Markov algorithm, including: Obtain the vehicle's status; An initial behavior decision tree is generated based on the vehicle state using a Markov method, such as... Figure 2 As shown, the action set of the vehicle decision-making process is constructed as A={LLC,LK,RLC}, where LLC represents changing lanes to the left, LK represents keeping the original lane, and RLC represents changing lanes to the right.

[0042] In this embodiment, vehicle status includes information such as vehicle position, speed, acceleration, and heading angle. This information can be acquired through various sensors on the vehicle, such as GPS (Global Positioning System), Inertial Measurement Unit (IMU), and wheel encoders. Vehicle status information is the fundamental data for the autonomous driving system to make decisions.

[0043] A Markov algorithm is a decision-making method based on a probabilistic model. It assumes that the next state of a system depends only on the current state and not on previous states. In autonomous driving decision control, a Markov algorithm can predict possible future behavior sequences based on the current vehicle state and construct a decision tree.

[0044] The initial behavior decision tree is a multi-level tree structure, where each node represents a possible behavior decision, and each path from the root node to a leaf node represents a complete behavior sequence. In this embodiment, the depth of the behavior decision tree is 3, meaning it considers behavior sequences over three future decision cycles. Each decision cycle lasts for 1 second, therefore the entire decision time domain is 3 seconds.

[0045] The action set A = {LLC, LK, RLC} defines the three basic behaviors that a vehicle can take in each decision cycle: changing lanes to the left (LLC), staying in the current lane (LK), and changing lanes to the right (RLC). Based on these three basic behaviors, a complete behavior decision tree can be constructed. For example, for a decision tree with a depth of 3, there are a total of 27 possible behavior sequences, such as {LLC, LLC, LLC}, {LLC, LLC, LK}, {LLC, LLC, RLC}, etc.

[0046] S102, the sequence of the initial behavior decision tree is pruned and reconstructed according to human driving characteristics to obtain a pre-selected behavior decision tree, such as... Figure 3 As shown, it includes: The behavior sequence of vehicle change behavior is pruned based on human driving characteristics, and all unreasonable behavior sequences in the overall behavior decision tree are pruned; unreasonable behaviors include sequences that simultaneously involve left and right lane changes and sequences that involve multiple vehicle change behaviors. After running the optimal decision sequence for one decision cycle, the behavior decision tree is reconstructed as the pre-selected behavior decision tree; Among these characteristics are: Heuristic rule 1: No two behavioral changes are allowed within a single decision-making time domain; Heuristic rule 2: In a single decision-making time domain, lane changes from left to right or from right to left are not allowed.

[0047] In real-world driving, human drivers typically follow certain driving habits and rules, which can be used to optimize the decision-making process of autonomous driving systems. By simulating human driving characteristics, some behavioral sequences that do not conform to actual driving habits can be eliminated, thereby reducing computational load and improving the rationality of decisions.

[0048] Heuristic rule one stipulates that no two behavioral changes are allowed within a single decision-making time domain. For example, if the behavioral sequence {LLC, LK, LLC} contains two behavioral changes within a single decision-making time domain (from changing lanes to the left to keeping the original lane, and then from keeping the original lane to changing lanes to the left), this behavioral sequence will be pruned. This is because frequent behavioral changes not only increase driving difficulty but may also lead to unstable vehicle control and increase the risk of accidents.

[0049] Heuristic Rule 2 stipulates that lane changes from left to right or right to left are not allowed within a single decision-making time domain. For example, the behavior sequence {LLC,RLC,LK}, which includes a direct change from a left lane change to a right lane change, will also be pruned. This is because in actual driving, vehicles usually need to return to their original lane before making a lane change in the other direction; a direct change from left to right (or from right to left) does not conform to safe driving standards.

[0050] By applying these two heuristics, the initial 27 behavioral sequences can be reduced to fewer reasonable sequences. For example, it may leave behavioral sequences that conform to human driving characteristics, such as {LK,LK,LK}, {LLC,LLC,LLC}, {RLC,RLC,RLC}, {LK,LLC,LLC}, and {LK,RLC,RLC}.

[0051] After driving through one decision cycle using the optimal decision sequence, the system reconstructs the behavior decision tree. This is because the optimal decision sequence needs to be continuously updated as the vehicle's state changes. By periodically reconstructing the behavior decision tree, the real-time nature and adaptability of decision-making can be ensured.

[0052] S103 uses the RSS model to prune and remove unsafe behavior decision sequences in the pre-selected behavior decision tree. Then, iteratively generating vehicle candidate trajectories corresponding to each behavior sequence in the vehicle decision tree by combining the vehicle kinematics model and the behavior decision tree.

[0053] Longitudinal safety distance d min The distance at which a collision can still be avoided even in the worst-case scenario is as follows:

[0054] in, v r It's the speed of the car behind. ρ It is the reaction time. a max It is the maximum acceleration. β min It's a comfortable deceleration. v f It is the speed of the vehicle in front, β max Maximum deceleration; Lateral safety distance: Between cars c1 and c2 traveling at lateral velocities v1 and v2, if both cars apply maximum lateral acceleration towards each other within the time interval [0, ρ], and then both cars apply minimum lateral deceleration to brake until their relative lateral velocities are zero, a collision can be avoided under these conditions. The lateral safety distance is specifically defined as follows:

[0055] Where v1 is the speed of the car on the right, v2 is the speed of the car on the left, ρ is the time interval, and β is the velocity. 1,lat,min β is the minimum lateral deceleration of the vehicle on the right. 2,lat,min The minimum lateral deceleration of the vehicle on the left. a 1,max This represents the maximum lateral acceleration of the vehicle on the right. a 2,max The maximum lateral acceleration of the vehicle on the left. µ This represents the lateral distance between the last two cars.

[0056] The safety of each behavior sequence can be evaluated using the RSS model. If a behavior sequence fails to meet the safe distance requirement at any time, then that behavior sequence will be pruned. For example, if the behavior sequence {LK, LLC, LLC} results in the vehicle's distance from the vehicle in front being less than the longitudinal safe distance, or its distance from the vehicle in the adjacent lane being less than the lateral safe distance, then that behavior sequence will be pruned.

[0057] After pruning unsafe behavior sequences, the system uses a vehicle kinematics model to generate corresponding candidate vehicle trajectories for each remaining behavior sequence. The vehicle kinematics model describes the vehicle's motion characteristics, including the relationships between parameters such as speed, acceleration, and steering angle. Using the vehicle kinematics model, the vehicle's trajectory can be predicted based on the behavior sequence.

[0058] The vehicle kinematics model adopts a bicycle model, which simplifies the vehicle into a two-dimensional model with one front wheel and one rear wheel. The model's state variables include the vehicle's position (x, y), heading angle θ, velocity v, and steering angle δ. The state update equations are as follows: x(t+1) = x(t) + v(t)·cos(θ(t))·Δt y(t+1) = y(t) + v(t)·sin(θ(t))·Δt θ(t+1) = θ(t) + v(t)·tan(δ(t)) / L·Δt v(t+1) = v(t) + a(t)·Δt Where L is the vehicle wheelbase, a(t) is the acceleration, and Δt is the time step.

[0059] For each behavior sequence, the system iteratively calculates the vehicle's state at each time interval based on the vehicle's kinematics model, thereby generating a complete candidate trajectory. For example, for the behavior sequence {LK,LK,LK}, the system generates a trajectory that remains in the current lane; for the behavior sequence {LLC,LLC,LLC}, the system generates a trajectory that changes lanes to the left.

[0060] S104, Generate a predicted trajectory from the sequence of candidate vehicle trajectories, including: Based on the sequence of vehicle candidate trajectories, a pure tracking model, a vehicle kinematics model, and IDM rolling are used to generate predicted trajectories.

[0061] After generating candidate vehicle trajectories, the system needs to further refine these trajectories to generate more accurate predicted trajectories. This process employs a combination of a pure tracking model, a vehicle kinematics model, and an Intelligent Driver Model (IDM).

[0062] The pure tracking model is a simple yet effective path tracking controller that calculates the required steering angle based on the lateral deviation between the vehicle's current position and the target path. The control law of the pure tracking model is as follows: δ = arctan(2·L·e / (Kp·v 2 )) Where δ is the steering angle, L is the vehicle wheelbase, e is the lateral deviation, Kp is the control gain, and v is the vehicle speed.

[0063] In this embodiment, the control gain Kp of the pure tracking model is set to 1.0, and the vehicle wheelbase L is 2.8 meters.

[0064] An IDM (Intermediate Dynamics Model) is a model used to simulate the longitudinal motion of a vehicle. It considers the relative distance and relative velocity between the vehicle and the vehicle in front, and calculates the required acceleration. The acceleration calculation formula for an IDM is as follows: a = a max ·[1 - (v / v0) 4 - (s / s) 2 ] Where a is acceleration, a max v is the maximum acceleration, v0 is the current velocity, v0 is the desired velocity, and s is the actual distance to the vehicle in front. Let s be the desired safe distance. The calculation formula is: s = s0 + v·T + v·Δv / (2· ) Where s0 is the minimum safe distance, T is the safe time interval, and Δv is the relative speed with respect to the vehicle in front.

[0065] In this embodiment, the parameters of IDM are set as follows: maximum acceleration a max 3.0 m / s 2 The desired speed v0 is determined based on the road speed limit, the minimum safe distance s0 is 2.0 meters, and the safe time interval T is 1.5 seconds.

[0066] By combining a pure tracking model, a vehicle kinematics model, and an IDM (Inductively Coupled Model), the system can generate accurate predicted trajectories that take into account path trajectories, vehicle dynamics, and traffic flow characteristics. These predicted trajectories will be used for subsequent safety checks and decision optimization.

[0067] S105, Perform a safety speed check on the predicted trajectory, including: The upper speed limit of the vehicle is determined based on the safe distance between the vehicle and the vehicle in front; the lower speed limit of the vehicle is determined based on the safe distance between the vehicle and the vehicle behind. If the vehicle speed exceeds the safe speed limit, adjust the vehicle speed to within the safe speed range and reduce the speed of each trajectory point in the candidate trajectory corresponding to the behavior sequence. If the vehicle speed is below the safe speed limit, adjust the vehicle speed to within the safe speed range and increase the speed of each trajectory point in the candidate trajectory corresponding to the behavior sequence.

[0068] Speed ​​safety checks are a crucial step in ensuring vehicle safety. By calculating the upper and lower limits of safe speed, the system can determine whether the current speed is within a safe range and make adjustments as necessary.

[0069] The safe speed limit is calculated based on the safe distance between the vehicle and the vehicle in front. If the distance between the vehicle and the vehicle in front is d... front So, the safe speed limit v max It can be calculated using the following formula: v max =

[0070] Where βmin is the comfortable deceleration and s0 is the minimum safe distance.

[0071] The lower limit of safe speed is calculated based on the safe distance between the vehicle and the vehicle behind. If the distance between the vehicle and the vehicle behind is d... rear Then the lower limit of safe speed v min It can be calculated using the following formula: v min = v rear

[0072] Among them, v rear β is the speed of the following vehicle, βmax is the maximum deceleration, and s0 is the minimum safe distance.

[0073] In this embodiment, the minimum safe distance s0 is set to 2.0 meters, and the comfortable deceleration βmin is 2.0 m / s². 2 The maximum deceleration βmax is 8.0 m / s². 2 .

[0074] If the vehicle's speed in the predicted trajectory exceeds the safe speed limit, the system will reduce the speed of each point in the trajectory to ensure it does not exceed the safe speed limit. For example, if the speed of a trajectory point is 25 m / s and the safe speed limit is 20 m / s, the system will adjust the speed of that point to 20 m / s.

[0075] Similarly, if the vehicle's speed in the predicted trajectory is below the safe speed limit, the system will increase the speed of each point in the trajectory to ensure it is not below the safe speed limit. For example, if the speed of a trajectory point is 10 m / s and the safe speed limit is 15 m / s, the system will adjust the speed of that point to 15 m / s.

[0076] By checking the safe speed, the system can ensure that the vehicle's speed is always within a safe range, avoiding the risk of collision due to improper speed.

[0077] In S105, collision checks include: After obtaining the trajectory predictions of important surrounding vehicles and candidate trajectories of the vehicles based on the input of the prediction module, a collision safety check is performed by determining whether there is a collision risk between the vehicle and the important surrounding vehicles at each time in the decision time domain. Behavioral sequences that fail the collision safety check are marked as infeasible behavioral sequences and removed from the behavioral decision tree.

[0078] Collision checks are another crucial step in ensuring driving safety. By comparing the predicted trajectory of a vehicle with that of surrounding vehicles, the system can determine whether there is a collision risk and eliminate potential collision sequences.

[0079] In this embodiment, collision detection employs a rectangular bounding box-based method. The vehicle is simplified as a rectangle with a length of 4.8 meters and a width of 1.8 meters. For each time step, the system calculates the position of the vehicle and surrounding important vehicles and checks whether their bounding boxes overlap. If overlap exists at any time step, a collision risk is considered to exist.

[0080] Specifically, let the coordinates of the four vertices of the rectangular bounding box of vehicle A be (x, y, y). A1 , y A1 ), (x A2 , y A2 ), (x A3 ,y A3 ), (x A4 , y A4 The coordinates of the four vertices of the bounding box of vehicle B are (x, y, y). B1 , y B1 ), (x B2 , y B2 (xB3, y) B3 (xB4, yB4). Two rectangular bounding boxes are considered not to overlap if one of the following conditions is met: 1) The rightmost boundary of vehicle A is less than the leftmost boundary of vehicle B: max(x A1 , x A2 , x A3 , x A4 ) <min(x B1 ,x B2 , x B3 , x B4 ) 2) The leftmost boundary of vehicle A is greater than the rightmost boundary of vehicle B: min(x A1 , x A2 , x A3 , x A4 )>max(xB1 ,x B2 , x B3 , x B4 ) 3) The uppermost boundary of vehicle A is less than the lowermost boundary of vehicle B: max(y A1 , y A2 , y A3 , y A4 ) <min(y B1 ,y B2 , y B3 , y B 4 ) 4) The lowest boundary of vehicle A is greater than the highest boundary of vehicle B: min(y A1 , y A2 , y A3 , y A4 )>max(y B1 ,y B2 , y B3 , y B4 ) If none of the above four conditions are met, the two rectangular bounding boxes are considered to overlap, and there is a risk of collision.

[0081] For each predicted trajectory corresponding to a behavior sequence, the system performs a collision check at each time step within the decision-making time domain. If a collision risk is detected at any time step, the behavior sequence will be marked as infeasible and removed from the behavior decision tree.

[0082] By performing collision checks, the system can ensure that the selected sequence of actions will not lead to a collision with surrounding vehicles, thereby guaranteeing driving safety.

[0083] S106, calculate the total cost of each pre-selected trajectory, and output the behavior sequence with the minimum total cost as the optimal vehicle decision result.

[0084] The overall cost can be expressed as a function: J total (τ) = λ1·Je(τ) + λ2·Js(τ) + λ3·Jn(τ) Je(τ) represents the path that the vehicle needs to follow as closely as possible to the globally optimal path. There is also a cost associated with changing lanes when the vehicle needs to do so under special circumstances. Js(τ) represents high decision consistency, encouraging the current behavior sequence to be similar to the behavior sequence in the previous decision time domain; Jn(τ) represents the most basic and important requirement for vehicle driving safety, and a safe distance model is used to measure collision risk. λ1, λ2, and λ3 are weighting coefficients.

[0085] After completing safety speed and collision checks, the system obtains a series of pre-selected trajectories, all of which are safe and feasible. Next, the system needs to select the optimal trajectories from these pre-selected trajectories as the final decision. The selection criterion is the overall cost function, which considers multiple factors.

[0086] Je(τ) is the path cost, which measures the deviation of the vehicle's travel path from the globally optimal path. The globally optimal path is typically generated by the path planning module, taking into account factors such as road network, traffic rules, and destination. The formula for calculating the path cost is: Je(τ) = wd·d lat + w θ ·|θ - θ ref | + w c ·c lane Where, d lat θ represents the lateral deviation of the vehicle from the reference path, and θ represents the vehicle's heading angle. ref The heading angle of the reference path, c lane For lane-changing costs (if the action sequence includes lane-changing actions, then c) lane (If it is 1, otherwise it is 0), w d w θ wc are weighting coefficients.

[0087] In this embodiment, the weighting coefficients for path costs are set as follows: w d = 1.0, w θ = 0.5, w c = 2.0.

[0088] Js(τ) is the decision consistency cost, which encourages the current behavior sequence to be similar to the behavior sequence in the previous decision time domain, thus avoiding frequent decision changes. The formula for calculating the decision consistency cost is: Js(τ) = w s ·d s (τ, τ prev ) Where, d s (τ, τ prev ) represents the current behavior sequence τ and the behavior sequence τ in the previous decision time domain. prev The degree of difference between them, w s The weighting coefficients. Dispersion d s The calculation method is as follows: d s (τ, τ prev ) = Σi I(τ i ≠ τ previ ) / n Where I(·) is an indicator function, which is 1 when the condition is true and 0 otherwise; n is the length of the action sequence; τ i and τ previ These represent the actions at the i-th position of the current action sequence and the action sequence within the previous decision time domain, respectively.

[0089] In this embodiment, the weighting coefficient ws for the decision consensus cost is set to 3.0.

[0090] Jn(τ) is the safety cost, which measures the safety of vehicle operation and is calculated using a safe distance model to determine the collision risk. The formula for calculating the safety cost is: Jn(τ) = w front ·max(0, d safefront- d front ) + w rear ·max(0, d saferear- d rear ) +w lat ·max(0, d safelat- d lat ) Where, d safefront d saferear and d safelat These represent the safe distances from the vehicle in front, behind, and to the side, respectively, d front d rear and d lat w represents the actual distance to the vehicle in front, behind, and to the side, respectively. front w rear and w lat These are the weighting coefficients.

[0091] In this embodiment, the weighting coefficient for security cost is set as follows: w front = 2.0, w rear = 1.0, w lat = 1.5.

[0092] The weighting coefficients λ1, λ2, and λ3 of the overall cost function are set to 0.3, 0.2, and 0.5, respectively, reflecting the important role of security in the decision-making process.

[0093] The sequence of actions that minimizes the overall cost is taken as the optimal vehicle decision result and output, including: By calculating the total cost of each behavior sequence, the behavior sequence with the minimum total cost is selected as the optimal vehicle behavior sequence. τ =

[0094] In the formula, J total (τ) represents the total cost.

[0095] After calculating the total cost of each pre-selected trajectory, the system selects the behavior sequence with the minimum total cost as the optimal vehicle decision. This process can be represented as a minimization problem: τ =

[0096] Where, τ For the optimal action sequence, J total (τ) represents the total cost of the behavior sequence τ.

[0097] For example, assuming that after safety speed checks and collision checks, the remaining pre-selected trajectories correspond to the following behavior sequences: {LK,LK,LK}, {LLC,LLC,LLC}, and {RLC,RLC,RLC}, with total costs of 5.2, 6.8, and 4.5 respectively. The system will then select the behavior sequence {RLC,RRC,RRC} with the lowest total cost as the optimal vehicle decision.

[0098] The optimal vehicle decision result will be output to the control module, which will generate specific control commands based on this decision result, such as steering angle and acceleration, to control the vehicle to drive according to the decision result.

[0099] In practical applications, the system re-executes the decision-making process every decision cycle (typically 0.1 seconds), updating the decision results based on the latest vehicle status and environmental information. This rolling optimization approach ensures the real-time nature and adaptability of the decisions.

[0100] Example 2 like Figure 5 As shown, an embodiment of the present invention provides an autonomous driving decision control system, comprising: The decision tree generation module 100 is used to generate an initial behavior decision tree based on the vehicle state input by the positioning input using a Markov algorithm. The sequence pruning module 200 is used to prune and reconstruct the sequence of the initial behavior decision tree according to human driving characteristics to obtain a pre-selected behavior decision tree; The sequence elimination module 300 is used to prune and eliminate unsafe behavior decision sequences in the pre-screened behavior decision tree through the RSS model, and then combine the vehicle kinematics model and behavior decision tree to iteratively generate the vehicle candidate trajectory corresponding to each behavior sequence in the vehicle decision tree. The trajectory generation module 400 is used to generate a predicted trajectory from a sequence of candidate vehicle trajectories. The sequence elimination module 500 is used to perform safety speed checks and collision checks on the predicted trajectory, and eliminate behavioral sequences that do not meet the safety speed checks and collision checks to obtain the pre-selected trajectory; The decision output module 600 is used to calculate the overall cost of each pre-selected trajectory, and outputs the behavior sequence with the minimum overall cost as the optimal vehicle decision result.

[0101] In this embodiment, the autonomous driving decision control system is a software system composed of multiple functional modules. It can be deployed on an onboard computing platform, such as a high-performance computer or a dedicated autonomous driving computing unit. The various modules of the system communicate with each other through data interfaces to jointly complete the autonomous driving decision control task.

[0102] The decision tree generation module is the starting point of the system. It receives vehicle state information from the localization module, including position, speed, acceleration, and heading angle. Based on this information, the decision tree generation module uses a Markov algorithm to generate an initial behavior decision tree. The Markov algorithm assumes that the next state of the system depends only on the current state and not on previous states. This characteristic makes it suitable for autonomous driving decision control.

[0103] The initial behavior decision tree is a multi-level tree structure, where each node represents a possible behavior decision, and each path from the root node to a leaf node represents a complete behavior sequence. In this embodiment, the depth of the behavior decision tree is 3, meaning it considers behavior sequences over three future decision cycles. Each decision cycle lasts for 1 second, therefore the entire decision time domain is 3 seconds.

[0104] The decision tree generation module constructs a set of actions for the vehicle decision-making process, A={LLC,LK,RLC}, where LLC represents changing lanes to the left, LK represents staying in the original lane, and RLC represents changing lanes to the right. Based on these three basic behaviors, the decision tree generation module can construct a complete behavioral decision tree. For example, for a decision tree of depth 3, there are a total of 27 possible behavioral sequences.

[0105] The sequence pruning module receives the initial behavior decision tree from the decision tree generation module and prunes and reconstructs it based on human driving characteristics. These characteristics include two heuristic rules: first, two behavior changes are not allowed within a single decision time domain; second, lane changes from left to right or right to left are not allowed within a single decision time domain.

[0106] By applying these two heuristics, the sequence pruning module can reduce the initial 27 behavior sequences to fewer reasonable behavior sequences. For example, the behavior sequence {LLC, LK, LLC}, which contains two behavior changes within a decision time domain, will be pruned; the behavior sequence {LLC, RLC, LK}, which contains a behavior that changes directly from changing lanes to changing lanes to the right, will also be pruned.

[0107] The sequence elimination module receives a pre-filtered behavior decision tree from the sequence pruning module and uses an RSS model to prune and eliminate unsafe behavior decision sequences. The RSS model consists of longitudinal safety distance and lateral safety distance, used to evaluate the safety of vehicle driving.

[0108] Using the RSS model, the sequence elimination module can assess the safety of each behavioral sequence and eliminate unsafe behavioral sequences.

[0109] After eliminating unsafe behavior sequences, the sequence elimination module combines the vehicle kinematics model to generate corresponding candidate vehicle trajectories for each remaining behavior sequence. The vehicle kinematics model uses a bicycle model, which simplifies the vehicle into a two-dimensional model with a front wheel and a rear wheel. Using the vehicle kinematics model, the sequence elimination module can predict the vehicle's trajectory based on the behavior sequence.

[0110] The trajectory generation module receives candidate vehicle trajectories from the sequence elimination module and further refines these trajectories to generate more accurate predicted trajectories. This process employs a combination of a pure tracking model, a vehicle kinematics model, and an IDM (Integrated Vehicle Model).

[0111] The pure tracking model is a simple yet effective path tracking controller that calculates the required steering angle based on the lateral deviation between the vehicle's current position and the target path. The IDM (Intermediate Dynamics Model) is a model used to simulate the longitudinal motion of a vehicle; it considers the relative distance and relative velocity between the vehicle and the vehicle in front, and calculates the required acceleration.

[0112] By combining a pure tracking model, a vehicle kinematics model, and an IDM (Inductively Coupled Model), the trajectory generation module can generate accurate predicted trajectories that take into account path tracking, vehicle dynamics, and traffic flow characteristics.

[0113] The sequence elimination module is also responsible for performing safety speed checks and collision checks on the predicted trajectory, and eliminating behavioral sequences that do not meet safety requirements. The safety speed check involves calculating the upper and lower limits of the safe speed and adjusting the speed of each point in the trajectory to keep it within the safe range. The collision check determines whether there is a collision risk by comparing the predicted trajectory of the vehicle with the predicted trajectories of surrounding vehicles.

[0114] The decision output module receives pre-selected trajectories from the sequence elimination module, calculates the overall cost of each trajectory, and selects the behavior sequence with the minimum overall cost as the optimal vehicle decision result. The overall cost function considers three aspects: path cost, decision consistency cost, and safety cost, which correspond to the rationality of the vehicle's driving path, the stability of the decision, and the safety of the driving, respectively.

[0115] In practical applications, the autonomous driving decision control system re-executes the decision-making process in each decision cycle, updating the decision results based on the latest vehicle status and environmental information. This rolling optimization approach ensures the real-time nature and adaptability of the decisions.

[0116] Example 3 like Figure 6 As shown, an embodiment of the present invention provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the following autonomous driving decision control method: Based on the vehicle status input from the positioning, an initial behavior decision tree is generated using a Markov algorithm; The sequence of the initial behavior decision tree is pruned and reconstructed according to human driving characteristics to obtain a pre-selected behavior decision tree; Using the RSS model, unsafe behavior decision sequences in the pre-screened behavior decision tree are pruned and removed. Then, combined with the vehicle kinematics model and behavior decision tree, the vehicle candidate trajectory corresponding to each behavior sequence in the vehicle decision tree is iteratively generated. The sequence of candidate vehicle trajectories is used to generate a predicted trajectory. Perform safety speed checks and collision checks on the predicted trajectory, and eliminate behavioral sequences that do not meet the safety speed checks and collision checks to obtain the pre-selected trajectory; Calculate the total cost of each pre-selected trajectory, and output the behavior sequence with the minimum total cost as the optimal vehicle decision result.

[0117] In this embodiment, the electronic device may be an on-board computer, an autonomous driving control unit, or other computing device capable of executing autonomous driving decision-making and control methods. The electronic device includes a memory, a processor, and a computer program stored in the memory and executable on the processor.

[0118] Memory can include various types of storage media such as read-only memory (ROM), random access memory (RAM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), flash memory, hard disk, or optical disk. Memory is used to store operating systems, application programs, and computer programs that implement autonomous driving decision-making and control methods.

[0119] The processor can be a central processing unit (CPU), graphics processing unit (GPU), digital signal processor (DSP), application-specific integrated circuit (ASIC), field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The processor is responsible for executing the computer program stored in memory to implement the various steps of the autonomous driving decision-making and control method.

[0120] A computer program contains a sequence of instructions for implementing an autonomous driving decision-making and control method. When the processor executes these instructions, it performs the steps of the autonomous driving decision-making and control method in the order of the instruction sequence.

[0121] Specifically, when the processor executes the computer program, it first uses a Markov algorithm to generate an initial behavior decision tree based on the vehicle state input. The vehicle state includes information such as position, speed, acceleration, and heading angle, which can be obtained through various sensors on the vehicle. Based on the current vehicle state, the Markov algorithm predicts possible future behavior sequences and constructs a decision tree.

[0122] The initial behavior decision tree is a multi-level tree structure, where each node represents a possible behavior decision, and each path from the root node to a leaf node represents a complete behavior sequence. The depth of the behavior decision tree is 3, meaning it considers behavior sequences over three future decision cycles. Each decision cycle lasts 1 second, therefore the entire decision time domain is 3 seconds.

[0123] The set of actions in the vehicle's decision-making process is A = {LLC, LK, RLC}, where LLC represents changing lanes to the left, LK represents staying in the current lane, and RLC represents changing lanes to the right. Based on these three basic behaviors, a complete behavior decision tree can be constructed. For example, for a decision tree of depth 3, there are a total of 27 possible behavior sequences.

[0124] Next, the processor prunes and reconstructs the initial behavior decision tree sequence based on human driving characteristics, resulting in a pre-selected behavior decision tree. Human driving characteristics include two heuristic rules: first, no two behavior changes are allowed in a single decision time domain; second, no lane change from left to right or right to left is allowed in a single decision time domain.

[0125] By applying these two heuristics, the processor can reduce the initial 27 behavior sequences to fewer reasonable ones. For example, the behavior sequence {LLC, LK, LLC}, which contains two behavior changes within a single decision time domain, will be pruned; the behavior sequence {LLC, RLC, LK}, which contains a behavior that changes directly from a left lane change to a right lane change, will also be pruned.

[0126] Then, the processor uses the RSS model to prune and remove unsafe behavioral decision sequences from the pre-selected behavioral decision tree. The RSS model consists of longitudinal safety distance and lateral safety distance, used to evaluate the safety of vehicle driving.

[0127] Using the RSS model, the processor can assess the safety of each behavioral sequence and eliminate unsafe behavioral sequences.

[0128] After eliminating unsafe behavior sequences, the processor, combined with a vehicle kinematics model, generates corresponding candidate vehicle trajectories for each remaining behavior sequence. The vehicle kinematics model uses a bicycle model, which simplifies the vehicle into a two-dimensional model with one front wheel and one rear wheel. Using this model, the processor can predict the vehicle's trajectory based on the behavior sequence.

[0129] Next, the processor generates a predicted trajectory from the sequence of candidate vehicle trajectories. This process employs a combination of a pure tracking model, a vehicle kinematics model, and an IDM (Integrated Vehicle Model).

[0130] The pure tracking model is a simple yet effective path tracking controller that calculates the required steering angle based on the lateral deviation between the vehicle's current position and the target path. The IDM (Intermediate Dynamics Model) is a model used to simulate the longitudinal motion of a vehicle; it considers the relative distance and relative velocity between the vehicle and the vehicle in front, and calculates the required acceleration.

[0131] By combining a pure tracking model, a vehicle kinematics model, and an IDM, the processor can generate accurate predicted trajectories that take into account path tracing, vehicle dynamics, and traffic flow characteristics.

[0132] The processor then performs safety speed checks and collision checks on the predicted trajectory, eliminating behavioral sequences that do not meet safety requirements. The safety speed check involves calculating upper and lower safety speed limits and adjusting the speeds of points in the trajectory to keep them within safe limits. The collision check compares the vehicle's predicted trajectory with the predicted trajectories of surrounding vehicles to determine if a collision risk exists.

[0133] Finally, the processor calculates the overall cost of each pre-selected trajectory and selects the behavior sequence with the minimum overall cost as the optimal vehicle decision result. The overall cost function considers three aspects: path cost, decision consistency cost, and safety cost, which correspond to the rationality of the vehicle's driving path, the stability of the decision, and the safety of the driving, respectively.

[0134] In practical applications, electronic devices re-execute the decision-making process in each decision cycle, updating the decision results based on the latest vehicle status and environmental information. This rolling optimization approach ensures the real-time nature and adaptability of the decisions.

[0135] Example 4 A computer-readable storage medium storing a computer program, which, when executed by a processor, implements the following autonomous driving decision-making and control method: S101, Based on the vehicle status input from the positioning, an initial behavior decision tree is generated using a Markov algorithm; S102, the sequence of the initial behavior decision tree is pruned and reconstructed according to the characteristics of human driving to obtain a pre-selected behavior decision tree; S103 uses the RSS model to prune and remove unsafe behavior decision sequences in the pre-selected behavior decision tree, and then combines the vehicle kinematics model and behavior decision tree to iteratively generate vehicle candidate trajectories corresponding to each behavior sequence in the vehicle decision tree. S104, Generate a predicted trajectory from the sequence of candidate vehicle trajectories; S105, Perform safety speed checks and collision checks on the predicted trajectory, and eliminate behavioral sequences that do not meet the safety speed checks and collision checks to obtain the pre-selected trajectory; S106, calculate the total cost of each pre-selected trajectory, and output the behavior sequence with the minimum total cost as the optimal vehicle decision result.

[0136] In this embodiment, the computer-readable storage medium can be any tangible medium that contains or stores a program, such as a read-only memory (ROM), random access memory (RAM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), flash memory, magnetic disk, or optical disk. The computer-readable storage medium stores a computer program that includes a sequence of instructions for implementing an autonomous driving decision control method.

[0137] When a computer program in a computer-readable storage medium is executed by a processor, the processor executes the various steps of the autonomous driving decision-making and control method in the order of the instruction sequence. The specific execution process is the same as described in Embodiment 3, including generating an initial behavior decision tree based on the vehicle state input from the positioning, pruning and reconstructing the behavior decision tree and removing unwanted branches, generating a predicted trajectory, performing a safety check, and selecting the optimal vehicle decision result.

[0138] Computer-readable storage media can be integrated into onboard computers, autonomous driving control units, or other computing devices capable of executing autonomous driving decision-making and control methods. By storing computer programs in computer-readable storage media, autonomous driving decision-making and control methods can be easily deployed to different computing devices, improving the portability and reusability of the methods.

[0139] Example 5 A computer program product includes computer instructions that instruct a computer to execute the following autonomous driving decision control method: S101, Based on the vehicle status input from the positioning, an initial behavior decision tree is generated using a Markov algorithm; S102, the sequence of the initial behavior decision tree is pruned and reconstructed according to the characteristics of human driving to obtain a pre-selected behavior decision tree; S103 uses the RSS model to prune and remove unsafe behavior decision sequences in the pre-selected behavior decision tree, and then combines the vehicle kinematics model and behavior decision tree to iteratively generate vehicle candidate trajectories corresponding to each behavior sequence in the vehicle decision tree. S104, Generate a predicted trajectory from the sequence of candidate vehicle trajectories; S105, Perform safety speed checks and collision checks on the predicted trajectory, and eliminate behavioral sequences that do not meet the safety speed checks and collision checks to obtain the pre-selected trajectory; S106, calculate the total cost of each pre-selected trajectory, and output the behavior sequence with the minimum total cost as the optimal vehicle decision result.

[0140] In this embodiment, the computer program product includes computer instructions, which may be stored in a computer-readable storage medium or transmitted over a network. The computer instructions instruct the computer to execute various steps of the autonomous driving decision-making and control method.

[0141] When the computer executes these instructions, it performs the various steps of the autonomous driving decision-making and control method in the order of the instructions. The specific execution process is the same as described in Embodiments 3 and 4, including generating an initial behavior decision tree based on the vehicle state input from the positioning, pruning and reconstructing the behavior decision tree and removing unwanted branches, generating a predicted trajectory, performing a safety check, and selecting the optimal vehicle decision result.

[0142] Computer program products can be provided as software packages, applications, plugins, or other forms, and can be distributed through various channels, such as app stores, corporate intranets, or direct installation. By encapsulating autonomous driving decision-making and control methods as computer program products, they can be easily deployed on different computing devices, improving the accessibility and ease of use of the methods.

[0143] It should be noted that Embodiment 1, Embodiment 2, Embodiment 3, Embodiment 4, and Embodiment 5 are all types of autonomous driving decision control methods.

[0144] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. An autonomous driving decision-making and control method, characterized in that, include: Based on the vehicle status input from the positioning, an initial behavior decision tree is generated using a Markov algorithm; The sequence of the initial behavior decision tree is pruned and reconstructed according to human driving characteristics to obtain a pre-selected behavior decision tree; Using the RSS model, unsafe behavior decision sequences in the pre-screened behavior decision tree are pruned and removed. Then, combined with the vehicle kinematics model and behavior decision tree, the vehicle candidate trajectory corresponding to each behavior sequence in the vehicle decision tree is iteratively generated. The sequence of candidate vehicle trajectories is used to generate a predicted trajectory. Perform safety speed checks and collision checks on the predicted trajectory, and eliminate behavioral sequences that do not meet the safety speed checks and collision checks to obtain the pre-selected trajectory; Calculate the total cost for each pre-selected trajectory, and output the action sequence with the minimum total cost as the optimal vehicle decision result; The RSS model is divided into longitudinal safety distance and lateral safety distance; Longitudinal safety distance d min The distance at which a collision can still be avoided even in the worst-case scenario is as follows: in, v r It's the speed of the car behind. ρ It is a time interval. a max It is the maximum acceleration. β min It's a comfortable deceleration. v f It is the speed of the vehicle in front, β max It is the maximum deceleration; Lateral safety distance: based on lateral speed v 1. v Two cars, c1 and c2, are moving towards each other. If both cars apply maximum lateral acceleration within the time interval [0, ρ], and then both cars apply minimum lateral deceleration until their relative lateral velocities are zero, a collision can be avoided under these conditions. The lateral safe distance is... D min Specifically: in, v 1 represents the speed of the car on the right. v 2 represents the speed of the car on the left, β 1,lat,min β is the minimum lateral deceleration of the vehicle on the right. 2,lat,min This represents the minimum lateral deceleration of the vehicle on the left.

2. The autonomous driving decision-making and control method according to claim 1, characterized in that, The step of generating an initial behavior decision tree using a Markov algorithm based on the vehicle state input from the positioning input includes: Obtain the vehicle's status; Based on the vehicle state, an initial behavior decision tree is generated using the Markov method, and the action set of the vehicle decision process is constructed as A={LLC,LK,RLC}, where LLC represents changing lanes to the left, LK represents keeping the original lane, and RLC represents changing lanes to the right.

3. The autonomous driving decision-making and control method according to claim 1, characterized in that, The sequence of the initial behavior decision tree is pruned and reconstructed according to human driving characteristics to obtain a pre-selected behavior decision tree, including: The behavior sequence of vehicle change behavior is pruned based on human driving characteristics, and all unreasonable behavior sequences in the overall behavior decision tree are pruned; unreasonable behaviors include sequences that simultaneously involve left and right lane changes and sequences that involve multiple vehicle change behaviors. After running the optimal decision sequence for one decision cycle, the behavior decision tree is reconstructed as the pre-selected behavior decision tree; Among these characteristics are: Heuristic rule 1: No two behavioral changes are allowed within a single decision-making time domain; Heuristic rule 2: In a single decision-making time domain, lane changes from left to right or from right to left are not allowed.

4. The autonomous driving decision-making and control method according to claim 1, characterized in that, The process of generating a predicted trajectory from a sequence of candidate vehicle trajectories includes: Based on the sequence of vehicle candidate trajectories, a pure tracking model, a vehicle kinematics model, and an intelligent driver IDM model are used to generate predicted trajectories in a rolling manner.

5. The autonomous driving decision-making and control method according to claim 1, characterized in that, The safe speed check of the predicted trajectory includes: The upper speed limit of the vehicle is determined based on the safe distance between the vehicle and the vehicle in front; the lower speed limit of the vehicle is determined based on the safe distance between the vehicle and the vehicle behind. If the vehicle speed exceeds the safe speed limit, adjust the vehicle speed to within the safe speed range and reduce the speed of each trajectory point in the candidate trajectory corresponding to the behavior sequence. If the vehicle speed is below the safe speed limit, adjust the vehicle speed to within the safe speed range and increase the speed of each trajectory point in the candidate trajectory corresponding to the behavior sequence.

6. The autonomous driving decision-making and control method according to claim 1, characterized in that, The collision check includes: After obtaining the trajectory predictions of important surrounding vehicles and candidate trajectories of the vehicles based on the input of the prediction module, a collision safety check is performed by determining whether there is a collision risk between the vehicle and the important surrounding vehicles at each time in the decision time domain. Behavioral sequences that fail the collision safety check are marked as infeasible behavioral sequences and removed from the behavioral decision tree.

7. The autonomous driving decision-making and control method according to claim 1, characterized in that, The overall cost is expressed as a function: In the formula, The cost of changing lanes when a vehicle needs to change lanes under special circumstances, so that the vehicle travels along the globally optimal path. To improve the consistency of decision-making, it is encouraged that the current sequence of actions is similar to the sequence of actions in the previous decision-making time domain; The requirements for vehicle driving safety are λ1, λ2, and λ3, which are weighting coefficients.

8. The autonomous driving decision-making and control method according to claim 1, characterized in that, The step of taking the action sequence that minimizes the overall cost as the optimal vehicle decision result and outputting it includes: By calculating the total cost of each behavior sequence, the behavior sequence with the minimum total cost is selected as the optimal vehicle behavior sequence. In the formula, This represents the overall cost.

9. An autonomous driving decision control system, implementing the autonomous driving decision control method according to any one of claims 1 to 8; characterized in that, include: The decision tree generation module is used to generate an initial behavior decision tree based on the vehicle state input from the positioning input, using a Markov algorithm. The sequence pruning module is used to prune and reconstruct the sequence of the initial behavior decision tree according to human driving characteristics, so as to obtain a pre-selected behavior decision tree; The first sequence elimination module is used to prune and eliminate unsafe behavior decision sequences in the pre-screened behavior decision tree using the RSS model, and then combine the vehicle kinematics model and behavior decision tree to iteratively generate vehicle candidate trajectories corresponding to each behavior sequence in the vehicle decision tree. The trajectory generation module is used to generate a predicted trajectory from a sequence of candidate vehicle trajectories. The second sequence elimination module is used to perform safety speed checks and collision checks on the predicted trajectory, and eliminate behavioral sequences that do not meet the safety speed checks and collision checks to obtain the pre-selected trajectory. The decision output module is used to calculate the overall cost of each pre-selected trajectory, and outputs the behavior sequence with the minimum overall cost as the optimal vehicle decision result.

10. An electronic device, characterized in that, It includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the computer program, implements the autonomous driving decision control method according to any one of claims 1 to 6.

11. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the autonomous driving decision control method according to any one of claims 1 to 8.

12. A computer program product, the computer program product comprising computer instructions, characterized in that, The computer instructions instruct the computer to execute the autonomous driving decision control method according to any one of claims 1 to 8.

Images