Vehicle automatic driving decision planning method and system

Abstract

The invention belongs to the technical field of automatic driving, and provides a vehicle automatic driving decision-making planning method and system, the whole system comprises a sensing layer, a decision-making layer, a planning layer and a control layer, and all the layers cooperatively realize the whole process of sensing, decision-making, planning and control; the method comprises the steps of firstly collecting and preprocessing sensing data; constructing a dynamic perception fusion model, and fusing multi-sensor data to output target comprehensive information; constructing a multi-objective optimization function based on safety, efficiency and comfort indexes to generate a decision scheme; generating an initial path in combination with a high-precision map, and dynamically adjusting the path in the emergency scene based on the scene emergency degree; the vehicle runs according to the path and the decision, decelerates and re-plans the path in an emergency scene; according to the invention, the safety, efficiency and comfort of automatic driving can be effectively guaranteed. The method is suitable for complex traffic scenes such as urban roads and expressways, and can solve the problems that in the prior art, decisions are lagged, path planning does not consider multi-target requirements, and adaptability to emergency scenes is poor.

Description

Technical Field

[0001] This invention belongs to the field of autonomous driving technology, specifically relating to a method and system for decision-making and planning for autonomous driving of vehicles. Background Technology

[0002] Autonomous driving is a crucial direction for the current development of the automotive industry. It integrates various advanced technologies, aiming to achieve autonomous vehicle operation, reduce traffic accidents caused by human error, and improve road transport efficiency and safety. Decision-making and planning technology is a core component of autonomous driving, directly affecting the vehicle's ability to cope with complex and changing traffic environments and its driving safety. Current autonomous driving decision-making and planning technologies have the following shortcomings: 1. Disconnect between dynamic environment perception and decision-making: Existing technologies mostly use a single sensor (such as a camera or LiDAR) to acquire environmental information, and the perception data and decision-making module are not linked in real time. The high latency of a single sensor in recognizing sudden dynamic targets leads to delayed decision response and increases the risk of collision.

[0003] 2. Insufficient optimization of multi-objective decision-making: Existing decision-making schemes focus more on safety and do not take into account driving efficiency and ride comfort in a coordinated manner, resulting in frequent rapid acceleration and deceleration during actual driving and insufficient comfort.

[0004] 3. Poor adaptability of path planning in emergency scenarios: Existing path planning is mostly based on static path generation from preset maps. When faced with emergency scenarios, path replanning has high latency, and new paths are prone to problems such as path intersections and insufficient passage space, which cannot meet the real-time and safety requirements of autonomous driving.

[0005] In summary, existing technologies cannot simultaneously meet the collaborative requirements of real-time perception, multi-objective decision-making, and dynamic path planning in autonomous driving. There is an urgent need for a decision-making and planning solution that integrates dynamic perception, multi-objective optimization, and rapid path adjustment. Summary of the Invention

[0006] To address the aforementioned technical problems, this invention provides a method and system for decision-making and planning for autonomous driving of vehicles, thereby resolving the issues in the prior art. The technical solution adopted by this invention is as follows: A method for decision-making and planning for autonomous driving of vehicles includes the following steps: Step 1, Perception Data Acquisition and Preprocessing: The acquired perception data includes: target category, 3D coordinates, distance information, target velocity, and target motion direction; Step 2: Construct a dynamic perception fusion model to calculate the fusion weights of the perception data and dynamically output the comprehensive information of the target. Step 3: Construct a multi-objective optimization function based on safety, efficiency, and comfort indicators; generate a decision scheme based on the output of the multi-objective optimization function; Step 4: Based on the decision-making plan and high-precision map, execute the dynamic path planning adjustment mechanism to generate the initial path and the dynamically adjusted path under emergency scenarios; Step 5: The vehicle travels according to the generated path, controlling the steering, throttle, and braking systems. When an emergency occurs, the vehicle slows down and replans the path, enabling the vehicle to travel according to the planned path and decision-making scheme.

[0007] Furthermore, step 2 includes: The formula for constructing a dynamic perception fusion model is as follows: ; in, To integrate the weighting coefficients, For sensor type, For cameras, For lidar, For millimeter-wave radar; To account for the overall sensor accuracy, its value ranges from 0 to 1. For cameras, For lidar, For millimeter-wave radar; For environmental adaptability, its value ranges from 0 to 1, where: Sunny day: ;rain: ; These are the weighting coefficients; ; Dynamic output of comprehensive information: based on fusion weights Calculate comprehensive information about the target; comprehensive information about pedestrians includes: relative distance ; The initial relative distance between the target pedestrian and the vehicle; speed ; The initial velocity of the target pedestrian; Category confidence .

[0008] Furthermore, step 3 includes: Quantitative definition of decision-making objectives: Safety indicators include safe distance and collision time TTC, where: In the formula, The distance is relative. The minimum dynamic safe distance between the vehicle and the target pedestrian; when Determine safety in time, and The larger the size, the higher the security. In the formula, It is a relative distance; It is the relative velocity; efficiency indicators : In the formula, The vehicle's current speed. Speed ​​limits on roads; The closer to 1, the higher the efficiency; Comfort Index : In the formula, For the rate of change of vehicle acceleration, The comfort threshold; The closer it is to 1, the higher the comfort level. Construct a multi-objective optimization function, the formula of which is: ; In the formula, For different target weights, ; The output of the multi-objective optimization function corresponds to different decision schemes, including: acceleration, deceleration, lane change, and following. When any one of the indicators S, TTC, E, and C is below the minimum threshold, the weight of that indicator is increased to ensure safety.

[0009] Furthermore, the dynamic minimum safe distance between the vehicle and the target pedestrian. and relative speed The results, obtained using vector calculation methods, are expressed as follows: ; ; In the formula, Braking distance, , The acceleration of the vehicle during braking. The vehicle's current speed; The distance traveled by the target pedestrian or vehicle during braking. , For vehicle braking time, .

[0010] Furthermore, step 4 includes: Based on relative distance ,speed Output results of multi-objective optimization functions Target speed Define the safe distance between the target pedestrian and the vehicle. At that time, the target vehicle speed Its deceleration process satisfies: ; This allows for adjustment of vehicle acceleration through vehicle braking. With the planned target speed ; Initial path generation: Based on a high-precision map and the destination, the A* algorithm is used to generate the initial path; Constructing the path cost function ,choose The path with the minimum cost is the initial path; path cost function The formula is: ; in, For road curvature, Lane width, , The rate of change of vehicle acceleration; , , , These are the weighting coefficients; Emergency Scenario Detection and Path Assessment: Emergency scenarios are detected using the dynamic perception fusion model from step 2, and the feasibility of the initial path is assessed: Path traversability is defined. ,when When the initial path is deemed infeasible, path replanning is triggered, and a rapid dynamic path generation is performed. Fast dynamic path generation: based on scene urgency coefficient Dynamically adjust the path search range and weights, and reduce path replanning latency. In the formula These are the optimization coefficients for the algorithm. ; Scenario urgency coefficient The value is: distance from the vehicle to the sudden obstacle. .

[0011] A vehicle autonomous driving decision-making and planning system, the system comprising: The perception layer includes: cameras, lidar, and millimeter-wave radar for acquiring perception data; and a dynamic perception fusion module for building a dynamic perception fusion model, calculating the fusion weights of the perception data, and dynamically outputting comprehensive information about the target. The decision-making layer includes a decision generation module, which is used to construct a multi-objective optimization function based on safety, efficiency, and comfort indicators; and to generate decision schemes based on the output of the multi-objective optimization function. The planning layer includes a dynamic path adjustment module, which is used to execute a dynamic path planning adjustment mechanism based on the decision-making scheme and high-precision map to generate an initial path and a dynamically adjusted path under emergency scenarios. The control layer, including the control module, is used to receive path information output by the planning layer and control the vehicle's steering, throttle, and braking systems to enable the vehicle to travel according to the planned path and decision scheme; at the same time, it feeds back the real-time status of the vehicle to the perception layer.

[0012] The present invention has the following beneficial effects: (1) Improved real-time perception: The multi-sensor fusion model effectively reduces the latency of dynamic target recognition, provides sufficient response time for decision planning, and reduces the safety risks caused by perception lag; (2) Multi-objective decision optimization: The multi-objective decision optimization function realizes the overall balance of safety, efficiency and comfort, and improves the riding experience of autonomous vehicles; (3) Enhanced path adaptability: The dynamic path planning adjustment mechanism effectively reduces the planning delay in emergency scenarios, and the new path meets the physical constraints of vehicle driving, avoiding driving problems caused by unreasonable paths; (4) Significantly improved safety: The collaborative optimization of the entire process improves driving safety, greatly reduces the risk of collision, and meets the safety requirements of autonomous driving functions. Attached Figure Description

[0013] Figure 1 This is a schematic flowchart of the method of the present invention; Figure 2 This is a schematic diagram of the system structure of the present invention; Figure 3 This is an example diagram of the present invention. Detailed Implementation

[0014] The following will be based on embodiments of the present invention. Figures 1-3 The technical solutions in the embodiments of the present invention will be clearly and completely described. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Unless otherwise specified, the technical means used in the embodiments are conventional means well known to those skilled in the art.

[0015] like Figure 1 A method for decision-making and planning for autonomous driving of vehicles includes the following steps: Step 1, Sensing Data Acquisition and Preprocessing: The acquired sensing data includes: target category, 3D coordinates, distance information, target velocity, and target motion direction; preprocessing involves denoising the acquired sensing data (using the Kalman filter algorithm) and time synchronization (based on the lidar data timestamp, with a deviation ≤10ms).

[0016] Step 2: Construct a dynamic perception fusion model to calculate the fusion weights of the perception data and dynamically output the comprehensive information of the target. Step 3: Construct a multi-objective optimization function based on safety, efficiency, and comfort indicators; generate a decision scheme based on the output of the multi-objective optimization function; Step 4: Based on the decision-making plan and high-precision map, execute the dynamic path planning adjustment mechanism to generate the initial path and the dynamically adjusted path under emergency scenarios; Step 5: The vehicle travels according to the generated path, controlling the steering, throttle, and braking systems. When an emergency occurs, the vehicle slows down and replans the path, enabling the vehicle to travel according to the planned path and decision-making scheme.

[0017] Furthermore, step 2 addresses the disconnect between dynamic environment perception and decision-making by designing a dynamic perception fusion model based on multi-sensor data fusion; this model includes: The formula for constructing a dynamic perception fusion model is as follows:

[0018] in, To integrate the weighting coefficients, For sensor type, For cameras, For lidar, For millimeter-wave radar; To account for the overall sensor accuracy, its value ranges from 0 to 1. For cameras, For lidar, For millimeter-wave radar; For environmental adaptability, its value ranges from 0 to 1, where: Sunny day: ;rain: ; These are the weighting coefficients; After 1000 tests in different environments, the results were verified. The time fusion effect is optimal.

[0019] The dynamic perception fusion model of this invention effectively solves the performance degradation problem of traditional models under complex weather conditions by introducing an environmental adaptability factor. It adopts a dynamic weight allocation mechanism that can automatically adjust the sensor data fusion ratio according to the needs of the scenario.

[0020] Dynamic output of comprehensive information: based on fusion weights Calculate comprehensive information about the target; comprehensive information about pedestrians includes: relative distance ; The initial relative distance between the target pedestrian and the vehicle; speed ; The initial velocity of the target; Category confidence .

[0021] Furthermore, step 3 addresses the shortcomings of multi-objective decision optimization by constructing a multi-objective decision optimization model that balances safety, efficiency, and comfort, which includes: Quantitative definition of decision-making objectives: Safety indicators include safe distance and collision time TTC, where: In the formula, The distance is relative. The minimum dynamic safe distance between the vehicle and the target pedestrian; when Determine safety in time, and The larger the size, the higher the security. In the formula, It is a relative distance; The relative velocity is TTC; TTC < 2s indicates a high-risk scenario.

[0022] efficiency indicators : In the formula, The vehicle's current speed. Speed ​​limits on roads; The closer to 1, the higher the efficiency; Comfort Index : In the formula, For the rate of change of vehicle acceleration, The comfort threshold; The closer it is to 1, the higher the comfort level. Construct a multi-objective optimization function, the formula of which is: ; In the formula, For different target weights, ; The values ​​correspond to different decision-making options, including: acceleration, deceleration, lane changing, and following; when , , When one of the indicators falls below the minimum threshold, the weight of that indicator is increased to ensure security.

[0023] Specifically, the dynamic minimum safe distance between the vehicle and the target pedestrian. and relative speed The results, obtained using vector calculation methods, are expressed as follows: ; ; In the formula, Braking distance, , The acceleration of the vehicle during braking. The vehicle's current speed; The distance traveled by the target pedestrian or vehicle during braking. , For vehicle braking time, .

[0024] For example: Given: the vehicle's current speed pedestrian speed (Towards the vehicle), relative speed Vehicle emergency braking acceleration Braking distance ; Pedestrian movement distance during vehicle braking ,Right now ; Dynamic minimum safe distance , must meet Otherwise, the risk of a collision is extremely high.

[0025] The multi-objective optimization function of this invention achieves balanced optimization of multi-dimensional objectives. By quantifying and integrating safety (safe distance, collision time), efficiency (matching degree between vehicle speed and road speed limit), and comfort (acceleration change rate) indicators, it enables autonomous driving to ensure safety while taking into account traffic efficiency and driving experience. Secondly, it has a dynamic weight adjustment mechanism. When any indicator of safety, efficiency, or comfort falls below the minimum threshold, the weight of that indicator is automatically increased to ensure that safety is always given priority and effectively avoid the safety risks brought about by single-objective optimization. Thirdly, the output decision schemes (acceleration, deceleration, lane change, following, etc.) are accurate and rich, providing intelligent decision-making basis for vehicles in complex traffic scenarios, significantly improving the decision rationality and scenario adaptability of the autonomous driving system.

[0026] Furthermore, step 4 addresses the issue of poor adaptability in path planning for unexpected scenarios by proposing an improved approach. The algorithm's dynamic path planning method includes: Based on relative distance ,speed Output results of multi-objective optimization functions Target speed Define the safe distance between the target pedestrian and the vehicle. At that time, the target vehicle speed Its deceleration process satisfies: ; This allows for adjustment of vehicle acceleration through vehicle braking. With the planned target speed ; Initial route generation: Based on high-precision maps and destinations, an improved method is used. The algorithm generates an initial path and constructs a path cost function. ,choose The path with the minimum cost is the initial path; path cost function The formula is: ; in, For road curvature, Lane width, , , , These are the weighting coefficients; .

[0027] The algorithm is a classic and widely used heuristic search algorithm. It introduces a heuristic function into path planning to guide the search direction, thereby effectively reducing the search space and improving search efficiency. Further improvements are needed. In addition to considering traditional path length factors, the algorithm also incorporates practical driving constraints such as road curvature and lane width, making the generated initial path more consistent with the physical characteristics and safety requirements of vehicle movement. Specifically, the improvement... During the search process, the algorithm not only calculates the actual path length from the starting point to the current node, but also estimates the heuristic distance from the current node to the destination based on information provided by a high-precision map. The weighted sum of these two estimates is then used as part of the path cost. Simultaneously, the algorithm dynamically adjusts the path cost based on road curvature and lane width; the greater the curvature or the narrower the lane, the higher the path cost, thus guiding the search process to avoid areas unfavorable for driving.

[0028] Emergency Scenario Detection and Path Assessment: Emergency scenarios are detected using the dynamic perception fusion model from step 2, and the feasibility of the initial path is assessed: Path traversability is defined. ,when When the initial path is deemed infeasible, path replanning is triggered, and a rapid dynamic path generation is performed. Fast dynamic path generation: based on improvements Algorithm, introducing scene urgency coefficient Dynamically adjust the path search range and weights, and reduce path replanning latency. In the formula These are the optimization coefficients for the algorithm. The new path must meet the minimum turning radius. (For small vehicles) Avoid driving risks caused by unreasonable routes.

[0029] The algorithm is an incremental path planning algorithm based on graph search. It can quickly find the optimal path from the starting point to the destination in a known environment, and can efficiently update path information when the environment changes, avoiding the need to replan the entire path. (Improved version) The algorithm introduces a scenario urgency coefficient, which dynamically adjusts the path search range and weights based on the severity of the emergency scenario.

[0030] Scenario urgency coefficient The value is: distance from the vehicle to the sudden obstacle. .

[0031] This invention first collects and preprocesses perception data, including target category, 3D coordinates, distance, speed, and direction of motion. Next, it constructs a dynamic perception fusion model, calculating the fusion weight coefficients of the camera, LiDAR, and millimeter-wave radar using formulas, and dynamically outputs comprehensive information such as the relative distance, speed, and category confidence of the target (e.g., a pedestrian). Then, based on safety, efficiency, and comfort indicators, it constructs a multi-objective optimization function to generate decision schemes such as acceleration, deceleration, lane changing, and following. Following this, based on the decision schemes and a high-precision map, it uses the A* algorithm to generate an initial path while simultaneously detecting unexpected scenarios and assessing the feasibility of the initial path. If infeasible, it dynamically adjusts the path search range and weights based on the scenario urgency coefficient, quickly generating a dynamically adjusted path. Finally, the vehicle travels along the generated path, controlling the steering, throttle, and braking systems. In the event of an unexpected scenario, it decelerates and replans the path.

[0032] This invention's perception fusion dynamically adjusts the weights of multiple sensors, combining sensor accuracy with environmental adaptability (e.g., sunny or rainy weather), to improve the accuracy of comprehensive target information and provide a reliable basis for decision-making. Decision optimization integrates safety, efficiency, and comfort indicators using a multi-objective optimization function, and can dynamically adjust weights based on indicator thresholds, ensuring safety while also considering driving efficiency and passenger comfort. Path planning is flexible and reliable, using the A* algorithm combined with a path cost function for initial path selection, and rapidly replanning in unexpected scenarios. Planning delay is controlled by a scenario urgency coefficient, improving traffic capacity in complex scenarios. Safety assurance is meticulous, using vector calculations for the dynamic minimum safe distance and relative speed between vehicles and pedestrians, clearly defining the constraint relationship between safe distance and target vehicle speed, effectively reducing collision risk and improving the safety of autonomous driving.

[0033] like Figure 2 , Figure 3 A vehicle autonomous driving decision-making and planning system, employing the aforementioned vehicle autonomous driving decision-making and planning method, the system comprising: The perception layer includes: cameras, LiDAR, and millimeter-wave radar for acquiring perception data; and a dynamic perception fusion module for constructing a dynamic perception fusion model, calculating the fusion weights of the perception data, and dynamically outputting comprehensive target information, including position, velocity, and category; wherein: Camera: Acquires target category (pedestrian, vehicle, obstacle) and color information, with a data refresh rate of 25fps; LiDAR: Acquires target's three-dimensional coordinates and distance information, with a data refresh rate of 10Hz; Millimeter-wave radar: acquires target speed and direction of motion information, with a data refresh rate of 50Hz.

[0034] The decision-making layer includes a decision generation module, which constructs a multi-objective optimization function based on safety, efficiency, and comfort metrics. It then generates decision schemes, such as acceleration, deceleration, and lane changing, based on the output of the multi-objective optimization function. The decision generation module can be an embedded processor with high-performance computing capabilities, possessing a built-in multi-objective optimization function, capable of processing data from the perception layer in real time and rapidly generating decision schemes. Alternatively, it can employ modules such as Mobileye RSS or NVIDIA DRIVEOrin decision processing units.

[0035] The planning layer includes a path dynamic adjustment module, which executes a dynamic path planning adjustment mechanism based on the decision-making scheme and high-precision map to generate an initial path and a dynamically adjusted path for unexpected scenarios. The path dynamic adjustment module can be a dedicated chip with a path planning algorithm. This chip has an improved path planning algorithm built in, which can generate the optimal path in real time based on the decision-making scheme and high-precision map, and quickly adjust the path in unexpected scenarios; or it can use a path planning module based on RRT* and A* algorithms, such as the WaymoBEV path planning module.

[0036] The control layer, including the control module, receives path information output from the planning layer and controls the vehicle's steering, throttle, and braking systems to ensure the vehicle travels according to the planned path and decision-making scheme. Simultaneously, it feeds back the vehicle's real-time status (speed, acceleration, and position) to the perception layer. The control module can be an ECU (Electronic Control Unit), which receives path information from the planning layer and uses built-in control algorithms to control the vehicle's steering, throttle, and braking systems, ensuring the vehicle travels stably according to the planned path and decision-making scheme. The ECU also feeds back real-time vehicle status information, such as speed, acceleration, and position, to the perception layer.

[0037] The control layer, including the control module, is used to receive path information output by the planning layer and control the vehicle's steering, throttle, and braking systems to enable the vehicle to travel according to the planned path and decision scheme; at the same time, it feeds back the vehicle's real-time status (speed, acceleration, and position) to the perception layer.

[0038] Taking "a pedestrian suddenly appears crossing the road while driving on an urban road" as an example, the specific embodiments of the present invention are as follows: Perception layer: Camera identifies pedestrians (category confidence) LiDAR obtains pedestrian location (Vehicle coordinate system), millimeter-wave radar acquires pedestrian speed. (Towards the vehicle); Through fusion calculation, Output comprehensive pedestrian information Output delay ; Decision-making level: Calculating security indicators Efficiency indicators ( Comfort Index (rate of change of acceleration) ); Urban scenes (0.98 after normalization), and simultaneously calculate the "deceleration" and "stopping" schemes. The value was ultimately selected as "Emergency Deceleration (Acceleration)". )"plan; Planning layer: Pedestrian not occupying the current lane is detected; initial path throughput. There's no need to replan the route, but you can adjust the vehicle speed plan to ensure the vehicle is at a safe distance from pedestrians. Decelerate to ; Control layer: Receives "emergency deceleration" decisions and vehicle speed planning information, controls the braking system to apply braking force, so that the vehicle decelerates according to the plan, and at the same time feeds back the vehicle's real-time speed and position to the perception layer, forming a closed-loop control.

[0039] The above embodiments are merely descriptions of preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Any modifications, alterations, alterations, or substitutions made by those skilled in the art to the technical solutions of the present invention without departing from the spirit of the present invention should fall within the protection scope defined by the claims of the present invention.

Claims

1. A decision-making and planning method for autonomous driving of vehicles, characterized in that, Includes the following steps: Step 1, Perception Data Acquisition and Preprocessing: The acquired perception data includes: target category, 3D coordinates, distance information, target velocity, and target motion direction; Step 2: Construct a dynamic perception fusion model to calculate the fusion weights of the perception data and dynamically output the comprehensive information of the target. Step 3: Construct a multi-objective optimization function based on safety, efficiency, and comfort indicators; generate a decision scheme based on the output of the multi-objective optimization function; Step 4: Based on the decision-making plan and high-precision map, execute the dynamic path planning adjustment mechanism to generate the initial path and the dynamically adjusted path under emergency scenarios; Step 5: The vehicle travels according to the generated path, controlling the steering, throttle, and braking systems. When an emergency occurs, the vehicle slows down and replans the path, enabling the vehicle to travel according to the planned path and decision-making scheme.

2. The vehicle autonomous driving decision-making and planning method according to claim 1, characterized in that, Step 2 includes: The formula for constructing a dynamic perception fusion model is as follows: ； in, To integrate the weighting coefficients, For sensor type, For cameras, For lidar, For millimeter-wave radar; To account for the overall sensor accuracy, its value ranges from 0 to 1. For cameras, For lidar, For millimeter-wave radar; For environmental adaptability, its value ranges from 0 to 1, where: sunny day: ;rain: ; These are the weighting coefficients; ; Dynamic output of comprehensive information: based on fusion weights Calculate comprehensive information about the target; comprehensive information about pedestrians includes: relative distance ; The initial relative distance between the target pedestrian and the vehicle; speed ; The initial velocity of the target pedestrian; Category confidence .

3. The vehicle autonomous driving decision-making and planning method according to claim 2, characterized in that, Step 3 includes: Quantitative definition of decision-making objectives: Safety indicators include safe distance and collision time TTC, where: In the formula, The distance is relative. The dynamic minimum safe distance between the vehicle and the target pedestrian; when Determine safety in time, and The larger the size, the higher the security. In the formula, It is a relative distance; It is the relative velocity; efficiency indicators : In the formula, The vehicle's current speed. Speed ​​limits on roads; The closer to 1, the higher the efficiency; Comfort Index : In the formula, For the rate of change of vehicle acceleration, The comfort threshold; The closer to 1, the higher the comfort level. Construct a multi-objective optimization function, the formula of which is: ； In the formula, For different target weights, ; The output of the multi-objective optimization function corresponds to different decision schemes, including: acceleration, deceleration, lane change, and following. When any one of the indicators S, TTC, E, and C is below the minimum threshold, the weight of that indicator is increased to ensure safety.

4. The vehicle autonomous driving decision-making and planning method according to claim 3, characterized in that, Dynamic minimum safe distance between vehicles and target pedestrians and relative speed The results, obtained using vector calculation methods, are expressed as follows: ； ； In the formula, Braking distance, , The acceleration of the vehicle during braking. The vehicle's current speed; The distance traveled by the target pedestrian or vehicle during braking. , For vehicle braking time, .

5. The vehicle autonomous driving decision-making and planning method according to claim 4, characterized in that, Step 4 includes: Based on relative distance ,speed Output results of multi-objective optimization functions Target speed Define the safe distance between the target pedestrian and the vehicle. At that time, the target vehicle speed Its deceleration process satisfies: ； This allows for the adjustment of vehicle acceleration through vehicle braking. With the planned target speed ; Initial path generation: Based on a high-precision map and the destination, the A* algorithm is used to generate the initial path; Constructing the path cost function ,choose The path with the minimum cost is the initial path; path cost function The formula is: ； in, For road curvature, Lane width, , The rate of change of vehicle acceleration; , , , These are the weighting coefficients; Emergency Scenario Detection and Path Assessment: Emergency scenarios are detected using the dynamic perception fusion model from step 2, and the feasibility of the initial path is assessed: Path traversability is defined. ,when When the initial path is deemed infeasible, path replanning is triggered, and a rapid dynamic path generation is performed. Fast dynamic path generation: based on scene urgency coefficient Dynamically adjust the path search range and weights, and reduce path replanning latency. In the formula These are the optimization coefficients for the algorithm. ; Scenario urgency coefficient The value is: distance from the vehicle to the sudden obstacle. .

6. A vehicle autonomous driving decision-making and planning system, characterized in that, The system employs a vehicle autonomous driving decision-making and planning method according to any one of claims 1-5, comprising: The perception layer includes: cameras, lidar, and millimeter-wave radar for acquiring perception data; and a dynamic perception fusion module for building a dynamic perception fusion model, calculating the fusion weights of the perception data, and dynamically outputting comprehensive information about the target. The decision-making layer includes a decision generation module, which is used to construct a multi-objective optimization function based on safety, efficiency, and comfort indicators; and to generate decision schemes based on the output of the multi-objective optimization function. The planning layer includes a dynamic path adjustment module, which is used to execute a dynamic path planning adjustment mechanism based on the decision-making scheme and high-precision map to generate an initial path and a dynamically adjusted path under emergency scenarios. The control layer, including the control module, is used to receive path information output by the planning layer and control the vehicle's steering, throttle, and braking systems to enable the vehicle to travel according to the planned path and decision scheme; at the same time, it feeds back the real-time status of the vehicle to the perception layer.

Images