Ultimate safety system for intelligent vehicle under extreme driving conditions

Abstract

An ultimate safety system for an intelligent vehicle under extreme driving conditions, comprising a vehicle sensing system supporting an ultimate safety function, which is used for collecting environmental information under an extreme vehicle driving condition and sending same in real time to an autonomous driving system and advanced driver assistance system; an ultimate safety function decision-making module of the autonomous driving system and advanced driver assistance system, which is used for acquiring the environmental information and receiving in an unknown environment data-driven reinforcement learning policy-based input information and fed-back vehicle state and execution information; and an ultimate safety function execution module, which is used for acquiring decision-making information and outputting the vehicle state and execution information. The ultimate safety system for an intelligent vehicle under extreme driving conditions makes full use of the motion potential of each system of the intelligent vehicle, and improves the emergency response capability and standby level of the intelligent vehicle under extreme driving conditions.

Description

Intelligent vehicle limit safety system under extreme working condition

[0001] Cross-reference to related applications

[0002] The present disclosure is based on and claims priority to Chinese Patent Application No. 202411312012.8, filed on September 19, 2024, the entire contents of which are incorporated herein by reference. TECHNICAL FIELD

[0003] The present disclosure relates to the field of automotive active safety technology, and in particular to an intelligent vehicle limit safety system under extreme working condition. BACKGROUND

[0004] According to the report of the World Health Organization, road traffic injuries are the leading cause of death among young people aged 5 to 29 years, resulting in about 1.3 million deaths each year. Among them, as high as 93% of traffic accidents are caused by driver errors. Autonomous driving technology is expected to reduce or avoid accidents caused by human errors, thereby improving vehicle safety.

[0005] Extreme working condition generally refers to a critical environment with a risk of collision or related conditions with serious deterioration of the vehicle's own condition. Related scenarios are, for example, sudden front collision scenarios, high-speed passing sharp turn scenarios, and vehicle critical instability. In such working conditions, the reaction speed and decision-making intelligence level of the vehicle are particularly critical. The control idea of the active safety system currently equipped on vehicles is mainly to limit the driving state of the vehicle within a linear and stable range to avoid vehicle instability when the wheels reach the dynamic limit. From the perspective of vehicle controllability, the current active safety function is too conservative and cannot fully utilize the potential of vehicle dynamics control under extreme working conditions.

[0006] Currently, the design of safety functions for extreme working conditions is still in its infancy. The active safety functions on the product level, such as the automatic emergency braking function, are designed for relatively single scenarios, and the control strategy is relatively simple. The functional design at the research level is relatively fragmented, and the utilization of the potential of vehicle dynamics control is limited, and there is a lack of a unified framework to integrate different technologies, making the coordination effect of the vehicle execution system in actual application limited.

[0007] At present, the theoretical method of limit safety function for extreme scenarios has not been established, and the trigger standard, decision planning and control method of limit safety function are not systematic. For the trigger standard, there is no unified trigger system considering the collision avoidance demand, cornering motion demand and dynamic stability demand, resulting in inconsistency and limitation of the application of trigger standard in different extreme situations. For decision planning, there is lack of an evaluation system for decision making for a wide range of extreme scenarios. For control method, there is lack of a unified framework to integrate the advantages of various technical routes, so that the current technology has limited collaborative effect in practical application.

[0008] In view of the above problems, it is urgent to establish a limit safety function, architecture and system with unified trigger standard, perfect evaluation system and collaborative control scheme. SUMMARY

[0009] The present disclosure aims to at least partially solve one of the technical problems in the related art.

[0010] To this end, the present disclosure proposes a limit safety system for intelligent vehicles in extreme working conditions, which is used for operations involving but not limited to dynamic critical stability or limit motion capability in extreme working conditions, so as to fully utilize the motion potential of each system of the intelligent vehicle and improve the emergency handling capability and backup level of the intelligent vehicle in extreme working conditions.

[0011] To achieve the above purpose, the present disclosure proposes an intelligent vehicle limit safety system in extreme working conditions, which comprises:

[0012] A vehicle sensing system supporting limit safety function is used to collect environmental information of the vehicle in extreme working conditions and send the environmental information to the autonomous driving system and the advanced auxiliary driving system in real time through Ethernet or CAN communication;

[0013] The autonomous driving system and the advanced auxiliary driving system comprise a limit safety function decision module and a limit safety function execution module, wherein;

[0014] The limit safety function decision module comprises a limit collision avoidance judgment module, a limit cornering judgment module, a limit vehicle condition recognition module, a limit state recognition module and a safety boundary monitoring module; the limit safety function decision module is used to obtain environmental information and receive input information based on data-driven reinforcement learning strategy in unknown environment and vehicle state and execution information fed back by the limit safety function execution module;

[0015] The limit safety function execution module is used to obtain the decision information sent by the limit safety function decision module and output vehicle state and execution information.

[0016] The intelligent vehicle extreme condition limit safety system of the embodiments of the present disclosure can further have the following additional technical features.

[0017] In an embodiment of the present disclosure, the vehicle sensing system supporting the limit safety function includes a camera, a millimeter wave radar, and a combined inertial navigation IMU.

[0018] In an embodiment of the present disclosure, the limit safety function decision module is further configured to:

[0019] The limit safety decision solver solves the target trajectory based on the evaluation results and using an iterative optimization or dynamic optimization method; wherein the evaluation results include evaluation of the vehicle risk avoidance reaction in the extreme scene based on the characteristics of the extreme scene, combined with the scene direct risk, the scene expected risk, the operation risk avoidance level, and the operation limit degree.

[0020] The target trajectory obtained by solving the decision matching the current extreme scene is transmitted to the limit safety function execution module.

[0021] In an embodiment of the present disclosure, the system is further configured to fuse the hierarchical decision control and the data-driven decision control using a fusion control mechanism; wherein

[0022] For the case where the hierarchical decision control and the data-driven decision control have the same target, the fusion control mechanism deploys the decision and generates the trajectory in the data generation environment, replaces the trajectory information output by the hierarchical decision with the trajectory, and uses a model-based trajectory tracker to solve the closed-loop input uses the data-driven method to directly generate the feedforward input The sum or weighted sum of the two generates the final to-be-executed input:

[0023] For the case where the hierarchical decision control and the data-driven decision control have different targets, the fusion control mechanism compares the target gap between the data-driven decision and the hierarchical decision, and the data-driven strategy directly generates the action information a Data , tracks the target trajectory of the hierarchical decision to obtain the action information a model , generates the final to-be-executed input based on the target difference, and the generation method includes neural network fusion and weighted fusion: a t = Diff target (a Data ,a model ).

[0024] In one embodiment of the present disclosure, the execution input output by the fusion control mechanism is combined with four-wheel drive and rear-wheel drive in driving to control drive slip, expand the limit driving control boundary of the vehicle; combined with friction braking and motor braking in braking to provide backup limit braking capability guarantee; combined with active steering and differential steering in steering to provide the maximum attitude adjustment capability for limit safety function; the vehicle executes the target input of each actuator to execute limit safety maneuver.

[0025] In one embodiment of the present disclosure, the environment information is a sharp curve environment, and the target of the limit safety decision module is to guarantee the limit drift over curve of comprehensive safety; the evaluation indexes related to limit over curve performance and safety include immediate evaluation and terminal evaluation: the immediate over curve performance index includes: the evaluation item r p :

[0026] wherein k pl and k pv are negative parameters; l is the lateral deviation of the vehicle relative to the center of the curve in the Frenet coordinate system, and v is the current vehicle speed; l ref (s) is the lateral displacement of the formation suboptimal trajectory at the current curve center at the current curve; is the maximum vehicle speed under the specific stability constraint under the current adhesion.

[0027] In one embodiment of the present disclosure, the immediate evaluation item r β rewarding high side slip angle encourages the vehicle to turn with a higher center of mass slip angle:

[0028] wherein k β is negative, v x is the longitudinal vehicle speed in the vehicle coordinate system, v y is the lateral vehicle speed in the vehicle coordinate system, and β is the center of mass side slip angle of the vehicle;

[0029] The terminal reward item r t rewarding the comprehensive performance of limit safety function encourages safe and fast over curve: r t =(1-χ)k t1 +k t2 χ(t f -t ref )

[0030] wherein χ represents a parameter of the final state of the vehicle: χ = 1 indicates that the vehicle has safely completed the turning task, χ = 0 indicates that an unsafe event occurs, t f represents the time required for the vehicle to reach the terminal state, and t ref represents the total travel time of the pre-optimized trajectory; k is a constant.t1 and k t2 are negative and positive values respectively, for penalizing unsafe end states and rewarding as short limit corner time as possible.

[0031] In an embodiment of the present disclosure, the limit safety decision solver adopts an end-to-end iterative solver based on reinforcement learning, and the output is directly action information; the iterative method is a Critic and Actor deep neural network, the Critic network parameters are trained according to a minimum time difference loss function; the Actor network parameters are trained by maximizing the value function.

[0032] In an embodiment of the present disclosure, the data collection environment used in the limit safety decision solver is a Carsim platform; the reinforcement learning-based control strategy is deployed to the corresponding scene in the data collection environment to obtain the optimal trajectory T p under the current scene, which is the target trajectory information of the limit safety function execution module: T p ={S i |simulate(P v ,V c ,S ini ,π),i=1,2,3,…,n}

[0033] Wherein, S represents the vehicle state along the estimated trajectory, S ini represents the vehicle state when entering the corner.

[0034] In an embodiment of the present disclosure, since the estimated target trajectory is used to generate feedback input, it is necessary to convert T p into Cartesian coordinate trajectory , and use a proportional-integral-derivative controller to track the trajectory to obtain the supplementary input a PID ; the end-to-end action output by reinforcement learning will be an important feedforward reference, which is recorded as a RL ; the hierarchical decision and the data-driven decision have the same goal, and the direct addition is used to obtain the execution input: a t =a RL +a PID .

[0035] The limit safety system of the intelligent vehicle in extreme working conditions in the embodiment of the present disclosure integrates traditional friction braking and modern motor braking technology in the braking system, not only provides accurate braking force, but also increases the redundancy of braking, ensuring high safety in extreme braking situations. In the steering technology, the embodiment combines active steering system and differential steering technology, greatly enhances the controllability and stability of the vehicle in emergency situations, and provides the maximum attitude adjustment capability for the limit safety function.

[0036] The additional aspects and advantages of the present disclosure will be in part apparent and in part pointed out below in the description of the present disclosure. BRIEF DESCRIPTION OF DRAWINGS

[0037] The above and / or additional aspects and advantages of the present disclosure will become apparent and be readily appreciated from the following description, including the references to the figures, in which:

[0038] Fig. 1 is a structure diagram of a limit safety system in an extreme working condition of an intelligent vehicle according to an embodiment of the present disclosure;

[0039] Fig. 2 is a deployment schematic diagram according to an embodiment of the present disclosure;

[0040] Fig. 3 is a deployment comparison trajectory diagram according to an embodiment of the present disclosure. DETAILED DESCRIPTION

[0041] It should be noted that the embodiments and features of the present disclosure can be combined with each other without conflict. The present disclosure will be described in detail below with reference to the accompanying drawings and in conjunction with the embodiments.

[0042] In order to enable persons skilled in the art to better understand the present disclosure scheme, the technical solutions in the embodiments of the present disclosure will be described clearly and completely below with reference to the accompanying drawings in the embodiments of the present disclosure. Obviously, the described embodiments are only a part of the embodiments of the present disclosure, not all the embodiments. Based on the embodiments in the present disclosure, all other embodiments obtained by persons skilled in the art without creative labor should belong to the scope of protection of the present disclosure.

[0043] The limit safety system in an extreme working condition of an intelligent vehicle according to the embodiments of the present disclosure will be described below with reference to the accompanying drawings.

[0044] Fig. 1 is a structure diagram of a limit safety system in an extreme working condition of an intelligent vehicle according to an embodiment of the present disclosure, as shown in Fig. 1, including:

[0045] A vehicle sensing system 10 supporting a limit safety function, for collecting environmental information in an extreme working condition of a vehicle, and sending the environmental information to an automatic driving system and a high-level auxiliary driving system 20 in real time through Ethernet or CAN communication;

[0046] The automatic driving system and the high-level auxiliary driving system 20 include a limit safety function decision module 21 and a limit safety function execution module 22, wherein;

[0047] The limit safety function decision module 21 comprises a limit collision avoidance judgment module, a limit cornering judgment module, a limit vehicle condition recognition module, a limit state recognition module and a safety boundary monitoring module; the limit safety function decision module is used for acquiring environmental information and receiving input information based on a data-driven reinforcement learning strategy in an unknown environment and vehicle state and execution information fed back by the limit safety function execution module 22;

[0048] The limit safety function execution module 22 is used for acquiring decision information sent by the limit safety function decision module 21 and outputting vehicle state and execution information.

[0049] To achieve the above object, the intelligent vehicle limit safety system in extreme working conditions of the present disclosure adopts the following technical solutions; limit safety function integrated in intelligent vehicle advanced driver assistance system (ADAS) or automatic driving system; modular limit safety function decision and control architecture, including extreme scenario decision intervention model, limit safety function optimizer and vehicle reaction evaluation system; limit operation execution method with the advantages of integrating various solving technologies, including input fusion control mechanism and cross-system collaborative input execution mechanism.

[0050] Exemplarily, a vehicle sensing system supporting limit safety function is proposed, including a single or multiple following sensors or sensor combinations: camera, millimeter wave radar and combined inertial navigation. Since extreme working conditions often occur at high speed, other sensors that can be used for sensing obstacles at high speed should also be included. The system sends environmental information to the limit safety function decision and execution module in real time through Ethernet or CAN communication.

[0051] Exemplarily, a limit safety function integrated in intelligent vehicle advanced driver assistance system (ADAS) or automatic driving system (ADS) is proposed, including but not limited to using operations related to dynamic near-stable, such as drift operation. This function can be integrated into the active safety module of the vehicle advanced driver assistance system (ADAS) or automatic driving system (ADS), and when the limit safety function intervenes, it raises its own priority, takes over or shields the vehicle stability function or related active safety function, including but not limited to anti-lock braking system or electronic stability control.

[0052] Exemplarily, a limit safety function intervention judgment module is proposed. Since extreme working conditions account for a small proportion in driving scenarios, only this intervention judgment module runs all the time, real-time monitoring limit collision avoidance requirements, limit motion requirements for cornering, extreme conditions of vehicle conditions, and dynamic stability conditions of driving working conditions, and comprehensively judging whether to trigger the limit safety function.

[0053] Exemplarily, an evaluation system for intelligent vehicle evasive response in extreme scenarios is constructed. In view of the characteristics of the extreme scenario, the vehicle evasive response in the extreme scenario is evaluated in combination with the direct risk of the scenario, the expected risk of the scenario, the operation evasive level, and the operation limit degree.

[0054] Exemplarily, a limit safety decision solver is constructed. In combination with the aforementioned vehicle evasive response evaluation system in extreme scenarios, a method including but not limited to iterative optimization and dynamic optimization is adopted to solve the target trajectory or end-to-end input information. It is worth noting that the decision module of the present disclosure is compatible with various model-based or data-driven solving methods. The decision solving output matching the current extreme scenario is all transmitted to the limit safety function execution module.

[0055] Exemplarily, a limit operation fusion mechanism with the advantages of integrating various solving techniques is proposed. Hierarchical decision usually has strong environmental adaptability, and usually outputs target trajectory information. Data-driven decision control usually has better global adaptability, and usually directly outputs action information. The present disclosure adopts a fusion control mechanism to combine the advantages of the two technical routes.

[0056] The fusion control mechanism is used for the case where the hierarchical decision and the data-driven decision have the same goal: deploying the decision in the data generation environment and generating a trajectory, and replacing the trajectory information output by the hierarchical decision with the trajectory. Further, a model-based trajectory tracker is used to solve the closed-loop input Data-driven direct generation of feedforward input The two are added or weighted to generate the final input to be executed.

[0057] The fusion control mechanism is used for the case where the hierarchical decision and the data-driven decision have different goals: comparing the goal gap between the data-driven decision and the hierarchical decision, and the data-driven strategy directly generates action information a Data , the action information a is obtained by tracking the target trajectory of the hierarchical decision model , the final input to be executed is generated by considering the target difference, and the generation method includes but is not limited to neural network fusion, weighted fusion, etc. t = Diff target (a Bata ,a model )

[0058] Exemplarily, a cross-system coordinated limit action execution is proposed. For the execution input output by the fusion control mechanism, four-wheel drive and rear-wheel drive are combined in driving to precisely control driving slip and expand the limit driving control boundary of the vehicle; friction braking and motor braking are combined in braking to provide precise backup limit driving capability guarantee; active steering and differential steering are combined in steering to provide the maximum attitude adjustment capability for limit safety functions. Further, the vehicle executes limit safety maneuvers according to the execution target input of each actuator. The executed maneuver and the scenario evolution caused thereby are transmitted to the evaluation system of the risk avoidance reaction for iterative / dynamic safety decision-making.

[0059] Specifically, the limit safety system of the intelligent vehicle in extreme conditions proposed by the present disclosure can improve the emergency capability in extreme conditions, as shown in FIG. 1, wherein,

[0060] Module 10 gives the architecture of the vehicle sensing system supporting the limit safety function, which is composed of a camera, a millimeter wave radar, and a combined inertial navigation IMU. It transmits data in a fusion manner through Ethernet and CAN FD with the automatic driving system and the advanced level assisted driving system.

[0061] Module 21 gives the architecture of the limit safety function decision module, which is composed of multiple modules such as intervention judgment, comprehensive evaluation, and optimization solution. Module 21 obtains environmental information from module 10 and vehicle state and execution information from module 22.

[0062] Module 22 gives the architecture of the limit safety function execution, which is composed of multiple modules such as data processing, input fusion, and coordinated execution. Module 22 obtains decision information from module 21, and in the present embodiment, module 21 receives input information based on a data-driven reinforcement learning strategy in an unknown environment.

[0063] In the present embodiment, module 10 detects a sharp curve environment with a diameter of 11 m and sends relevant information to the limit safety function decision module. The limit curve judgment module gives an activation flag and transmits it to the limit safety decision module. The limit safety decision module triggers the limit safety function based on the limit curve situation, vehicle speed, and driver intent, aiming to guarantee the limit drift curve.

[0064] After the limit safety decision module judges the intervention, the vehicle stability function and other active safety functions such as collision avoidance are shielded.

[0065] It can be understood that in the present embodiment, only the items related to the scenario of the embodiment in the evaluation system are listed. Specifically, only the safety and performance evaluation indexes of the limit curve scenario are listed.

[0066] In this embodiment, the evaluation index related to the limit cornering performance and safety includes immediate evaluation and terminal evaluation. The immediate cornering performance index includes: the evaluation term r p :

[0067] where k pl and k pv are negative parameters; l is the lateral deviation of the vehicle relative to the center of the curve in the Frenet coordinate system, and v is the current vehicle speed; l ref (s) is the lateral displacement of the current curve center at the current curve of the formation suboptimal trajectory; is the maximum vehicle speed under the specific stability constraint under the current adhesion. This reward encourages the vehicle to travel along the pre-optimized path at the maximum speed allowed by the stability constraint, thereby improving the exploration efficiency in the iterative optimization process.

[0068] The immediate evaluation term r β for rewarding high side slip angle encourages the vehicle to turn with a higher center of mass slip angle:

[0069] where k β is negative, v x is the longitudinal vehicle speed in the vehicle coordinate system, v y is the lateral vehicle speed in the vehicle coordinate system, and β is the center of mass side slip angle of the vehicle.

[0070] The terminal reward term r t for rewarding the comprehensive performance of the limit safety function encourages safe and fast cornering: r t = (1-χ)k t1 +k t2 χ(t f -t ref )

[0071] where χ represents a parameter of the final state of the vehicle: χ = 1 indicates that the vehicle has safely completed the cornering task, and χ = 0 indicates that an unsafe event has occurred, such as a collision with the track boundary or a rollover. In addition, t f represents the time required for the vehicle to reach the terminal state, and t ref represents the total travel time of the pre-optimized trajectory. The constants k t1 and k t2 are negative and positive, respectively, to punish unsafe terminal states and reward as short a limit cornering time as possible.

[0072] In this embodiment, the limit safety solver uses an end-to-end iterative solver based on reinforcement learning, and the output is directly the action information. The state space and action space constructed according to the actual situation of the deployed vehicle are determined, and it is a common technology in the field of using reinforcement learning optimization. This embodiment will not be expanded in detail. In this embodiment, the iterative method is a Critic and Actor deep neural network, and the Critic network parameters are trained according to the minimization of the time difference loss function; the Actor network parameters are trained by maximizing the value function. The above is only one specific example given in this embodiment, and the strategy of using imitation learning and reinforcement learning is included in this embodiment.

[0073] In this embodiment, the data acquisition environment used in the limit safety solver is the Carsim platform, in which the vehicle parameters are common Class-C parameters, and the interfaces of the stability function can be opened. This embodiment is compatible with self-built vehicle dynamics model or other dynamics response data acquisition platform.

[0074] In this embodiment, the friction coefficient in the real environment is close to but not completely consistent with the data acquisition platform, and the inconsistent tire type is introduced as a disturbance term.

[0075] To achieve precise control over the corner, the control strategy based on reinforcement learning is deployed to the corresponding scene in the data acquisition environment to obtain the optimal trajectory T p in the current scene, which is used as the target trajectory information of the limit safety function execution module: T p ={S i |simulate(P v ,V c ,S ini ,π),i=1,2,3,…,n}

[0076] where S represents the vehicle state along the estimated trajectory, and S ini represents the vehicle state when entering the corner. Note that since the estimated trajectory is used to generate feedback input, T p needs to be converted into a Cartesian coordinate trajectory The conversion details are omitted here.

[0077] In this embodiment, the proportional-integral-derivative controller is used to track the trajectory to obtain the supplementary input a PID . Note that this embodiment includes all methods of obtaining trajectory tracking input, including but not limited to linear quadratic Gaussian controller and model predictive controller. Specifically, the calculated input will be further processed by the fusion mechanism.

[0078] In this embodiment, the end-to-end action output by reinforcement learning will be an important feedforward reference and will be recorded as a RLThis input usually has good global optimality and worrying environmental adaptability. Specifically, the input will be directly passed to the input fusion mechanism for further processing.

[0079] In this embodiment, since the hierarchical decision and the data-driven decision target are the same, a direct summation is used to obtain the execution input: a t = a RL + a PID

[0080] In this embodiment, the to-be-executed input a t of the fusion control mechanism output is obtained by using the data-driven method. In this embodiment, four-wheel drive and rear-wheel drive are combined in the driving aspect to precisely control the slip phenomenon of the driving wheels, thereby expanding the control boundary of the vehicle under extreme driving conditions. Through this driving mode, the vehicle can maintain optimal power output and stability under various complex road conditions. In terms of the braking system, the system integrates traditional friction braking and modern motor braking technology, not only providing precise braking force, but also increasing the redundancy of braking to ensure high safety under extreme driving conditions. In terms of steering technology, this embodiment combines active steering systems and differential steering technology, greatly enhancing the maneuverability and stability of the vehicle in emergency situations, providing the maximum attitude adjustment capability for extreme safety functions.

[0081] Further, each actuator of the vehicle executes limit safety maneuvering operations according to the target input of the coordination mechanism. These operations and the resulting scenario evolution information will be passed to the previously established risk avoidance evaluation system, which evaluates the safety response of the vehicle in real time and makes iterative or dynamic safety decision adjustments. This process ensures that the vehicle can take the most appropriate risk avoidance measures in a changing environment and sudden situations, maximizing the emergency response capability and backup level of the intelligent vehicle under extreme conditions.

[0082] The performance of this embodiment in the limit cornering scenario is shown in FIGS. 2 and 3. In the comparative trajectories shown in FIG. 3, it can be found that the control effect of the strategy proposed in this embodiment is obviously better than that of the data-driven strategy-based control effect and that of the feedback controller. Among the three comparison strategies, only the scheme of this embodiment can control the vehicle to pass through a 180-degree U-turn curve with a diameter of 11 meters at a high side slip angle limit state.

[0083] The beneficial effects of the present disclosure are:

[0084] 1. Enhancing emergency handling capability and backup level: The present disclosure allows intelligent vehicles to exhibit superior emergency handling capability under extreme working conditions by integrating advanced dynamics control technology. This includes but is not limited to efficient response capability in emergency braking, extreme steering, or high-speed cornering, etc. In addition, through the backup mechanism of multiple braking and driving systems, the safety and reliability of the vehicle under extreme conditions are improved.

[0085] 2. Optimizing decision-making and control accuracy: The use of a modular extreme safety function decision-making and control architecture, which integrates extreme scenario decision-making intervention models and extreme safety function optimizers, provides precise decision support for a wide range of extreme scenarios. This enables the vehicle to make the most appropriate dynamic response in various extreme environments, significantly improving the accuracy and efficiency of control.

[0086] 3. Enhancing the dynamics performance of the vehicle: Through advanced sensing systems and control algorithms, the present disclosure enables the vehicle to operate near the dynamics limit, such as high-speed cornering and emergency obstacle avoidance, fully utilizing the vehicle's motion potential. This not only enhances the performance of the vehicle, but also greatly expands its operating range and safety boundaries.

[0087] 4. Achieving high integration and system synergy: Through a fusion control mechanism that integrates the advantages of various solution techniques, the present disclosure achieves cross-system collaborative execution. This synergy not only improves the consistency and coordination of operations, but also optimizes the reaction speed and efficiency of the entire vehicle system, especially when dealing with complex and variable driving environments.

[0088] The intelligent vehicle extreme working condition limit safety system according to the embodiments of the present disclosure, under extreme working conditions, performs operations involving but not limited to dynamics critical stability or limit motion capability, to fully utilize the motion potential of each system of the intelligent vehicle, and to enhance the emergency handling capability and backup level of the intelligent vehicle under extreme working conditions.

[0089] In the description of the present specification, the description referring to the terms "one embodiment", "some embodiments", "an example", "a specific example", or "some examples" and the like means that the specific features, structures, materials or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present disclosure. In the present specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any appropriate manner in any one or more embodiments or examples. Furthermore, the different embodiments or examples described in the present specification and the features of the different embodiments or examples can be combined and combined by those skilled in the art without contradiction.

[0090] Moreover, the terms "first", "second", "third", etc. are used herein for descriptive purposes only and cannot be construed as indicating or implying relative importance or an ordered ranking such that the technical features indicated with such terms are in any way prioritized over or above each other. Thus, features defined with "first", "second", etc. can implicitly or explicitly include at least one of such features. In the description of the disclosure, the meaning of "a plurality" is at least two, e.g. two, three, etc., unless explicitly specified otherwise.

Claims

An extreme condition limit safety system for intelligent vehicles, comprising: a vehicle sensing system supporting limit safety functions, configured to collect environment information of the vehicle in extreme conditions and send the environment information to an autonomous driving system and a high-level assisted driving system in real time through Ethernet or CAN communication; the autonomous driving system and the high-level assisted driving system, comprising a limit safety function decision module and a limit safety function execution module, wherein: the limit safety function decision module comprises a limit collision avoidance judgment module, a limit cornering judgment module, a limit vehicle condition recognition module, a limit state recognition module and a safety boundary monitoring module; the limit safety function decision module is configured to obtain environment information and receive input information based on a data-driven reinforcement learning strategy in an unknown environment and vehicle state and execution information fed back by the limit safety function execution module; the limit safety function execution module is configured to obtain decision information sent by the limit safety function decision module and output vehicle state and execution information. The system of claim 1, wherein The vehicle sensing system supporting limit safety functions comprises a camera, a millimeter wave radar and a combined inertial navigation system (IMU). The system of claim 1 or 2, wherein, The limit safety function decision module is further configured to: solving the target trajectory based on the evaluation results and using an iterative optimization or dynamic optimization method through a limit safety decision solver; wherein the evaluation results include evaluation of vehicle avoidance reaction in extreme scenarios based on characteristics of the extreme scenarios, combined with direct risk of the scenarios, expected risk of the scenarios, operation risk avoidance level and operation limit degree; transmitting all target trajectories matching the decision solving of the current extreme scenario to the limit safety function execution module. The system of claim 3, wherein, The system is further configured to fuse the hierarchical decision control and the data-driven decision control using a fusion control mechanism; wherein, The fusion control mechanism aims at the same situation of hierarchical decision control and data-driven decision control: deploying decisions and generating trajectories in a data generation environment, replacing the trajectory information output by hierarchical decisions with trajectories, and using a model-based trajectory tracker to solve the closed-loop input Direct generation of feedforward input using data drive The two are added or weighted to generate the final input to be executed: The fusion control mechanism addresses situations where hierarchical decision control and data-driven decision control have different objectives: by comparing the objective gap between data-driven decision and hierarchical decision, the data-driven strategy directly generates action information a. Data Tracking the hierarchical decision target trajectory to obtain action information a model The final input to be executed is generated based on the differences in the target. The generation methods include neural network fusion and weighted fusion. a t = Diff target (a Data , a model ). The system of claim 4, wherein, for the execution input output by the fusion control mechanism, combining four-wheel drive and rear-wheel drive in driving to control driving slip and expand the limit driving control boundary of the vehicle; combining friction braking and motor braking in braking to provide backup limit braking capability guarantee; combining active steering and differential steering in steering to provide the maximum attitude adjustment capability for the limit safety function; the vehicle executes the target input through each actuator to perform limit safety maneuvers. The system of any one of claims 1-5, wherein The environment information is a sharp corner environment, and the target of the limit safety decision module is to ensure limit drift cornering with comprehensive safety; Evaluation indices related to the limit cornering performance and safety include immediate evaluation and final evaluation: immediate cornering performance indices include: evaluation term r for tracking the scenario suboptimal trajectory p : wherein k pl and k pv are negative parameters; l is the lateral deviation of the vehicle with respect to the center of the curve in the Frenet coordinate system, v is the current vehicle speed; l ref (s) is the lateral displacement of the current curve center of the current curve at the platoon suboptimal trajectory; the maximum vehicle speed under the current adhesion and specific stability constraints. The system of claim 6, wherein, instantaneous reward term r for rewarding high side slip angles β , the vehicle is encouraged to corner with higher center of mass slip angles: where k β is negative, v x is the longitudinal vehicle speed in the vehicle coordinate system, v y is the lateral vehicle speed in the vehicle coordinate system, and β is the vehicle's mass center side slip angle. End reward term r for rewarding the overall performance of the limit safety function t , encouraging safe fast cornering: r t = (1 - x) - k t1 + k t2 - x - (t f - t ref ) where χ represents a parameter of the final state of the vehicle: χ = 1 indicates that the vehicle has safely completed the turn task, χ = 0 indicates that an unsafe event has occurred, t f represents the time required for the vehicle to reach the end state, t ref represents the total travel time of the pre-optimized trajectory; the constant k t1 and k t2 are negative and positive values, respectively, for penalizing the unsafe end state and rewarding the limit turn time as short as possible. The system of claim 3, wherein, The limit safety decision solver uses an end-to-end iterative solver based on reinforcement learning, and the output is directly action information; the iterative method is a Critic and Actor deep neural network, the Critic network parameters are trained according to the minimization of the time difference loss function; the Actor network parameters are trained by maximizing the value function. The system of claim 3, wherein, The data collection environment used in the limit safety decision solver is a Carsim platform; the control strategy based on reinforcement learning is deployed to the corresponding scene in the data collection environment, and the optimal trajectory T under the current scene is obtained p The trajectory is the target trajectory information of the limit safety function execution module: T p = {S i | simulate(P v , V c , S ini , π), i = 1, 2, 3,..., n} where S denotes the vehicle state along the estimated trajectory, S ini denotes the vehicle state at the entry of the curve. The system of claim 9, wherein, Since the estimated target trajectory is used to generate the feedback input, it is necessary to transform T p Converting to Cartesian coordinate trajectory Tracking a trajectory using a proportional-integral-derivative controller Obtaining supplemental input a PID ; The end-to-end action output by reinforcement learning will serve as an important feedforward reference, recorded as a RL ; the hierarchical decision and data-driven decision have the same goal, and the direct summation is used to obtain the execution input: a t = a RL + a PID .

Images