Vehicle-road collaborative obstacle avoidance path optimization method, system and equipment

Abstract

The invention discloses a vehicle-road collaborative obstacle avoidance path optimization method, system and device, and relates to the technical field of obstacle avoidance path optimization, and the method comprises the steps: obtaining vehicle-mounted sensing data and road test sensing data; obstacle avoidance planning is carried out, and a safe path and a steering decision instruction are determined; meanwhile, the traffic weights of the road nodes under different time granularities are dynamically distributed; introducing a risk coefficient to dynamically correct the path cost; performing global search, and configuring an extreme obstacle avoidance window meeting distance following perception; and executing driving safety response optimization of the safety path and the steering decision instruction. The technical problem that in the prior art, vehicle-road collaborative obstacle avoidance path planning does not fully fuse vehicle-mounted sensing data and road test sensing data, so that the obstacle avoidance safety and the passing efficiency are insufficient is solved, and the effects of fusing the vehicle-mounted sensing data and the road test sensing data to perform obstacle avoidance planning are achieved; the technical effect of improving the safety and passing efficiency of the obstacle avoidance path under vehicle-road cooperation is achieved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of obstacle avoidance path optimization, in particular to a vehicle-road cooperative obstacle avoidance path optimization method, system and device. BACKGROUND

[0002] In the intelligent transportation system, the vehicle-road cooperative obstacle avoidance path planning technology faces many challenges. In the prior art, the traditional obstacle avoidance path planning method often only relies on vehicle-mounted sensor data, which is difficult to comprehensively obtain road global environment information, and has insufficient adaptability to road topological changes, obstacle dynamic trajectories and multi-vehicle interaction scenes, and lacks a dynamic evaluation mechanism for risk factors such as vehicle control limits and weather influences, resulting in difficulty in balancing the safety and traffic efficiency of the planned path, especially in complex traffic environment, which is prone to problems such as untimely obstacle avoidance or traffic flow disorder.

[0003] The prior art has the technical problem that the vehicle-road cooperative obstacle avoidance path planning does not fully integrate vehicle-mounted and road test sensor data, resulting in insufficient obstacle avoidance safety and traffic efficiency. SUMMARY

[0004] The present application provides a vehicle-road cooperative obstacle avoidance path optimization method, system and device, which is used to solve the technical problem that the vehicle-road cooperative obstacle avoidance path planning in the prior art does not fully integrate vehicle-mounted and road test sensor data, resulting in insufficient obstacle avoidance safety and traffic efficiency.

[0005] In view of the above problems, the present application provides a vehicle-road cooperative obstacle avoidance path optimization method, system and device.

[0006] In a first aspect of the present application, a vehicle-road cooperative obstacle avoidance path optimization method is provided, which comprises: Obtaining vehicle-mounted sensor data and road test sensor data; performing obstacle avoidance planning with the vehicle-mounted sensor data and the road test sensor data to determine a safe path and a steering decision instruction; at the same time, dynamically allocating the traffic weight of road nodes at different time granularities through the obstacle dynamic trajectory and road topological data in the road test sensor data; introducing a risk coefficient to dynamically correct the path cost through the vehicle control state and target perception data in the vehicle-mounted sensor data; performing global search with the traffic weight of road nodes at different time granularities and the corrected path cost, configuring a limit obstacle avoidance window under distance following perception at an edge computing node with vehicle traffic efficiency and safety distance as a reward function; and using the edge computing node and the limit obstacle avoidance window to execute driving safety response optimization of the safe path and the steering decision instruction.

[0007] In a second aspect of the present application, a vehicle-road cooperative obstacle avoidance path optimization system is provided, which comprises: The sensor data acquisition module is used for acquiring vehicle-mounted sensor data and road test sensor data; the obstacle avoidance planning module is used for obstacle avoidance planning based on the vehicle-mounted sensor data and the road test sensor data, and is used for determining a safe path and a steering decision instruction; the traffic weight allocation module is used for dynamically allocating traffic weights of road nodes at different time granularities based on dynamic trajectories of obstacles and road topological data in the road test sensor data; the path cost correction module is used for introducing a risk coefficient to dynamically correct a path cost based on vehicle control states and target perception data in the vehicle-mounted sensor data; the limit obstacle avoidance window configuration module is used for global search based on the traffic weights of the road nodes at different time granularities and the corrected path cost, and is used for configuring a limit obstacle avoidance window meeting distance following perception based on vehicle traffic efficiency and safety distance as a reward function at an edge computing node; and the safe response optimization execution module is used for executing a driving safety response optimization of the safe path and the steering decision instruction based on the edge computing node and the limit obstacle avoidance window.

[0008] In a third aspect, the present application provides an electronic device, comprising: a processor; a memory for storing instructions executable by the processor; wherein the processor is configured to execute the vehicle-road cooperative obstacle avoidance path optimization method provided by the present application.

[0009] The one or more technical solutions provided in the present application have at least the following technical effects or advantages: The vehicle-mounted sensor data and the road test sensor data are acquired; obstacle avoidance planning is performed to determine a safe path and a steering decision instruction; traffic weights of road nodes at different time granularities are dynamically allocated based on dynamic trajectories of obstacles and road topological data in the road test sensor data; a risk coefficient is introduced to dynamically correct a path cost; global search is performed based on the traffic weights of the road nodes at different time granularities and the corrected path cost, and a limit obstacle avoidance window meeting distance following perception is configured based on vehicle traffic efficiency and safety distance as a reward function at an edge computing node; and a driving safety response optimization of the safe path and the steering decision instruction is executed based on the edge computing node and the limit obstacle avoidance window. The technical effect of realizing obstacle avoidance planning by fusing vehicle-mounted and road test sensor data is achieved, and the safety and traffic efficiency of an obstacle avoidance path under vehicle-road cooperation are improved. BRIEF DESCRIPTION OF DRAWINGS

[0010] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0011] Figure 1A vehicle-road cooperation obstacle avoidance path optimization method flow diagram is provided for the embodiments of the present application.

[0012] Figure 2 A vehicle-road cooperation obstacle avoidance path optimization system structure diagram is provided for the embodiments of the present application.

[0013] Figure 3 A structure diagram of an electronic device is provided for the present application.

[0014] Marked for explanation: Sensory data acquisition module 10, obstacle avoidance planning module 20, passage weight distribution module 30, path cost correction module 40, limit obstacle avoidance window configuration module 50, safety response optimization execution module 60, processor 21, memory 22, input device 23, output device 24. DETAILED DESCRIPTION

[0015] The present application provides a vehicle-road cooperation obstacle avoidance path optimization method, system and device, which is used to solve the technical problem that the vehicle-road cooperation obstacle avoidance path planning in the prior art does not fully integrate vehicle and road test sensory data, resulting in insufficient obstacle avoidance safety and passage efficiency.

[0016] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative labor are within the scope of protection of the present application.

[0017] Embodiment one, as shown, the present application provides a vehicle-road cooperation obstacle avoidance path optimization method, the method comprises: Figure 1 Step S100: acquiring vehicle-mounted sensory data and road test sensory data.

[0018] Specifically, the vehicle-mounted sensory data of the vehicle itself state and the surrounding environment is acquired in real time by the vehicle-mounted sensor (such as laser radar, camera, millimeter wave radar, etc.), including vehicle speed, acceleration, steering wheel angle, position and speed of target object, etc.; at the same time, the road test sensory data is collected by the sensor deployed on the roadside (such as traffic camera, microwave radar, optical fiber sensor, etc.), covering the dynamic trajectory of the obstacle on the road, the road topological structure (such as lane line, intersection shape, traffic sign position), traffic flow and other information, providing multi-dimensional data support for subsequent obstacle avoidance planning.

[0019] Step S200: obstacle avoidance planning is performed based on the vehicle-mounted sensory data and the road test sensory data, and safety path and steering decision instruction are determined.

[0020] ​Specifically, based on the obtained vehicle-mounted sensor data (including vehicle steering angle, braking acceleration, and other control state data, and target perception data) and road test sensor data (including obstacle dynamic trajectory, road topology data, and the like), the vehicle control limit is first evaluated through the vehicle-mounted sensor data to quantize the influence of weather on path safety by fuzzy logic and introduce a risk coefficient to dynamically correct the path cost. Meanwhile, according to the obstacle speed vector and collision probability in the road test sensor data, combined with the road topology data, a time-varying weight coefficient is set to dynamically allocate the road node passing weight. Then, the passing weight at different time granularities is taken as a heuristic function adjustment factor, the simulated annealing mechanism is introduced, and the search step is dynamically adjusted according to the road congestion degree for global search. The limit obstacle avoidance window is configured with vehicle passing efficiency and safety distance as the reward function at the edge computing node, and finally the obstacle avoidance planning is completed within the window range to determine the safe path and steering decision instruction that meet the safety distance constraint and adapt to the vehicle dynamics characteristics.

[0021] Step S300: Meanwhile, the passing weight of the road node at different time granularities is dynamically allocated through the obstacle dynamic trajectory in the road test sensor data and the road topology data.

[0022] Specifically, at the same time, through the obstacle dynamic trajectory (such as the moving trajectory of pedestrians, vehicles, and other obstacles) and the road topology data (such as lane layout, intersection connection relationship, and the like) in the road test sensor data, the passing situation of each node (such as discrete points on the lane, key points at the intersection) of the road is analyzed for different time granularities (such as short time, medium time, and long time). According to the time when the obstacles arrive at each road node, the road congestion degree, and other factors, the passing weight is dynamically allocated. For example, if a road node will be occupied by an obstacle in a short time, its passing weight is reduced, and if the road node is smooth and conducive to fast vehicle passing, its passing weight is increased, thereby providing a dynamic quantitative reference basis for subsequent path planning.

[0023] Step S400: The risk coefficient is introduced to dynamically correct the path cost through the vehicle control state and target perception data in the vehicle-mounted sensor data.

[0024] Specifically, according to the vehicle basic information (such as vehicle type, load) in the vehicle-mounted sensor data, combined with the real-time collected steering angle, braking acceleration, and other data to evaluate the vehicle control limit, the fuzzy logic algorithm is used to quantize the influence degree of weather factors (such as rainy day, foggy day) on path safety, the vehicle control limit and weather influence parameters are input into the risk coefficient model to dynamically adjust the value range of the risk coefficient, and then the weights of the safety item, energy consumption item, and the like in the path cost function are corrected, so that the path planning result is more consistent with the actual vehicle control ability and environmental risk level.

[0025] Step S500: Global search is performed on the traffic weight of the road node at different time granularities and the modified path cost, and an extreme obstacle avoidance window is configured in the edge computing node based on the vehicle traffic efficiency and safety distance as the reward function under the distance following perception.

[0026] Specifically, the traffic weight of the road node at different time granularities is taken as an adjustment factor of the heuristic function in the A* algorithm, a simulated annealing mechanism is introduced to accept a suboptimal solution with a certain probability to avoid the search from falling into a local optimum, and the search step is dynamically adjusted according to the road congestion degree (the step is reduced to improve the precision in congestion, and the step is increased to improve the efficiency in smoothness), and the modified path cost with the risk coefficient is globally searched; in the edge computing node, the road length passed by the vehicle per unit time is taken as the traffic efficiency, and the minimum safety distance between two vehicles is taken as the safety distance constraint to build the reward function, a vehicle following condition containing a safety distance threshold (such as a dynamic safety distance based on the current vehicle speed) is set by using a PID controller, the vehicle acceleration constraint boundary is calculated in real time by fusing the wheel speed sensor and millimeter wave radar data through a Kalman filter, a trajectory replanning instruction based on model predictive control (MPC) is triggered when a front vehicle emergency braking signal is detected, the acceleration constraint boundary is updated by rolling horizon optimization, and finally the extreme obstacle avoidance window meeting the vehicle dynamics characteristics is configured by using a genetic algorithm for obstacle avoidance optimization based on the distance following perception data in the boundary.

[0027] Step S600: The driving safety response optimization of the safety path and the steering decision instruction is performed using the edge computing node and the extreme obstacle avoidance window.

[0028] Specifically, the local path data of the vehicle is preprocessed by the edge computing node and uploaded to the cloud, the global optimal path is generated by aggregating through a graph neural network, the influence of different obstacle avoidance strategies (such as emergency braking and lane changing) on the traffic flow is preplayed based on the path using a traffic flow simulation model, and a distributed model predictive control algorithm is used to coordinate the steering timing of multiple vehicles to ensure that the interval between the steering actions of adjacent vehicles meets the safety time threshold; in the extreme obstacle avoidance window, a smooth steering trajectory with a steering angular velocity change rate meeting a preset threshold limit is formulated based on the steering decision instruction, a trajectory compensation instruction is triggered based on the wheel speed sensor and a friction coefficient-steering correction model when the road friction coefficient suddenly changes, the driving safety response optimization is realized by adjusting the steering radius and speed, and finally the traffic efficiency and safety in the process of multi-vehicle cooperative obstacle avoidance are ensured.

[0029] In one possible implementation manner, step S300 further includes: Step S310: The obstacle speed vector and the collision probability are extracted from the dynamic trajectory of the obstacle in the road test sensor data.

[0030] Step S320: based on the road topology data, set a time-varying weight coefficient based on road traffic efficiency in combination with the obstacle speed vector and collision probability, the weight update delay corresponding to the time-varying weight coefficient is strongly associated with the congestion road segment weight decay rate.

[0031] Specifically, a dynamic trajectory is constructed using the coordinate sequence of the obstacle in the road test sensor data at different timestamps, Kalman filtering algorithm is used to denoise and smooth the trajectory data, the displacement change amount in unit time is calculated through the coordinate difference of adjacent time points, the speed size is obtained in combination with the time interval, and then the speed vector direction is determined according to the displacement direction. At the same time, based on the real-time position, speed vector and motion trajectory of the vehicle and the obstacle, a collision prediction model is established, the time to collision (TTC) value is calculated and compared with a preset threshold, the Gaussian mixture model is used to fit the probability distribution of TTC, and then the collision probability is extracted, thereby providing a quantitative risk parameter for subsequent road traffic weight adjustment.

[0032] Based on the road topology data (such as lane layout, road segment connection relationship, intersection geometry, etc.), in combination with the obstacle speed vector and the collision probability, a time-varying weight coefficient model is constructed to represent the road traffic efficiency. For road segments with high collision probability, the traffic weight is reduced in proportion to guide the vehicle to preferentially select a low-risk path; for road segments with obstacle speed consistent with the vehicle driving direction and fast speed, the weight is moderately increased to improve the traffic efficiency. Among them, the update mechanism of the time-varying weight coefficient is deeply coupled with the road congestion state: the weight update delay time is determined by analyzing the historical congestion data (such as setting a shorter update period for frequently congested road segments), and the weight decay rate of the congestion road segment is set to a higher value (for example, 3%~5% decay per minute), so that the weight can be quickly adjusted according to the real-time traffic condition, and ensure that the path planning algorithm always makes decisions based on the latest road traffic efficiency data.

[0033] In one possible implementation manner, step S400 further includes: Step S410: according to the vehicle basic information, in combination with the steering angle and the brake acceleration, the vehicle handling limit is evaluated.

[0034] Step S420: quantifying the influence of weather on path safety by fuzzy logic, in combination with the vehicle handling limit, configuring the dynamic adjustment range of the risk coefficient.

[0035] Specifically, the vehicle basic information (including vehicle model, wheelbase, track, tire specification, and tire pressure, etc.) is acquired, the steering angle and braking acceleration data are collected in real time by the vehicle sensors, the vehicle basic information and the real-time collected data are input into the vehicle dynamics model (such as a linear two-degree-of-freedom vehicle model), the adhesion coefficient of the tire and the road surface, the vehicle mass center side slip angle, and the yaw angular velocity and other parameters are calculated, the handling limit of the vehicle under the current steering angle and braking acceleration is evaluated, and the maximum safe steering angle of the vehicle, the minimum braking distance, and the limit side slip critical value and other key parameters are determined.

[0036] A fuzzy logic system is constructed, the weather parameters (such as rainfall, visibility, and road surface wetness, etc.) are taken as input variables, the fuzzy sets such as “light rain”, “heavy rain”, and “dense fog” and the corresponding membership functions are defined, the weather conditions are quantified into the influence level on the path safety, and the greater the rainfall, the higher the risk influence value output by the membership function. Then, the mapping relationship between the risk coefficient and the weather influence level and the vehicle handling limit is established, the adjustment range of the risk coefficient is dynamically configured, the upper limit value of the risk coefficient is expanded when the weather influence level is increased and the vehicle handling limit is decreased, the path planning algorithm is more strictly evaluated for the high-risk path, and the driving safety is ensured.

[0037] In one possible implementation manner, the step S500 further includes: Step S510: The passing weight of the road node under different time granularity is taken as the adjustment factor of the heuristic function, and the simulated annealing mechanism is introduced.

[0038] Step S520: Meanwhile, the search step is dynamically adjusted according to the road congestion degree.

[0039] Specifically, the passing weight of the road node under different time granularity (such as 100 milliseconds, 1 second, 5 seconds, etc.) is taken as the adjustment factor of the heuristic function, and the simulated annealing mechanism is introduced. In the path search process, the weight coefficient of the heuristic function is dynamically adjusted according to the passing weight corresponding to the current time granularity, so as to balance the exploration of new paths and the use of existing paths. The simulated annealing mechanism sets the initial temperature and the cooling rate, so that the algorithm can accept suboptimal solutions with a certain probability, and effectively avoid falling into local optimum. For example, in the road node area with high passing weight, the probability of accepting suboptimal solutions is increased, and the possibly existing better path is explored.

[0040] According to the traffic flow, vehicle density, average driving speed and other information in the road test sensor data, the congestion degree of each road section is evaluated and the congestion level is divided (such as smooth, light congestion, heavy congestion). Based on the congestion level, the step size of the path search algorithm is dynamically adjusted: in the smooth road section, the search step size is increased (such as from 10 meters to 20 meters) to improve the search efficiency and quickly traverse the feasible path; in the light congestion road section, the medium step size (such as 10 meters) is maintained to balance the search accuracy and efficiency; in the heavy congestion road section, the search step size is reduced (such as 5 meters) to finely explore the feasible path in the complex traffic environment and avoid skipping potential safe paths due to large step size. By continuously and adaptively adjusting the step size according to the real-time monitoring of the congestion degree change, it is ensured that the path search algorithm can efficiently and accurately find a safe path in different traffic scenarios.

[0041] In one possible implementation, step S500 further includes: Step S530: setting a vehicle following condition, wherein the vehicle following condition includes a safety distance threshold.

[0042] Step S540: configuring a vehicle acceleration constraint boundary according to the vehicle following condition, so that the distance following perception data performs obstacle avoidance optimization in the vehicle acceleration constraint boundary.

[0043] Specifically, according to the vehicle dynamics characteristics (such as braking distance, response time) and road safety specifications, a vehicle following condition including a safety distance threshold is set. The safety distance threshold is dynamically determined based on current vehicle speed, road friction coefficient and other parameters (for example, a dynamic safety distance model is used to associate the safety distance with the square of the vehicle speed), and the speed, relative position and motion state of the preceding vehicle are obtained in real time through vehicle-mounted sensor data (such as millimeter wave radar, laser radar), and the Kalman filtering algorithm is used to correct the safety distance threshold, so that the vehicle can maintain a safe following distance under different traffic environments and driving conditions.

[0044] According to the set vehicle following condition including the safety distance threshold, combined with the real-time speed, wheel speed, braking state and other information in the vehicle-mounted sensor data, a vehicle acceleration constraint boundary model is established. When the vehicle-mounted sensor data detects that the preceding vehicle is suddenly braking or abnormally decelerating, a trajectory re-planning instruction is triggered, a model predictive control (MPC) is used for rolling horizon optimization, the acceleration constraint boundary is dynamically updated, the upper limit of acceleration is reduced and the lower limit of braking is increased to ensure the safety distance; at the same time, the distance following perception data (such as relative distance, relative speed, angle deviation) obtained by the millimeter wave radar, laser radar and other devices are taken as inputs, and the genetic algorithm is used for obstacle avoidance optimization in the updated acceleration constraint boundary, so as to select the optimal acceleration and deceleration strategy and the corresponding obstacle avoidance path that meet the safety distance constraint and the vehicle dynamics characteristics, and balance the safety and traffic efficiency in the obstacle avoidance process.

[0045] In a possible implementation, step S540 further includes: Step S541: triggering a trajectory re-planning instruction when detecting that the preceding vehicle is suddenly braking or abnormally decelerating.

[0046] Step S542: performing a receding horizon optimization based on the trajectory re-planning instruction, and updating the vehicle acceleration constraint boundary.

[0047] Specifically, the driving data of the preceding vehicle is continuously collected by the vehicle-mounted sensors (such as a laser radar, a millimeter wave radar, a camera, etc.), and the speed change rate and the brake signal state of the preceding vehicle are calculated in real time. When it is monitored that the brake light of the preceding vehicle is suddenly turned on and the speed of the preceding vehicle is reduced by more than a preset threshold (such as a deceleration of 1.5 m / s²) in a short time, or the speed change rate exceeds the abnormal deceleration judgment standard, it is determined that the preceding vehicle is suddenly braking or abnormally decelerating, and a trajectory re-planning instruction is triggered immediately, the current driving path planning logic is interrupted, and an emergency obstacle avoidance trajectory generation process is started, to provide an initial trigger signal for subsequent receding horizon optimization, so as to avoid collision with the preceding vehicle.

[0048] Based on the trajectory re-planning instruction, a receding horizon optimization algorithm (MPC) is started, and in each control period, the feasible driving trajectory in a future limited time domain (such as 5 seconds) is predicted with the current vehicle state (position, speed, acceleration) and the dynamic trajectory after the sudden braking of the preceding vehicle as initial conditions. By constructing a target function including a safety distance, an obstacle avoidance constraint, and a comfort index, an optimal control sequence (steering angle, acceleration) is solved. At the same time, the vehicle acceleration constraint boundary is updated in real time according to the optimized trajectory: when emergency obstacle avoidance is needed, the maximum deceleration threshold is adjusted from the normal -3 m / s² to -5 m / s², and the acceleration feasible interval is reduced to ensure safe following; if the obstacle avoidance space is sufficient, the acceleration boundary is relaxed to improve the traffic efficiency, and finally the dynamic obstacle avoidance is realized through the closed-loop control of the receding optimization and the boundary update.

[0049] In a possible implementation, step S542 further includes: Step S5421: within the limit obstacle avoidance window, a smooth steering trajectory is formulated in combination with the steering decision instruction, and a steering angular velocity change rate of the smooth steering trajectory satisfies a preset change rate threshold.

[0050] Step S5422: according to the smooth steering trajectory, a trajectory compensation instruction is triggered when the road friction coefficient suddenly changes.

[0051] Specifically, within the safety boundary defined by the limit obstacle avoidance window, based on the determined steering decision instruction, a quintic polynomial interpolation or Bezier curve algorithm is used to formulate a smooth steering trajectory, so that the steering angular velocity rate of the steering trajectory is strictly controlled within the preset rate threshold (such as not more than 5° / s²), to avoid the risk of tire skidding or body roll caused by sudden change of angular velocity when the vehicle is steering. In this process, through the steering wheel angle sensor, yaw rate gyroscope in the vehicle-mounted sensor data, the steering action parameters are collected in real time, and are dynamically compared with the preset trajectory model. If it is detected that the steering angular velocity rate exceeds the threshold, the trajectory curvature parameter is immediately adjusted through the model predictive control (MPC) algorithm, to ensure that the steering trajectory always meets the vehicle dynamics constraints and the safety operation requirements.

[0052] According to the formulated smooth steering trajectory, the road friction coefficient data is collected in real time through the wheel speed sensor, vehicle body acceleration sensor and road surface state monitoring equipment (such as infrared / laser road detector). When it is detected that the friction coefficient suddenly changes (such as the mutation amplitude exceeds the preset threshold of 20%) due to factors such as water, ice, oil stains, etc., the trajectory compensation instruction is triggered. At this time, based on the friction coefficient-steering correction model (combining the vehicle dynamics equation and the tire magic formula), the curvature radius of the steering trajectory and the vehicle driving speed are dynamically adjusted. If the friction coefficient suddenly decreases, the steering radius is increased and the vehicle speed is reduced, and at the same time the electronic stability program (ESP) is used to apply brake torque to the inside drive wheel, to compensate for the understeering or oversteering caused by the deterioration of the road adhesion condition; if the friction coefficient suddenly increases, the steering radius is correspondingly reduced to optimize the obstacle avoidance path, to ensure that the steering trajectory always adapts to the real-time road surface state, to avoid vehicle sideslip or loss of control, and to realize dynamic safety compensation of the trajectory.

[0053] In one possible implementation manner, the step S600 further includes: Step S610: uploading local path data according to the edge computing node and the limit obstacle avoidance window, and aggregating a global optimal path.

[0054] Step S620: based on the global optimal path, simulating the traffic flow influence of different obstacle avoidance strategies, and coordinating the steering timing of multiple vehicles.

[0055] Specifically, the vehicle-mounted sensing data and road testing sensing data are preprocessed by the edge computing node, the data such as the distribution of obstacles around the vehicle, the road topology and the local feasible path are extracted, the local path data is encoded into a feature vector after being combined with the safety boundary determined by the limit obstacle avoidance window, and then uploaded to the cloud or regional control center. The control center models the interaction relationship and road topology structure of multiple vehicles by using a graph neural network (GNN), takes the safe distance and traffic efficiency as the optimization objective, fuses the local path data uploaded by multiple vehicles by using a deep reinforcement learning algorithm (such as PPO), searches and aggregates a globally optimal path that meets the multi-vehicle cooperative obstacle avoidance constraint, and realizes the cooperative optimization of the driving path of the vehicles in the region.

[0056] A digital twin scene including road topology, lane attributes and traffic rules is constructed by using a traffic flow micro-simulation software (such as SUMO), the globally optimal path is imported, and different obstacle avoidance strategies (such as emergency braking, single-lane deceleration and yielding, and lane-changing obstacle avoidance) are parameterized modeled. By setting input parameters such as vehicle density, speed distribution and driver behavior model, the traffic flow propagation effect when multiple vehicles simultaneously perform obstacle avoidance actions is simulated, for example, the speed disturbance amplitude and duration of vehicles within a range of 50 meters behind the vehicle when the lane-changing obstacle avoidance strategy is analyzed. Based on the traffic flow influence matrix (including conflict probability, traffic efficiency reduction, etc.) output by the simulation, a distributed model predictive control (DMPC) algorithm is used to calculate the optimal steering timing for each vehicle, the steering start time of each vehicle is allocated (for example, the vehicle starts steering at t=3.2s, and the vehicle B starts steering at t=4.7s) by optimizing the objective function (balancing the safe distance and time delay), ensuring that the interval between the steering actions of adjacent vehicles is not less than the safety time threshold (1.8 seconds), and synchronizing the timing instructions in real time through the Internet of Vehicles communication, realizing the cooperative scheduling of the steering actions of multiple vehicles and avoiding traffic flow disorder.

[0057] Embodiment two, based on the same inventive concept as the obstacle avoidance path optimization method of the preceding embodiments, as shown in Figure 2 The application provides an obstacle avoidance path optimization system for vehicle-road cooperation, and the system and method embodiments in the application are based on the same inventive concept. The system comprises: A sensing data acquisition module 10 is configured to acquire vehicle-mounted sensing data and road testing sensing data.

[0058] An obstacle avoidance planning module 20 is configured to plan obstacle avoidance based on the vehicle-mounted sensing data and road testing sensing data, and determine a safe path and a steering decision instruction.

[0059] A traffic weight allocation module 30 is configured to simultaneously allocate traffic weights of road nodes at different time granularities based on dynamic trajectories of obstacles and road topology data in the road testing sensing data.

[0060] The path cost correction module 40 is configured to introduce a risk coefficient to dynamically correct the path cost based on the vehicle operating state in the vehicle-mounted sensor data and the target perception data.

[0061] The limit obstacle avoidance window configuration module 50 is configured to perform a global search on the passing weight of the road node at different time granularities and the corrected path cost, and configure a limit obstacle avoidance window in distance following perception based on the vehicle passing efficiency and safety distance as a reward function at the edge computing node.

[0062] The safety response optimization execution module 60 is configured to use the edge computing node and the limit obstacle avoidance window to execute the driving safety response optimization of the safety path and the steering decision instruction.

[0063] Further, the system is also configured to implement the following functions: The obstacle speed vector and the collision probability are extracted based on the dynamic trajectory of the obstacle in the road test sensor data, and a time-varying weight coefficient based on the road passing efficiency is set based on the road topology data and the obstacle speed vector and the collision probability. The weight update delay corresponding to the time-varying weight coefficient is strongly related to the congestion road segment weight decay rate.

[0064] Further, the system is also configured to implement the following functions: The vehicle operating limit is evaluated according to the vehicle basic information, combined with the steering angle and the braking acceleration, and the dynamic adjustment range of the risk coefficient is configured based on the influence of the weather on the path safety quantified by fuzzy logic and the vehicle operating limit.

[0065] Further, the system is also configured to implement the following functions: The passing weight of the road node at different time granularities is used as an adjustment factor of the heuristic function, and a simulated annealing mechanism is introduced. At the same time, the search step is dynamically adjusted according to the road congestion degree.

[0066] Further, the system is also configured to implement the following functions: The vehicle following condition is set, and the vehicle following condition includes a safety distance threshold. The vehicle acceleration constraint boundary is configured according to the vehicle following condition, and the distance following perception data is optimized for obstacle avoidance in the vehicle acceleration constraint boundary.

[0067] Further, the system is also configured to implement the following functions: When the front vehicle is detected to be suddenly braked or abnormally decelerated, a trajectory re-planning instruction is triggered. The rolling horizon optimization is performed based on the trajectory re-planning instruction, and the vehicle acceleration constraint boundary is updated.

[0068] Further, the system is also configured to implement the following functions: In the limit obstacle avoidance window, a smooth steering trajectory is formulated in combination with the steering decision instruction, a steering angular velocity change rate of the smooth steering trajectory satisfying a preset change rate threshold; according to the smooth steering trajectory, a trajectory compensation instruction is triggered when a road surface friction coefficient suddenly changes.

[0069] Further, the system is also used to realize the following functions: According to the edge computing node and the limit obstacle avoidance window, local path data is uploaded, and a global optimal path is aggregated; based on the global optimal path, traffic flow influences of different obstacle avoidance strategies are preplayed, and multi-vehicle steering timing is coordinated.

[0070] Embodiment three, Figure 3 The structure schematic diagram of the electronic device provided by the vehicle-road cooperation obstacle avoidance path optimization method of the application shows a block diagram of an exemplary electronic device suitable for implementing the embodiments of the application. Figure 3 The electronic device shown is merely an example, and should not impose any limitation on the functions and use range of the embodiments of the application. As Figure 3 As shown in the figure, the electronic device includes a processor 21, a memory 22, an input device 23 and an output device 24; the number of processors 21 in the electronic device can be one or more, Figure 3 In the example, the processor 21 in the electronic device, the memory 22, the input device 23 and the output device 24 can be connected through a bus or other means, Figure 3 In the example, the connection is through a bus.

[0071] It should be noted that the above sequence of the embodiments of the application is only for description, and does not represent the advantages and disadvantages of the embodiments. And the above describes specific embodiments of the present description. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multi-task processing and parallel processing are possible or can be advantageous.

[0072] The above is only the preferred embodiment of the application, and does not limit the application, any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the application should be included in the protection scope of the application.

[0073] The present description and drawings are only exemplary of the application, and are considered to cover any and all modifications, changes, combinations or equivalents within the scope of the application. Obviously, those skilled in the art can make various modifications and changes to the application without departing from the scope of the application. Thus, if these modifications and changes of the application belong to the scope of the application and its equivalents, the application intends to include these modifications and changes.

Claims

1. The vehicle-road collaborative obstacle avoidance path optimization method is characterized by: The method comprises: Acquire vehicle-mounted sensor data and road test sensor data; Using the onboard sensor data and road test sensor data to perform obstacle avoidance planning, determine a safe path and steering decision instructions; At the same time, the traffic weights of road nodes at different time granularities are dynamically allocated based on the dynamic trajectory of obstacles and road topology data in the road test sensor data; By using the vehicle control state and target perception data in the on-board sensor data, a risk factor is introduced to dynamically correct the path cost; A global search is performed using the traffic weights of road nodes at different time granularities and the corrected path costs. At the edge computing node, the vehicle traffic efficiency and safe distance are used as reward functions to configure the extreme obstacle avoidance window that meets the distance following perception. The edge computing node and the extreme obstacle avoidance window are used to perform driving safety response optimization of the safe path and steering decision instructions.

2. The vehicle-road collaborative obstacle avoidance path optimization method according to claim 1, characterized in that: Dynamically allocating traffic weights of road nodes at different time granularities based on the dynamic trajectory of obstacles and road topology data in the road test sensor data, the method includes: Extracting obstacle velocity vectors and collision probabilities from the dynamic trajectory of obstacles in the drive test sensor data; Based on the road topology data, combined with the obstacle velocity vector and collision probability, a time-varying weight coefficient based on road traffic efficiency is set, and the weight update delay corresponding to the time-varying weight coefficient is strongly correlated with the weight attenuation rate of the congested road section.

3. The vehicle-road cooperative obstacle avoidance path optimization method according to claim 2, characterized in that: By using the vehicle control state and target perception data in the vehicle-mounted sensor data, a risk coefficient is introduced to dynamically correct the path cost. The method includes: Evaluate the vehicle's handling limits based on basic vehicle information, combined with steering angle and braking acceleration; The impact of weather on route safety is quantified using fuzzy logic, and the dynamic adjustment range of the risk coefficient is configured in combination with the vehicle control limit.

4. The vehicle-road collaborative obstacle avoidance path optimization method according to claim 3, characterized in that: A global search is performed using the traffic weights of road nodes at different time granularities and the corrected path costs. The method includes: The traffic weights of road nodes at different time granularities are used as adjustment factors of the heuristic function, and a simulated annealing mechanism is introduced; At the same time, the search step size is dynamically adjusted according to the degree of road congestion.

5. The vehicle-road cooperative obstacle avoidance path optimization method according to claim 4, characterized in that: At the edge computing node, a limit obstacle avoidance window is configured that complies with distance following perception, using vehicle traffic efficiency and safety distance as reward functions. The method further includes: Setting a vehicle following condition, wherein the vehicle following condition includes a safety distance threshold; According to the vehicle following condition, a vehicle acceleration constraint boundary is configured, and obstacle avoidance is prioritized within the vehicle acceleration constraint boundary using distance following perception data.

6. The vehicle-road cooperative obstacle avoidance path optimization method according to claim 5, characterized in that: Configuring a vehicle acceleration constraint boundary according to the vehicle following condition, the method further includes: When the preceding vehicle is detected to have braked suddenly or decelerated abnormally, the trajectory replanning command is triggered; Based on the trajectory replanning instruction, a rolling horizon optimization is performed, and the vehicle acceleration constraint boundary is updated.

7. The vehicle-road cooperative obstacle avoidance path optimization method according to claim 6, characterized in that: The method comprises: Within the extreme obstacle avoidance window, a smooth steering trajectory is formulated in combination with the steering decision instruction, wherein the steering angular velocity change rate of the smooth steering trajectory meets a preset change rate threshold; According to the smooth steering trajectory, when the road friction coefficient suddenly changes, a trajectory compensation instruction is triggered.

8. The vehicle-road cooperative obstacle avoidance path optimization method according to claim 7, characterized in that: Using the edge computing node and the extreme obstacle avoidance window, the method for optimizing the driving safety response of the safe path and the steering decision instruction includes: Upload local path data based on the edge computing nodes and the extreme obstacle avoidance window to aggregate the global optimal path; Based on the global optimal path, the traffic flow impact of different obstacle avoidance strategies is previewed, and the turning timing of multiple vehicles is coordinated.

9. The vehicle-road collaborative obstacle avoidance path optimization system is characterized by: The system is used to implement the vehicle-road collaborative obstacle avoidance path optimization method according to any one of claims 1 to 8, and the system includes: Sensor data acquisition module, used to obtain vehicle-mounted sensor data and road test sensor data; An obstacle avoidance planning module, configured to perform obstacle avoidance planning based on the vehicle-mounted sensor data and the road test sensor data, and determine a safe path and steering decision instructions; A traffic weight allocation module is used to dynamically allocate traffic weights of road nodes at different time granularities based on the dynamic trajectory of obstacles and road topology data in the road test sensor data; A path cost correction module, configured to dynamically correct the path cost by introducing a risk factor based on the vehicle control state and target perception data in the onboard sensor data; The extreme obstacle avoidance window configuration module is used to perform a global search based on the traffic weights of road nodes at different time granularities and the corrected path costs. At the edge computing node, the module uses vehicle traffic efficiency and safe spacing as reward functions to configure the extreme obstacle avoidance window that meets the requirements of distance following perception. A safety response optimization execution module is used to use the edge computing node and the extreme obstacle avoidance window to execute the driving safety response optimization of the safe path and steering decision instructions.

10. An electronic device, characterized in that: The electronic device comprises: processor; a memory for storing instructions executable by the processor; Wherein, the processor is used to execute the vehicle-road collaborative obstacle avoidance path optimization method described in any one of claims 1 to 8.

Images