Emergency obstacle avoidance control method and system for unmanned vehicle

Abstract

The present invention discloses an emergency obstacle avoidance control method and system for an unmanned vehicle, which relates to the field of autonomous driving technology. The method and system include: accessing an unmanned vehicle control terminal, receiving an obstacle avoidance planning path and obstacle avoidance control parameters output by an emergency obstacle avoidance control module; using a rollover risk identification module to identify rollover risks based on the current vehicle's real-time status and output a rollover risk level label; when the rollover risk level label is greater than a preset rollover risk level, activating a flexible control instruction, driving the flexible control module to perform flexible control, outputting a flexible obstacle avoidance planning path and flexible obstacle avoidance control parameters, and performing flexible obstacle avoidance control on the current vehicle based on the flexible obstacle avoidance planning path and flexible obstacle avoidance control parameters. The present invention solves the technical problem in the prior art that a vehicle may roll over or become unstable during an emergency obstacle avoidance process, thereby achieving the technical effect of ensuring that the vehicle avoids obstacles smoothly and safely during the obstacle avoidance process and effectively preventing rollover.

Description

Technical Field

[0001] The present invention relates to the field of autonomous driving technology, and in particular to an emergency obstacle avoidance control method and system for an unmanned vehicle. Background Art

[0002] With the rapid development of autonomous driving technology, unmanned vehicles are increasingly being used in transportation, logistics, and other fields. Unmanned driving systems utilize sensors, cameras, lidar, and other equipment to perceive the environment and make decisions in real time, significantly improving transportation efficiency and safety. However, with the increasing complexity of road environments and the diversification of driving scenarios, existing control methods face new challenges, especially in emergency obstacle avoidance. Traditional obstacle avoidance control methods often focus on path planning and speed control, but when faced with complex obstacles and sharp turns, the control accuracy and stability often fail to meet requirements, potentially leading to vehicle instability, rollover, or other safety hazards. Summary of the Invention

[0003] The present application provides an emergency obstacle avoidance control method and system for an unmanned vehicle, which is used to solve the technical problem in the prior art that the vehicle may roll over or become unstable during the emergency obstacle avoidance process.

[0004] In view of the above problems, the present application provides an emergency obstacle avoidance control method and system for an unmanned vehicle.

[0005] In a first aspect of the present application, a method for emergency obstacle avoidance control of an unmanned vehicle is provided, the method comprising:

[0006] An unmanned vehicle control terminal is connected to the current vehicle; the unmanned vehicle control terminal receives the obstacle avoidance planning path and obstacle avoidance control parameters output by an emergency obstacle avoidance control module, the emergency obstacle avoidance control module includes an obstacle avoidance path planning module and an obstacle avoidance control analysis module; the obstacle avoidance planning path and the obstacle avoidance control parameters are input into a rollover risk identification module, the rollover risk identification module identifies the rollover risk according to the real-time status of the current vehicle, and outputs a rollover risk level label; when the rollover risk level label is greater than the preset rollover risk level, a flexible control instruction is activated; according to the flexible control instruction, a flexible control module connected to the emergency obstacle avoidance control module is driven to perform flexible control, and a flexible obstacle avoidance planning path and flexible obstacle avoidance control parameters are output, and the unmanned vehicle control terminal performs flexible obstacle avoidance control on the current vehicle according to the flexible obstacle avoidance planning path and the flexible obstacle avoidance control parameters.

[0007] A second aspect of the present application provides an emergency obstacle avoidance control system for an unmanned vehicle, the system comprising:

[0008] A terminal connection unit is used to access the unmanned vehicle control terminal of the current vehicle; a data receiving unit is used for the unmanned vehicle control terminal to receive the obstacle avoidance planning path and obstacle avoidance control parameters output by the emergency obstacle avoidance control module, and the emergency obstacle avoidance control module includes an obstacle avoidance path planning module and an obstacle avoidance control analysis module; a rollover risk identification unit is used to input the obstacle avoidance planning path and the obstacle avoidance control parameters into the rollover risk identification module, and the rollover risk identification module performs rollover risk identification according to the real-time status of the current vehicle and outputs a rollover risk level label; a control instruction activation unit is used to activate a flexible control instruction when the rollover risk level label is greater than a preset rollover risk level; an obstacle avoidance control unit is used to drive the flexible control module connected to the emergency obstacle avoidance control module to perform flexible control according to the flexible control instruction, output a flexible obstacle avoidance planning path and flexible obstacle avoidance control parameters, and the unmanned vehicle control terminal performs flexible obstacle avoidance control on the current vehicle according to the flexible obstacle avoidance planning path and the flexible obstacle avoidance control parameters.

[0009] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0010] This application is connected to an unmanned vehicle control terminal of the current vehicle; the unmanned vehicle control terminal receives the obstacle avoidance planning path and obstacle avoidance control parameters output by an emergency obstacle avoidance control module, and the emergency obstacle avoidance control module includes an obstacle avoidance path planning module and an obstacle avoidance control analysis module; the obstacle avoidance planning path and the obstacle avoidance control parameters are input into a rollover risk identification module, and the rollover risk identification module performs rollover risk identification according to the real-time status of the current vehicle and outputs a rollover risk level label; when the rollover risk level label is greater than the preset rollover risk level, a flexible control instruction is activated; according to the flexible control instruction, the flexible control module connected to the emergency obstacle avoidance control module is driven to perform flexible control, and a flexible obstacle avoidance planning path and flexible obstacle avoidance control parameters are output, and the unmanned vehicle control terminal performs flexible obstacle avoidance control on the current vehicle according to the flexible obstacle avoidance planning path and the flexible obstacle avoidance control parameters. The present invention solves the technical problem in the prior art that vehicles may roll over or become unstable during emergency obstacle avoidance. By accessing the unmanned vehicle control terminal of the current vehicle, receiving the obstacle avoidance planning path and obstacle avoidance control parameters, evaluating the rollover risk of the path and performing flexible control, the technical effect of ensuring that the vehicle can smoothly and safely avoid obstacles and effectively prevent rollover during the obstacle avoidance process is achieved. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0012] Figure 1 A flow chart of an emergency obstacle avoidance control method for an unmanned vehicle provided in an embodiment of the present application;

[0013] Figure 2 Schematic diagram of the structure of the emergency obstacle avoidance control system for an unmanned vehicle provided in an embodiment of the present application.

[0014] Explanation of reference numerals: terminal connection unit 11 , data receiving unit 12 , rollover risk identification unit 13 , control instruction activation unit 14 , obstacle avoidance control unit 15 . DETAILED DESCRIPTION

[0015] This application provides an emergency obstacle avoidance control method and system for unmanned vehicles, aiming to solve the technical problem in the prior art that the vehicle may roll over or become unstable during the emergency obstacle avoidance process. By connecting to the unmanned vehicle control terminal of the current vehicle, receiving the obstacle avoidance planning path and obstacle avoidance control parameters, evaluating the rollover risk of the path and performing flexible control, the technical effect of ensuring that the vehicle can smoothly and safely avoid obstacles and effectively prevent rollover during the obstacle avoidance process is achieved.

[0016] The following will be combined with the accompanying drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only some of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0017] It should be noted that any variations of the terms "include" and "have" are intended to cover non-exclusive inclusions. For example, a process, method, system, product or server that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or modules that are not clearly listed or are inherent to these processes, methods, products or devices.

[0018] Example 1, as Figure 1 As shown, the present application provides an emergency obstacle avoidance control method for an unmanned vehicle, the method comprising:

[0019] Step S100: Access the unmanned vehicle control terminal of the current vehicle.

[0020] In this embodiment of the present application, the unmanned vehicle control terminal is accessed via the vehicle's communication bus or wireless network interface. The unmanned vehicle control terminal is the vehicle's core control system, responsible for coordinating and managing all driving tasks. It integrates a computing platform, control algorithms, and communication interfaces to process sensor data and execute decision commands. Ultimately, it implements unmanned driving by controlling the vehicle's driving, steering, acceleration, and braking systems.

[0021] Step S200: The unmanned vehicle control terminal receives the obstacle avoidance planning path and obstacle avoidance control parameters output by the emergency obstacle avoidance control module, wherein the emergency obstacle avoidance control module includes an obstacle avoidance path planning module and an obstacle avoidance control analysis module.

[0022] In an embodiment of the present application, the unmanned vehicle control terminal receives the obstacle avoidance planning path and obstacle avoidance control parameters output by the emergency obstacle avoidance control module. This process involves the collaborative work of multiple modules to ensure that the vehicle can safely and stably avoid obstacles and continue driving.

[0023] The emergency obstacle avoidance control module includes an obstacle avoidance path planning module and an obstacle avoidance control analysis module. The obstacle avoidance path planning module is responsible for planning a safe obstacle avoidance path based on the vehicle's surrounding environment. This module utilizes environmental data from vehicle sensors (such as lidar, cameras, and radar) and performs path planning using the A* algorithm. Specifically, sensors such as lidar scan the environment around the vehicle and collect information on the locations of obstacles. Based on this data, the path planning module converts the environmental information into a grid map or topology map, where each grid cell or node represents a location, with obstacles and free areas occupying different nodes. The A* algorithm then uses a heuristic search to evaluate the cost of each node and select the optimal path, ensuring that the vehicle reaches the target in the shortest possible time while avoiding obstacles. Ultimately, the obstacle avoidance path planning module outputs a planned obstacle avoidance path, a safe path that avoids obstacles and guides the vehicle from its current location to the target.

[0024] The obstacle avoidance control analysis module generates actual control commands based on the planned path, ensuring the vehicle follows the planned path and maintains stability while avoiding obstacles. This module employs a rule-based control approach, generating control parameters based on real-time sensor data and pre-set obstacle avoidance rules. Specifically, the obstacle avoidance control analysis module first uses sensor data to determine the location of surrounding obstacles, identifying obstacles in front, to the sides, and behind. If the distance to the obstacle in front is less than a set threshold (e.g., 2 meters), it issues a "slow down" or "stop" command. If an obstacle to the left or right is closer, it issues a "turn right" or "turn left" command to avoid the obstacle. In emergency situations, rules are used to output "maximum steering angle" and "minimum speed" commands to avoid the obstacle. The key to this process lies in the setting of rules, which determine obstacle avoidance behavior based on the relative position and distance between the vehicle and the obstacle. Finally, based on this real-time data and judgment, the obstacle avoidance control analysis module generates specific obstacle avoidance control parameters, including steering angle, braking force, and acceleration, which guide the vehicle's obstacle avoidance.

[0025] Through the obstacle avoidance path planning module and obstacle avoidance control analysis module, the obstacle avoidance planning path and obstacle avoidance control parameters are output and input into the unmanned vehicle control terminal.

[0026] Step S300: inputting the obstacle avoidance planning path and the obstacle avoidance control parameters into a rollover risk identification module, and the rollover risk identification module performs rollover risk identification according to the current real-time status of the vehicle and outputs a rollover risk level label.

[0027] In this embodiment of the present application, the rollover risk identification module dynamically assesses rollover risk using a pre-embedded recognition model. The module takes the obstacle avoidance planning path and control parameters as input and calculates the input data using a support vector machine. The model performs a comprehensive analysis based on the vehicle's real-time state (such as speed, lateral acceleration, and steering angle), assesses whether the current maneuver will result in a rollover, and outputs a rollover risk level label. This label quantifies the risk level and is used to guide subsequent flexible obstacle avoidance control to ensure vehicle safety.

[0028] Furthermore, in the method provided in the embodiment of the application, the rollover risk identification module is embedded with a classification recognition model obtained by decision boundary learning through support vector machine training, and the expression is as follows:

[0029] ;

[0030] in, is the rollover risk level of the current input path x, n is the sample data size, is the Lagrange multiplier in the support vector machine, indicating that each training sample The weight of For training samples The real rollover risk level label, It is used to calculate the input sample x and training sample The kernel function of the similarity between , Used to measure the input sample x and training sample The difference between them in feature space, is the bandwidth parameter of the kernel function, which controls the influence range of the similarity metric. A bias term for adjusting the classification boundary position.

[0031] In an embodiment of the present application, the rollover risk identification module is embedded with a classification recognition model obtained by decision boundary learning through support vector machine training.

[0032] In the support vector machine (SVM) model, the purpose of the rollover risk identification module is to evaluate the rollover risk based on the real-time status of the vehicle by inputting the obstacle avoidance planning path and obstacle avoidance control parameters into the SVM model and output a rollover risk level label.

[0033] Specifically, the input data includes the real-time status of the vehicle, obstacle avoidance path planning and control parameters. These data are pre-processed and converted into feature vectors suitable for SVM model processing. Each input sample Contains a set of features, such as vehicle speed, acceleration, steering angle, lateral acceleration, wheel rotation, etc., which are crucial for predicting rollover risk. Each input sample Each is given a label , indicating the rollover risk level of the sample. Labels are categorized into four levels: 1, 2, 3, and 4, representing varying degrees of risk, from low to high. These labels are derived from annotations in historical data or simulated training datasets, with 1 indicating low risk and 4 indicating high risk.

[0034] During the training phase, the support vector machine learns the optimal decision boundary by inputting the training data into an optimization process. The decision boundary is a hyperplane that separates data points of different categories. The goal of the support vector machine is to find the hyperplane with the largest margin so that data points of different categories are separated as much as possible. This process involves calculating the coefficients of each sample point, especially the important support vector. Each training sample The contribution to the decision boundary is given by the coefficient These coefficients are obtained through optimization algorithms during the training process of SVM. Their role is to determine which samples are support vectors. Support vectors are samples closest to the decision boundary and have the greatest impact on the classification decision. Through training, SVM will automatically adjust The value of maximises the influence of the support vector on the decision boundary while ensuring that the boundary interval is maximised.

[0035] In order to process nonlinearly separable data, SVM introduces kernel function. The kernel function maps the input data to a high-dimensional feature space so that a linear segmentation surface can be found in the space to separate data of different categories. In this step, SVM uses Gaussian kernel function to calculate the similarity between each sample and the training sample. is the input sample x and the training sample The Euclidean distance between is a parameter that controls the influence range of the similarity measure. Determines the bandwidth or scale of the kernel function and affects the similarity measurement between sample points. During the training process, It can be optimized by cross-validation method to select a value that best suits the current data set. Generally speaking, the smaller A value of will make the similarity measure more limited, considering only very close samples; while a larger A value of will make the similarity metric broader, taking into account more sample points. , which can improve the generalization ability and classification effect of the model.

[0036] Through the optimization algorithm, the support vector machine solves the optimal coefficient , and determine the parameters of the kernel function , thus obtaining the best decision boundary. The goal of this process is to maximize the interval between samples and minimize the classification error. The optimized coefficients and kernel function parameters This ensures that the model can accurately classify new input samples.

[0037] Finally, the trained support vector machine model can classify new input data. For new obstacle avoidance planning paths and control parameters, the classification recognition model is used to obtain a rollover risk level label.

[0038] Furthermore, in the method provided in the embodiment of the application, the training samples are obtained by simulating a vehicle dynamics model, and further includes:

[0039] Construct a vehicle dynamics model; define path segment feature samples, vehicle center of gravity structure samples and vehicle real-time status samples, the path segment feature samples include path length, turning radius and ground friction coefficient, the vehicle center of gravity structure samples are vehicle geometric data and center of gravity height, and the vehicle real-time status samples include lateral acceleration, vehicle speed and steering angle; the vehicle dynamics model simulates the path segment feature samples, vehicle center of gravity structure samples and vehicle real-time status samples, and outputs a simulated data set and a rollover risk level label corresponding to the simulated data set; outputs a training sample based on the simulated data set and the rollover risk level label corresponding to the simulated data set.

[0040] In the embodiment of the present application, a vehicle dynamics model is first constructed. This model is based on the vehicle's dynamic equations and kinematic equations and is used to simulate the dynamic response of the vehicle under different driving conditions. The real-time data obtained by the vehicle sensors is used as input, including lateral acceleration, vehicle speed, steering angle, etc., combined with the dynamic equilibrium equations (in It represents the moment generated by the vehicle due to lateral acceleration or turning. It indicates the force acting on the wheels when the vehicle is turning or moving sideways. represents the height of the vehicle's center of gravity) and the lateral acceleration formula (where v is the vehicle speed and R is the turning radius) and calculates the vehicle's lateral acceleration and roll moment. This model accurately simulates the vehicle's rollover risk under different operating conditions.

[0041] Next, we define path segment feature samples, vehicle center of gravity structure samples, and real-time vehicle status samples. Path segment feature samples include path length, turning radius, and ground friction coefficient. Path length is typically directly derived from historical driving data and represents the distance traveled by the vehicle. The turning radius is calculated from historical records of the vehicle's steering angle and speed. The ground friction coefficient is estimated using road surface information and historical sensor data, or set based on data from a road condition database (e.g., 0.7-0.8 for dry asphalt and 0.1-0.3 for wet roads).

[0042] The vehicle center of gravity structure samples include vehicle geometry data and center of gravity height. The vehicle geometry data is obtained through vehicle design parameters or vehicle specification information in the historical database; the center of gravity height is obtained through static test data or vehicle data extracted from the historical database.

[0043] Real-time vehicle status samples include lateral acceleration, vehicle speed, and steering angle. Lateral acceleration is extracted from historical data from the IMU (Inertial Measurement Unit), vehicle speed is obtained from historical data recorded by the wheel speed sensor, and steering angle is obtained from historical data recorded by the steering wheel angle sensor.

[0044] Next, samples of path segment characteristics, vehicle center of gravity structure, and real-time vehicle status are input into the constructed vehicle dynamics model, and numerical simulation is used to simulate the vehicle's dynamic response. The dynamics model uses the input data to calculate metrics such as lateral force, roll moment, and lateral acceleration, reflecting the vehicle's dynamic behavior under different path segments and driving conditions. For example, when a vehicle negotiates a sharp turn at high speed, the model calculates the vehicle's roll moment based on the input turning radius, vehicle speed, and lateral acceleration, thereby assessing the likelihood of rollover. These simulations generate a dataset of the vehicle's dynamic response under various driving conditions.

[0045] By analyzing the dynamic response indicators in the simulated dataset and combining them with pre-set rollover risk level standards, corresponding rollover risk level labels are generated. Risk levels are graded based on key parameters such as lateral acceleration. For example, a lateral acceleration of <0.3g is labeled as low risk (Level 1), 0.3g-0.5g is labeled as medium-low risk (Level 2), 0.5g-0.7g is labeled as medium-high risk (Level 3), and >0.7g is labeled as high risk (Level 4). These labels represent the risk of a vehicle rolling over under different conditions.

[0046] Finally, the simulated data set and the rollover risk level labels corresponding to the simulated data set are integrated and output as training samples.

[0047] Step S400: When the rollover risk level label is greater than a preset rollover risk level, the flexible control instruction is activated.

[0048] In the embodiment of the present application, the rollover risk level tag is compared with a preset rollover risk level, and when the rollover risk level tag is greater than the preset rollover risk level, the flexible control instruction is activated. The preset rollover risk level is pre-set, for example, level 2.

[0049] Step S500: According to the flexible control instruction, the flexible control module connected to the emergency obstacle avoidance control module is driven to perform flexible control, and a flexible obstacle avoidance planning path and flexible obstacle avoidance control parameters are output. The unmanned vehicle control terminal performs flexible obstacle avoidance control on the current vehicle according to the flexible obstacle avoidance planning path and the flexible obstacle avoidance control parameters.

[0050] In an embodiment of the present application, when the flexible control instruction is activated, the flexible control module connected to the emergency obstacle avoidance control module is driven to perform control and generate a flexible obstacle avoidance planning path and flexible obstacle avoidance control parameters. By initializing the variational probability distribution and setting the variational lower bound, these parameters are continuously adjusted during the calculation process until the similarity between the initialized variational probability distribution and the preset posterior probability distribution is greater than the preset value. Ultimately, by maximizing the variational lower bound, a flexible probability distribution that meets the optimization objective is output, thereby obtaining an accurate obstacle avoidance planning path and flexible obstacle avoidance control parameters. The obstacle avoidance planning path and flexible obstacle avoidance control parameters are then transmitted to the unmanned vehicle control terminal, which performs flexible obstacle avoidance control on the vehicle based on these paths and parameters to ensure that the vehicle can safely avoid obstacles and avoid rollover.

[0051] Furthermore, in the method provided in the embodiment of the application, the flexible control module connected to the emergency obstacle avoidance control module is driven to perform flexible control according to the flexible control instruction, and the flexible obstacle avoidance planning path and flexible obstacle avoidance control parameters are output, and the method further includes:

[0052] Initialize the variational probability distribution; set a variational lower bound, the variational lower bound including the log-likelihood expectation and the entropy of the variational probability distribution, wherein the log-likelihood expectation is the joint probability distribution corresponding to the obstacle avoidance planning path and the obstacle avoidance control parameters obtained by integrating the variational probability distribution; adjust the parameters of the initialized variational probability distribution by maximizing the variational lower bound until the similarity between the initialized variational probability distribution and the posterior probability distribution is greater than a preset similarity, and output the current variational probability distribution as a flexible probability distribution, wherein the posterior probability distribution is a preset true probability distribution; output the flexible obstacle avoidance planning path and the flexible obstacle avoidance control parameters using the flexible probability distribution.

[0053] In the embodiment of the present application, the variational probability distribution is first initialized using a high-dimensional space variational inference algorithm. At this stage, a simple probability distribution, such as a Gaussian distribution, is selected as the initial variational probability distribution form. The initial parameters of the variational probability distribution (such as the mean and covariance matrix) are determined based on the real-time status data of the vehicle (such as vehicle speed, steering angle, location of environmental obstacles, etc.). The purpose of this step is to provide an initial estimate for subsequent optimization, although this initial distribution may not fully reflect the actual driving environment.

[0054] Next, we set a variational lower bound as the optimization objective function. The variational lower bound consists of two parts: the log-likelihood expectation and the entropy of the variational probability distribution. The log-likelihood expectation is used to measure the joint probability distribution of the current obstacle avoidance path and control parameters under the variational probability distribution. Specifically, by integrating the variational probability distribution, we calculate the probability distribution associated with the obstacle avoidance path and control parameters. For example, if a vehicle is making a sharp turn, environmental data provided by sensors (such as lidar) will help determine the optimal obstacle avoidance path and control parameters. Entropy ensures that the distribution is not too concentrated by measuring the uncertainty of the variational probability distribution, which helps avoid the emergence of local optimal solutions.

[0055] The variational probability distribution is then optimized by maximizing the variational lower bound. To accomplish this optimization, an optimization algorithm such as gradient descent or natural gradient descent is used. By adjusting parameters such as the mean and covariance of the variational distribution, the variational probability distribution is gradually brought closer to the posterior probability distribution. During the optimization process, the variational probability distribution is continuously adjusted, using real-time sensor data (such as vehicle speed, lateral acceleration, and steering angle) to update the likelihood expectation component while optimizing the entropy component to enhance the diversity and robustness of the distribution. Multiple iterations are performed until the similarity between the variational probability distribution and the posterior probability distribution exceeds a preset similarity. For example, if the goal is to ensure that the obstacle avoidance path with a turning radius of 10 meters is highly consistent with the optimal path in reality, then after optimization, the variational probability distribution will approach the actual situation.

[0056] After optimization is complete, the current variational probability distribution is considered the final flexible probability distribution. This flexible probability distribution contains the optimal distribution of path planning and control parameters, effectively reflecting the dynamic behavior of the vehicle in complex environments. The vehicle can then generate safe and efficient obstacle avoidance paths and control parameters based on this flexible probability distribution when performing obstacle avoidance.

[0057] The flexible obstacle avoidance planning path and flexible obstacle avoidance control parameters are output using a flexible probability distribution. The flexible obstacle avoidance planning path and flexible obstacle avoidance control parameters are generated for the vehicle using the flexible probability distribution. For example, suppose the vehicle is facing an obstacle with a high risk of rollover. In this case, the optimal obstacle avoidance path and control strategy are calculated based on the flexible probability distribution. For example, in front of an obstacle, the vehicle may choose to avoid it by turning to the right. Based on factors such as the turning radius, vehicle speed, and ground friction, the vehicle generates an appropriate steering angle (e.g., 20 degrees), braking force (e.g., 50%), and acceleration control (e.g., reducing acceleration by 10%). These flexible control parameters are used to adjust the vehicle's motion in real time to ensure a safe and smooth obstacle avoidance process, preventing rollover or other unstable behavior.

[0058] Ultimately, the vehicle executes these flexible obstacle avoidance control instructions through the unmanned vehicle control terminal, adjusting its motion trajectory in real time to ensure that the vehicle can safely avoid obstacles and maintain stability.

[0059] Furthermore, in the method provided in the embodiment of the application, setting the variational lower bound further includes:

[0060] According to the flexible control module, a high-dimensional mapping is performed on the real-time state of the current vehicle to output a high-dimensional state space; based on the high-dimensional state space, the joint probability distribution corresponding to the obstacle avoidance planning path and the obstacle avoidance control parameters is updated to optimize the set variational lower bound.

[0061] In this embodiment of the present application, the flexible control module first performs high-dimensional mapping of the vehicle's real-time state, mapping the vehicle's state data (such as speed, lateral acceleration, steering angle, etc.) and surrounding environment data (such as obstacle location and ground friction coefficient) into a multidimensional state space. This step is achieved through sensor data (such as lidar, IMU sensors, cameras, etc.), ensuring that the vehicle's current motion state and various dynamic factors of the external environment are captured, generating a high-dimensional state space that encompasses all key factors related to obstacle avoidance decisions.

[0062] Based on the resulting high-dimensional state space, the joint probability distribution of the obstacle avoidance path and control parameters is then updated. Using Bayesian inference, each state in the high-dimensional state space is used to calculate the joint probability distribution of the obstacle avoidance path and control parameters, thereby updating the obstacle avoidance strategy. This process uses real-time sensor data and environmental information to adjust the obstacle avoidance path, ensuring the vehicle can flexibly respond to obstacles in a dynamic environment while optimizing control parameters to maintain vehicle stability.

[0063] Finally, the set variational lower bound is optimized through variational inference. The variational lower bound is an objective function used to measure the difference between the variational distribution and the true posterior distribution. It contains two parts: the log-likelihood expectation and the entropy of the variational probability distribution. By adjusting the parameters of the variational distribution (such as the mean and covariance matrix), the variational lower bound is continuously maximized, thereby optimizing the choice of obstacle avoidance path and control parameters. Finally, the setting of the variational lower bound is completed, providing an optimized theoretical basis for the subsequent path planning and control parameter generation.

[0064] Furthermore, the method provided in the application embodiment also includes:

[0065] The flexible control module also includes a stop interval control module; after outputting the flexible obstacle avoidance control parameter, the flexible obstacle avoidance control parameter is subjected to stop interval processing according to a preset interval of the stop interval control module to obtain the processed flexible obstacle avoidance control parameter.

[0066] In an embodiment of the present application, the flexible control module also includes a stop-and-go interval control module. This module's function is to smoothly adjust the flexible obstacle avoidance control parameters at preset time intervals after generating them. Specifically, after the vehicle's obstacle avoidance control parameters (such as steering angle, braking force, or acceleration) are calculated by the flexible control module, the stop-and-go interval control module processes these control parameters according to a preset interval. For example, if the steering angle needs to be adjusted from 0 degrees to 30 degrees, the stop-and-go interval control module will not immediately cause the steering angle to jump to 30 degrees. Instead, it will gradually change the steering angle at a preset interval (e.g., adjusting 5 degrees every 100 milliseconds). This not only prevents the vehicle from making sudden, drastic movements in a short period of time, but also effectively reduces the impact of dynamic adjustments on vehicle stability, minimizing the risk of rollover or loss of control due to drastic maneuvers.

[0067] Through this point-to-point interval processing, the flexible control module optimizes each flexible obstacle avoidance control parameter (such as braking force and acceleration) in stages, ensuring smoother and safer vehicle avoidance. Ultimately, the point-to-point interval control module outputs the processed flexible obstacle avoidance control parameters. These parameters are smoothly adjusted to ensure vehicle stability during obstacle avoidance in dynamic environments, reduce the risk of sudden changes, and provide safe and reliable control input for subsequent control execution.

[0068] Furthermore, in the method provided in the embodiment of the application, after outputting the obstacle avoidance planning path and obstacle avoidance control parameters, the method further includes:

[0069] A path response length analysis is performed on the obstacle avoidance planning path. If the current path response length is less than a preset path response length, the flexible control module performs full-stage flexible control on the obstacle avoidance planning path and the obstacle avoidance control parameters, and outputs full-stage flexible obstacle avoidance control parameters. If the path response length is greater than or equal to the preset path response length, the obstacle avoidance planning path is segmented to obtain multiple obstacle avoidance planning paths. The flexible control module performs multi-stage flexible control on the obstacle avoidance planning path and the obstacle avoidance control parameters, and outputs stage-by-stage flexible obstacle avoidance control parameters.

[0070] In this embodiment, the obstacle avoidance planning path is first analyzed for its path response length. Specifically, the path response length of the obstacle avoidance planning path is determined. This length refers to the time or distance required for the vehicle to execute the obstacle avoidance path from its current state. Based on the comparison of this path response length with a preset path response length, different flexible control strategies are adopted.

[0071] When the path response length is less than the preset path response length, it indicates that the vehicle is able to complete the obstacle avoidance operation in a shorter time. In this case, the same method as above is used to perform full-stage flexible control of the entire obstacle avoidance path and control parameters. Specifically, the variational probability distribution is first initialized, and the variational lower bound is set as the optimization target to calculate the joint probability distribution of the obstacle avoidance path and control parameters. By maximizing the variational lower bound, the parameters of the variational probability distribution are optimized so that they gradually approach the true posterior probability distribution. This process ensures that all control parameters of the entire obstacle avoidance path can be globally optimized to meet the needs of fast obstacle avoidance on shorter paths. Finally, the full-stage-flexible obstacle avoidance control parameters are output, which can ensure that the vehicle completes the obstacle avoidance operation smoothly and avoids obstacles in a shorter time.

[0072] When the path response length is greater than or equal to the preset path response length, the obstacle avoidance path is long, requiring the vehicle to avoid obstacles in a longer time or at a greater distance. In this case, the obstacle avoidance path is segmented, with the long path divided into multiple smaller segments, each of roughly equal length or time. This process divides the path into multiple segments, allowing each segment to be independently optimized, thereby improving control accuracy and stability. Next, the obstacle avoidance planning path and the obstacle avoidance control parameters undergo multi-stage flexible control. Flexible control is applied to each path segment, with control parameters (such as steering angle, acceleration, and braking force) independently optimized for each segment. Using variational inference, the optimal control parameters are calculated for each path segment, ensuring the vehicle can smoothly avoid obstacles and maintain good driving stability at each stage. Finally, the system outputs stage-by-stage flexible obstacle avoidance control parameters that independently optimize control behavior within each path segment, ensuring stable obstacle avoidance in complex environments.

[0073] By analyzing the path response length, the appropriate control strategy is selected based on the complexity of the path. For short paths, full-stage flexible control is used to optimize the control parameters for the entire path. For longer paths, segmented processing and multi-stage flexible control are used to optimize the control parameters for each path segment. Ultimately, these optimized control parameters enable the vehicle to smoothly and safely avoid obstacles in various obstacle avoidance scenarios, ensuring both efficiency and stability.

[0074] Furthermore, in the method provided in the embodiment of the application, after the obstacle avoidance planning path is segmented to obtain multiple obstacle avoidance planning paths, the method further includes:

[0075] The rollover risk identification module identifies the rollover risk of each obstacle avoidance planning path in the multiple obstacle avoidance planning paths according to the real-time status of the current vehicle, and outputs multiple rollover risk level labels corresponding to the multiple obstacle avoidance planning paths; identifies the obstacle avoidance planning paths whose multiple rollover risk level labels are greater than the preset rollover risk level; and performs separate flexible control on each identified obstacle avoidance planning path according to the flexible control module.

[0076] In an embodiment of the present application, after the obstacle avoidance planning path is segmented to obtain multiple obstacle avoidance planning paths, the rollover risk identification module performs rollover risk identification on each of the multiple obstacle avoidance planning paths based on the current vehicle's real-time status. Specifically, the rollover risk identification module uses an embedded classification recognition model to identify the rollover risk of each of the multiple obstacle avoidance planning paths based on the current vehicle's real-time status, and obtains multiple rollover risk level labels corresponding to the multiple obstacle avoidance planning paths.

[0077] Next, the multiple rollover risk level labels corresponding to the obtained multiple obstacle avoidance planning paths are compared with the preset rollover risk level, and the obstacle avoidance planning paths with a rollover risk level greater than the preset rollover risk level among the multiple rollover risk level labels are identified.

[0078] Finally, the flexible control module uses the same method as above to perform individual flexible control on each marked obstacle avoidance planning path, ensuring that the vehicle can avoid obstacles smoothly and safely during the obstacle avoidance process, avoiding rollover or other unsafe behaviors caused by the complexity or instability of the path.

[0079] In the embodiments of the present application, in summary, the embodiments of the present application have at least the following technical effects:

[0080] This application is connected to an unmanned vehicle control terminal of the current vehicle; the unmanned vehicle control terminal receives the obstacle avoidance planning path and obstacle avoidance control parameters output by an emergency obstacle avoidance control module, and the emergency obstacle avoidance control module includes an obstacle avoidance path planning module and an obstacle avoidance control analysis module; the obstacle avoidance planning path and the obstacle avoidance control parameters are input into a rollover risk identification module, and the rollover risk identification module performs rollover risk identification according to the real-time status of the current vehicle and outputs a rollover risk level label; when the rollover risk level label is greater than the preset rollover risk level, a flexible control instruction is activated; according to the flexible control instruction, the flexible control module connected to the emergency obstacle avoidance control module is driven to perform flexible control, and a flexible obstacle avoidance planning path and flexible obstacle avoidance control parameters are output, and the unmanned vehicle control terminal performs flexible obstacle avoidance control on the current vehicle according to the flexible obstacle avoidance planning path and the flexible obstacle avoidance control parameters. The present invention solves the technical problem in the prior art that vehicles may roll over or become unstable during emergency obstacle avoidance. By accessing the unmanned vehicle control terminal of the current vehicle, receiving the obstacle avoidance planning path and obstacle avoidance control parameters, evaluating the rollover risk of the path and performing flexible control, the technical effect of ensuring that the vehicle can smoothly and safely avoid obstacles and effectively prevent rollover during the obstacle avoidance process is achieved.

[0081] Embodiment 2 is based on the same inventive concept as the emergency obstacle avoidance control method for an unmanned vehicle in the aforementioned embodiment. Figure 2 As shown, the present application provides an emergency obstacle avoidance control system for an unmanned vehicle. The system and method embodiments in the present application are based on the same inventive concept. The system includes:

[0082] The terminal connection unit 11 is used to access the unmanned vehicle control terminal of the current vehicle; the data receiving unit 12 is used for the unmanned vehicle control terminal to receive the obstacle avoidance planning path and obstacle avoidance control parameters output by the emergency obstacle avoidance control module, and the emergency obstacle avoidance control module includes an obstacle avoidance path planning module and an obstacle avoidance control analysis module; the rollover risk identification unit 13 is used to input the obstacle avoidance planning path and the obstacle avoidance control parameters into the rollover risk identification module, and the rollover risk identification module performs rollover risk identification according to the real-time status of the current vehicle and outputs a rollover risk level label; the control instruction activation unit 14 is used to activate the flexible control instruction when the rollover risk level label is greater than the preset rollover risk level; the obstacle avoidance control unit 15 is used to drive the flexible control module connected to the emergency obstacle avoidance control module to perform flexible control according to the flexible control instruction, output the flexible obstacle avoidance planning path and flexible obstacle avoidance control parameters, and the unmanned vehicle control terminal performs flexible obstacle avoidance control on the current vehicle according to the flexible obstacle avoidance planning path and the flexible obstacle avoidance control parameters.

[0083] Furthermore, the system is also used to implement the following functions:

[0084] A path response length analysis is performed on the obstacle avoidance planning path. If the current path response length is less than a preset path response length, the flexible control module performs full-stage flexible control on the obstacle avoidance planning path and the obstacle avoidance control parameters, and outputs full-stage flexible obstacle avoidance control parameters. If the path response length is greater than or equal to the preset path response length, the obstacle avoidance planning path is segmented to obtain multiple obstacle avoidance planning paths. The flexible control module performs multi-stage flexible control on the obstacle avoidance planning path and the obstacle avoidance control parameters, and outputs stage-by-stage flexible obstacle avoidance control parameters.

[0085] Furthermore, the system is also used to implement the following functions:

[0086] The rollover risk identification module identifies the rollover risk of each obstacle avoidance planning path in the multiple obstacle avoidance planning paths according to the real-time status of the current vehicle, and outputs multiple rollover risk level labels corresponding to the multiple obstacle avoidance planning paths; identifies the obstacle avoidance planning paths whose multiple rollover risk level labels are greater than the preset rollover risk level; and performs separate flexible control on each identified obstacle avoidance planning path according to the flexible control module.

[0087] Furthermore, the system is also used to implement the following functions:

[0088] ;

[0089] in, is the rollover risk level of the current input path x, n is the sample data size, is the Lagrange multiplier in the support vector machine, indicating that each training sample The weight of For training samples The real rollover risk level label, It is used to calculate the input sample x and training sample The kernel function of the similarity between , Used to measure the input sample x and training sample The difference between them in feature space, is the bandwidth parameter of the kernel function, which controls the influence range of the similarity metric. A bias term for adjusting the classification boundary position.

[0090] Furthermore, the system is also used to implement the following functions:

[0091] Construct a vehicle dynamics model; define path segment feature samples, vehicle center of gravity structure samples and vehicle real-time status samples, the path segment feature samples include path length, turning radius and ground friction coefficient, the vehicle center of gravity structure samples are vehicle geometric data and center of gravity height, and the vehicle real-time status samples include lateral acceleration, vehicle speed and steering angle; the vehicle dynamics model simulates the path segment feature samples, vehicle center of gravity structure samples and vehicle real-time status samples, and outputs a simulated data set and a rollover risk level label corresponding to the simulated data set; outputs a training sample based on the simulated data set and the rollover risk level label corresponding to the simulated data set.

[0092] Furthermore, the system is also used to implement the following functions:

[0093] Initialize the variational probability distribution; set a variational lower bound, the variational lower bound including the log-likelihood expectation and the entropy of the variational probability distribution, wherein the log-likelihood expectation is the joint probability distribution corresponding to the obstacle avoidance planning path and the obstacle avoidance control parameters obtained by integrating the variational probability distribution; adjust the parameters of the initialized variational probability distribution by maximizing the variational lower bound until the similarity between the initialized variational probability distribution and the posterior probability distribution is greater than a preset similarity, and output the current variational probability distribution as a flexible probability distribution, wherein the posterior probability distribution is a preset true probability distribution; output the flexible obstacle avoidance planning path and the flexible obstacle avoidance control parameters using the flexible probability distribution.

[0094] Furthermore, the system is also used to implement the following functions:

[0095] According to the flexible control module, a high-dimensional mapping is performed on the real-time state of the current vehicle to output a high-dimensional state space; based on the high-dimensional state space, the joint probability distribution corresponding to the obstacle avoidance planning path and the obstacle avoidance control parameters is updated to optimize the set variational lower bound.

[0096] Furthermore, the system is also used to implement the following functions:

[0097] The flexible control module also includes a stop interval control module; after outputting the flexible obstacle avoidance control parameter, the flexible obstacle avoidance control parameter is subjected to stop interval processing according to a preset interval of the stop interval control module to obtain the processed flexible obstacle avoidance control parameter.

[0098] It should be noted that the order in which the embodiments of the present application are presented is for illustrative purposes only and does not necessarily represent the superiority or inferiority of the embodiments. Furthermore, the foregoing descriptions of specific embodiments of this specification are provided. The processes depicted in the accompanying drawings do not necessarily require the specific order or sequential sequence shown to achieve the desired results. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0099] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present application shall be included in the scope of protection of the present application.

[0100] This specification and drawings are merely illustrative of the present application and are intended to cover any and all modifications, variations, combinations, or equivalents within the scope of this application. Obviously, those skilled in the art may make various modifications and variations to this application without departing from the scope of this application. Thus, this application is intended to include such modifications and variations as fall within the scope of this application and its equivalents.

Claims

1. An emergency obstacle avoidance control method for an unmanned vehicle, characterized in that: The method comprises: Access the unmanned vehicle control terminal of the current vehicle; The unmanned vehicle control terminal receives the obstacle avoidance planning path and obstacle avoidance control parameters output by the emergency obstacle avoidance control module, wherein the emergency obstacle avoidance control module includes an obstacle avoidance path planning module and an obstacle avoidance control analysis module; Inputting the obstacle avoidance planning path and the obstacle avoidance control parameters into a rollover risk identification module, the rollover risk identification module performs rollover risk identification according to the current real-time state of the vehicle and outputs a rollover risk level label; When the rollover risk level label is greater than a preset rollover risk level, activating a flexible control instruction; According to the flexible control instruction, the flexible control module connected to the emergency obstacle avoidance control module is driven to perform flexible control, and a flexible obstacle avoidance planning path and flexible obstacle avoidance control parameters are outputted, and the unmanned vehicle control terminal performs flexible obstacle avoidance control on the current vehicle according to the flexible obstacle avoidance planning path and the flexible obstacle avoidance control parameters; The rollover risk identification module is embedded with a classification recognition model obtained by decision boundary learning through support vector machine training, and the expression is as follows: ； in, is the rollover risk level of the current input path x, n is the sample data size, is the Lagrange multiplier in the support vector machine, indicating that each training sample The weight of For training samples The real rollover risk level label, It is used to calculate the input sample x and training sample The kernel function of the similarity between , Used to measure the input sample x and training sample The difference between them in feature space, is the bandwidth parameter of the kernel function, which controls the influence range of the similarity metric. A bias item to adjust the classification boundary position; The method includes: driving a flexible control module connected to the emergency obstacle avoidance control module to perform flexible control according to the flexible control instruction, and outputting a flexible obstacle avoidance planning path and flexible obstacle avoidance control parameters. Initialize the variational probability distribution; Setting a variational lower bound, the variational lower bound comprising a log-likelihood expectation and an entropy of a variational probability distribution, wherein the log-likelihood expectation is a joint probability distribution corresponding to the obstacle avoidance planning path and the obstacle avoidance control parameters obtained by integrating the variational probability distribution; Adjusting the parameters of the initialized variational probability distribution by maximizing the variational lower bound until the similarity between the initialized variational probability distribution and the posterior probability distribution is greater than a preset similarity, and outputting the current variational probability distribution as a flexible probability distribution, wherein the posterior probability distribution is a preset true probability distribution; Outputting a flexible obstacle avoidance planning path and flexible obstacle avoidance control parameters using the flexible probability distribution; The flexible control module also includes a stop interval control module; After the flexible obstacle avoidance control parameter is output, the flexible obstacle avoidance control parameter is subjected to a stop interval processing according to a preset interval of the stop interval control module to obtain the processed flexible obstacle avoidance control parameter.

2. The emergency obstacle avoidance control method for an unmanned vehicle according to claim 1, wherein: After outputting the obstacle avoidance planning path and obstacle avoidance control parameters, the method further includes: performing a path response length analysis on the obstacle avoidance planning path; if the current path response length is less than a preset path response length, the flexible control module performs full-stage flexible control on the obstacle avoidance planning path and the obstacle avoidance control parameters, and outputs full-stage flexible obstacle avoidance control parameters; If the path response length is greater than or equal to the preset path response length, the obstacle avoidance planning path is segmented to obtain multiple obstacle avoidance planning paths. The flexible control module performs multi-stage flexible control on the obstacle avoidance planning path and the obstacle avoidance control parameters, and outputs the stage-by-stage flexible obstacle avoidance control parameters.

3. The emergency obstacle avoidance control method for an unmanned vehicle according to claim 2, wherein: After segmenting the obstacle avoidance planning path to obtain multiple segments of the obstacle avoidance planning path, the method includes: The rollover risk identification module performs rollover risk identification on each of the multiple obstacle avoidance planning paths according to the real-time status of the current vehicle, and outputs multiple rollover risk level labels corresponding to the multiple obstacle avoidance planning paths; Identifying obstacle avoidance planning paths whose rollover risk level labels are greater than the preset rollover risk level; Each marked obstacle avoidance planning path is independently and flexibly controlled according to the flexible control module.

4. The emergency obstacle avoidance control method for an unmanned vehicle according to claim 1, wherein: The training samples are obtained by simulating the vehicle dynamics model. The method includes: Construct vehicle dynamics models; Define path segment feature samples, vehicle center of gravity structure samples, and vehicle real-time status samples. The path segment feature samples include path length, turning radius, and ground friction coefficient. The vehicle center of gravity structure samples include vehicle geometry data and center of gravity height. The vehicle real-time status samples include lateral acceleration, vehicle speed, and steering angle. The vehicle dynamics model simulates the path segment feature samples, the vehicle center of gravity structure samples and the vehicle real-time state samples, and outputs a simulated data set and a rollover risk level label corresponding to the simulated data set; Output training samples based on the simulated data set and the rollover risk level labels corresponding to the simulated data set.

5. The emergency obstacle avoidance control method for an unmanned vehicle according to claim 1, wherein: The method for setting the variational lower bound further includes: Performing high-dimensional mapping on the real-time state of the current vehicle according to the flexible control module and outputting a high-dimensional state space; Based on the high-dimensional state space, the joint probability distribution corresponding to the obstacle avoidance planning path and the obstacle avoidance control parameters is updated to optimize the set variational lower bound.

6. Emergency obstacle avoidance control system for unmanned vehicles, characterized in that: A system for executing the emergency obstacle avoidance control method for an unmanned vehicle according to any one of claims 1 to 5, comprising: A terminal connection unit, used to connect to the unmanned vehicle control terminal of the current vehicle; A data receiving unit, configured for the unmanned vehicle control terminal to receive the obstacle avoidance planning path and obstacle avoidance control parameters output by the emergency obstacle avoidance control module, wherein the emergency obstacle avoidance control module includes an obstacle avoidance path planning module and an obstacle avoidance control analysis module; a rollover risk identification unit, configured to input the obstacle avoidance planning path and the obstacle avoidance control parameters into a rollover risk identification module, wherein the rollover risk identification module performs rollover risk identification based on the current real-time status of the vehicle and outputs a rollover risk level label; a control instruction activating unit, configured to activate a flexible control instruction when the rollover risk level label is greater than a preset rollover risk level; An obstacle avoidance control unit is used to drive a flexible control module connected to the emergency obstacle avoidance control module to perform flexible control according to the flexible control instruction, output a flexible obstacle avoidance planning path and flexible obstacle avoidance control parameters, and the unmanned vehicle control terminal performs flexible obstacle avoidance control on the current vehicle according to the flexible obstacle avoidance planning path and the flexible obstacle avoidance control parameters.

Images