Multi-vehicle collision mitigation system for autonomous vehicle and control method

Abstract

The invention discloses a multi-vehicle collision mitigation system for an autonomous vehicle and a control method, and aims to improve the decision accuracy of the vehicle in an emergency and ensure the safety and comfort of passengers. The system comprises an information sensing module, a vehicle dynamics module, a human body injury prediction module, a collision mitigation planning module and a motion tracking module. The information sensing module is responsible for collecting surrounding environment data, the vehicle dynamics module predicts the vehicle state and detects potential collision, the human body injury prediction module predicts the collision injury degree based on machine learning, the collision mitigation planning module minimizes injury through an optimization algorithm, and the motion tracking module executes corresponding vehicle control actions. According to the system, by comprehensively analyzing the surrounding environment of the vehicle and predicting the collision consequence, the optimal vehicle movement track can be planned when multi-vehicle collision is about to occur, so that damage to traffic participants is minimized, more serious secondary collision caused by emergency obstacle avoidance is avoided, and the overall road safety level is improved.

Description

Technical Field

[0001] The present invention relates to the field of vehicle active safety technology, and in particular to a multi-vehicle collision mitigation system and a control method for an autonomous driving vehicle. Background Art

[0002] Autonomous driving technology aims to enable vehicles to drive autonomously without direct human intervention. By integrating advanced sensors, computer vision, machine learning and other technologies, autonomous vehicles can perceive the surrounding environment, make decisions, and control the vehicle to achieve safe and efficient driving. With the advancement of technology, autonomous driving has moved from the laboratory to practical applications, but it still faces many technical challenges, especially in path planning.

[0003] At present, the path planning of autonomous vehicles mainly relies on setting the initial point and the target point, and realizing the functions such as following the vehicle and changing lanes based on the predetermined algorithm. This method mainly relies on the vehicle's own state information (such as speed, acceleration, etc.) and the state information of surrounding vehicles (such as position, speed, etc.). However, the existing path planning algorithms mainly focus on how to achieve the optimal path selection from the starting point to the end point under the premise of complying with traffic rules, and often ignore the safety of the passengers themselves.

[0004] In practical applications, when vehicles encounter emergency obstacle avoidance requirements, existing path planning algorithms often fail to make accurate lane keeping and lane changing decisions. For example, when an obstacle suddenly appears in front of the vehicle or another vehicle brakes suddenly, the vehicle needs to quickly assess the surrounding environment and take measures to avoid a collision. However, existing algorithms may not be able to handle complex decision-making processes in real time or provide adequate safety guarantees when dealing with such emergency situations. Summary of the invention

[0005] The purpose of the present invention is to provide a multi-vehicle collision mitigation system and control method for autonomous driving vehicles, which can take into account both occupant safety and comfort and have a path planning capability with higher decision-making accuracy in emergency situations.

[0006] The above technical objectives of the present invention are achieved through the following technical solutions:

[0007] A multi-vehicle collision mitigation system for autonomous driving vehicles, comprising an information perception module, a vehicle dynamics module, a human injury prediction module, a collision mitigation planning module, and a motion tracking module;

[0008] The information perception module is used to perceive the vehicle's surrounding environment information in a scenario where a collision is about to occur, and includes an information collection unit and a calculation unit. The information collection unit is used to collect surrounding object information: surrounding object type, surrounding object movement direction, surrounding object movement speed, and surrounding object position. The calculation unit is used to calculate the relative information between the vehicle and the surrounding objects: relative movement direction, relative movement speed, and relative distance.

[0009] The vehicle dynamics module is based on the vehicle freedom dynamics model and includes a state prediction unit and a collision detection unit, wherein the state prediction unit is used to calculate and predict the state of the vehicle and surrounding objects, and the collision detection unit is used to detect whether the vehicle collides with surrounding objects;

[0010] The human injury prediction module is used to predict the injuries of traffic participants caused by collision accidents. The module applies machine learning algorithm training based on the traffic database to generate a human injury prediction model. The model can predict the injuries of traffic participants based on the collision object, relative movement direction, and relative movement speed;

[0011] The collision mitigation planning module is used to plan vehicles in a scenario where a collision is about to occur, so as to minimize the damage to traffic participants caused by the collision, and includes a cost function unit and a discretization unit, wherein the cost function unit is a damage cost function used to calculate the optimal vehicle control amount, and the discretization unit is used to generate different vehicle control amounts;

[0012] The motion tracking module is used to control the vehicle according to the vehicle motion planned in the collision mitigation planning module, specifically including brake pedal and steering wheel angle control.

[0013] Further preferably, the surrounding object types include two-wheeled vehicles, motor vehicles, and pedestrians.

[0014] Further preferably, the vehicle control quantity includes a brake pedal and a steering wheel angle.

[0015] The present invention also discloses a control method for a multi-vehicle collision mitigation system for an autonomous driving vehicle, comprising the following steps:

[0016] Step 1: In a scenario where a collision is about to occur, the autonomous vehicle establishes a self-vehicle coordinate system with the self-vehicle as the center;

[0017] Step 2: Based on the information collection unit in the information perception module, the vehicle camera is used to determine the type of surrounding objects, and the vehicle radar is used to collect the movement direction, movement speed, and coordinates of the surrounding objects in the vehicle coordinate system centered on the vehicle;

[0018] Step 3, based on the calculation unit in the information perception module, the relative motion direction θ, relative motion speed V, and relative distance P between the vehicle and the surrounding objects are calculated according to the information collected by the information collection unit in step 2;

[0019] Step 4: Compare the vehicle state information at the current time t with the state information of the surrounding objects and the control amount u t , the movement direction and speed of the surrounding objects are input into the state prediction unit of the vehicle dynamics module, and the state s of the vehicle at the next time step is predicted in the prediction time domain Np t+1 The state of the surrounding objects is as shown below:

[0020] s t+1 =f9s t ,u t )

[0021] Step 5: The predicted vehicle state s at the next time step is t+1 The collision detection unit in the vehicle dynamics module inputs the state of the surrounding objects, and the coordinates of the upper left, lower left, upper right and lower right boundary points of the ego vehicle in the ego vehicle coordinate system form the ego vehicle rectangular frame. Similarly, the coordinates of the upper left, lower left, upper right and lower right boundary points of the surrounding objects in the ego vehicle coordinate system form the object rectangular frame.

[0022] Step 6, determine whether the vehicle rectangle and the object rectangle intersect, if so, output a collision signal, if not, repeat steps 4 and 5;

[0023] Step 7: Output the collision signal to the human body injury prediction module. Based on the state of the vehicle at the time of the collision and the state of the object that collided, the human body injury prediction model h(·) is used to predict the injury condition J of the traffic participant, as shown in the following formula:

[0024] J=h(V,θ,P)

[0025] Step 8: Generate different vehicle state control quantities k steps forward from the current time t based on the discretization unit The generated control volume Input it into the state prediction unit of the vehicle dynamics module, repeat steps 4-7, and obtain the damage status of different traffic participants at the current time t and k steps forward under the current vehicle state control quantity

[0026] Step 9: Based on the different vehicle state control quantities obtained in step 8 Injury of traffic participants The cost function unit in the collision mitigation planning module establishes a collision mitigation cost function J, as shown in the following formula. The cost function includes the control amount constraint of the ego vehicle: Vehicle state quantity With reference state Constraints and the amount of traffic participants injured

[0027] Step 10: Obtain the optimal vehicle state control value by minimizing the collision mitigation cost function minJ

[0028] Step 11: The optimal vehicle state control quantity obtained in step 10 is The input is sent to the motion tracking module to control the vehicle’s brake pedal and steering wheel angle;

[0029] Step 12: In a scenario where a collision is about to occur, all autonomous vehicles repeat steps 1-11 to achieve multi-vehicle collision mitigation.

[0030] In summary, the present invention has the following beneficial effects:

[0031] First, the multi-vehicle collision mitigation system proposed in the present invention can solve the problem of traditional autonomous driving vehicles not considering human injuries under marginal conditions (such as inevitable collision) through human injury prediction, thereby minimizing the injuries of traffic participants in collision accidents.

[0032] Secondly, the collision mitigation cost function proposed in the present invention comprehensively considers the damage to traffic participants caused by the vehicle and all surrounding potential collision objects when multiple vehicles are about to collide. Therefore, the collision mitigation system proposed in the present invention can avoid accidents such as traditional autonomous driving vehicles causing a more serious collision with the second collision object due to emergency avoidance of the first collision object.

[0033] Third, the multi-vehicle collision mitigation system proposed in the present invention can coordinately control all autonomous driving vehicles involved in complex collision environments. By predicting the movement of the vehicle and the surrounding potential collision objects at each time step, all autonomous driving vehicles designed for collision can plan the optimal movement mode. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 It is a structural block diagram of the present invention;

[0035] Figure 2 It is a logic block diagram of the control method of the present invention. DETAILED DESCRIPTION

[0036] In order to make those skilled in the art better understand the technical solution of the present invention, the present invention is further described in detail below in conjunction with embodiments. It will be understood by those skilled in the art that the following embodiments are only used to illustrate the present invention and should not be regarded as limiting the scope of the present invention.

[0037] Example 1, reference Figure 1 ,A multi-vehicle collision mitigation system for autonomous vehicles includes an information perception module, a vehicle dynamics module, a human injury prediction module, a collision mitigation planning module, and a motion tracking module.

[0038] The information perception module is used to perceive the vehicle's surrounding environment information in a scenario where a collision is about to occur, and includes an information collection unit and a calculation unit. The information collection unit is used to collect surrounding object information: surrounding object type (two-wheeled vehicle, motor vehicle, pedestrian), surrounding object movement direction, surrounding object movement speed, and surrounding object position. The calculation unit is used to calculate the relative information between the vehicle and the surrounding objects: relative movement direction, relative movement speed, and relative distance.

[0039] The vehicle dynamics module is based on a 2-DOF vehicle dynamics model and includes a state prediction unit and a collision detection unit, wherein the state prediction unit is used to calculate and predict the state of the vehicle and surrounding objects, and the collision detection unit is used to detect whether the vehicle collides with surrounding objects;

[0040] The human injury prediction module is used to predict the injuries of traffic participants caused by collision accidents. The module applies machine learning algorithm training based on the traffic database to generate a human injury prediction model. The model can predict the injuries of traffic participants based on the collision object, relative movement direction, and relative movement speed.

[0041] The collision mitigation planning module is used to plan vehicles in scenarios where a collision is about to occur so as to minimize the damage to traffic participants caused by the collision. It includes a cost function unit and a discretization unit, wherein the cost function unit is a damage cost function used to calculate the optimal vehicle control quantity, and the discretization unit is used to generate different vehicle control quantities (brake pedal and steering wheel angle).

[0042] The motion tracking module is used to control the vehicle according to the vehicle motion planned in the collision mitigation planning module, specifically including brake pedal and steering wheel angle control.

[0043] Example 2, reference Figure 2 The present invention also proposes a multi-vehicle collision mitigation system control method for an autonomous driving vehicle:

[0044] Step 1: In a scenario where a collision is about to occur, the autonomous vehicle establishes a self-vehicle coordinate system with the self-vehicle as the center;

[0045] Step 2: Based on the information collection unit in the information perception module, the vehicle camera is used to determine the type of surrounding objects (two-wheeled vehicles, motor vehicles, pedestrians), and the vehicle radar is used to collect the movement direction, movement speed, and coordinates of the surrounding objects (first object, second object, third object...) in the vehicle coordinate system centered on the vehicle;

[0046] Step 3, based on the calculation unit in the information perception module, the relative motion direction θ, relative motion speed V, and relative distance P between the vehicle and surrounding objects (first object, second object, third object...) are calculated according to the information collected by the information collection unit in step 2;

[0047] Step 4: Compare the vehicle state information at the current time t with the state information of the surrounding objects (first object, second object, third object...) and the control amount u t The vehicle’s direction of movement, speed of movement, and the directions and speeds of movement of the surrounding objects (first object, second object, third object…) are input into the state prediction unit of the vehicle dynamics module to predict the vehicle’s state s at the next time step in the prediction time domain Np. t+1 and the surrounding object states (first object state, second object state, third object state…), as shown in the following formula:

[0048] s t+1 =f(s t ,u t )

[0049] Step 5: The predicted vehicle state s at the next time step is t+1 The collision detection unit in the vehicle dynamics module inputs the surrounding object states (first object predicted state, second object predicted state, third object predicted state...), and the coordinates of the upper left, lower left, upper right, and lower right boundary points of the self-vehicle in the self-vehicle coordinate system form the self-vehicle rectangular frame. Similarly, the coordinates of the upper left, lower left, upper right, and lower right boundary points of the surrounding objects (first object, second object, third object...) in the self-vehicle coordinate system form the object rectangular frames (first object rectangular frame, second object rectangular frame, third object rectangular frame...);

[0050] Step 6, judging whether the vehicle rectangular frame and the object rectangular frame (the first object rectangular frame, the second object rectangular frame, the third object rectangular frame...) intersect, if so, outputting a collision signal, if not, repeating steps 4 and 5;

[0051] Step 7: Output the collision signal to the human body injury prediction module. Based on the state of the vehicle at the time of collision and the state of the object that collided, the human body injury prediction model h(·) is used to predict the injury condition J of the traffic participant, as shown in the following formula.

[0052] J=h(V,θ,P)

[0053] Step 8: Generate different vehicle state control quantities k steps forward from the current time t based on the discretization unit The generated control volume Input it into the state prediction unit of the vehicle dynamics module, repeat steps 4-7, and obtain the damage status of different traffic participants at the current time t and k steps forward under the current vehicle state control quantity

[0054] Step 9: Based on the different vehicle state control quantities obtained in step 8 Injury of traffic participants The cost function unit in the collision mitigation planning module establishes a collision mitigation cost function J, as shown in the following formula. The cost function includes the control amount constraint of the ego vehicle: Vehicle state quantity With reference state Constraints and the amount of traffic participants injured

[0055] Step 10: Obtain the optimal vehicle state control value by minimizing the collision mitigation cost function minJ

[0056] Step 11: The optimal vehicle state control quantity obtained in step 10 is The input is sent to the motion tracking module to control the vehicle’s brake pedal and steering wheel angle;

[0057] Step 12: In a scenario where a collision is about to occur, all autonomous vehicles repeat steps 1-11 to achieve multi-vehicle collision mitigation.

[0058] At present, the path planning of autonomous vehicles mainly relies on setting the initial point and the target point, and realizing the functions such as following the vehicle and changing lanes based on the predetermined algorithm. However, the existing methods often ignore the safety and comfort of the passengers, and cannot make accurate lane keeping and lane changing decisions when encountering emergency obstacle avoidance needs. To this end, we proposed a new path planning method, which improves the accuracy and robustness of environmental perception by introducing the passenger comfort index and combining multi-sensor fusion technology; at the same time, by optimizing the algorithm performance and real-time performance, it ensures that the vehicle can quickly evaluate the surrounding environment and make the best decision in an emergency. These improvements not only improve driving safety and passenger comfort, but also significantly enhance the driving performance of autonomous vehicles in complex road environments.

[0059] The above embodiments are merely explanations of the present invention and are not limitations of the present invention. After reading this specification, those skilled in the art may make modifications to the embodiments without any creative contribution as needed. However, such modifications are protected by the patent law as long as they are within the scope of the claims of the present invention.

Claims

1. A multi-vehicle collision mitigation system for an autonomous driving vehicle, characterized in that: It includes information perception module, vehicle dynamics module, human injury prediction module, collision mitigation planning module, and motion tracking module; The information perception module is used to perceive the vehicle's surrounding environment information in a scenario where a collision is about to occur, and includes an information collection unit and a calculation unit. The information collection unit is used to collect surrounding object information: surrounding object type, surrounding object movement direction, surrounding object movement speed, and surrounding object position. The calculation unit is used to calculate the relative information between the vehicle and the surrounding objects: relative movement direction, relative movement speed, and relative distance. The vehicle dynamics module is based on the vehicle freedom dynamics model and includes a state prediction unit and a collision detection unit, wherein the state prediction unit is used to calculate and predict the state of the vehicle and surrounding objects, and the collision detection unit is used to detect whether the vehicle collides with surrounding objects; The human injury prediction module is used to predict the injuries of traffic participants caused by collision accidents. The module applies machine learning algorithm training based on the traffic database to generate a human injury prediction model. The model can predict the injuries of traffic participants based on the collision object, relative movement direction, and relative movement speed; The collision mitigation planning module is used to plan vehicles in a scenario where a collision is about to occur, so as to minimize the damage to traffic participants caused by the collision, and includes a cost function unit and a discretization unit, wherein the cost function unit is a damage cost function used to calculate the optimal vehicle control amount, and the discretization unit is used to generate different vehicle control amounts; The motion tracking module is used to control the vehicle according to the vehicle motion planned in the collision mitigation planning module, specifically including brake pedal and steering wheel angle control.

2. A multi-vehicle collision mitigation system for an autonomous driving vehicle according to claim 1, characterized in that: The surrounding object types include two-wheeled vehicles, motor vehicles, and pedestrians.

3. The multi-vehicle collision mitigation system for an autonomous driving vehicle according to claim 1, characterized in that: The vehicle control variables include a brake pedal and a steering wheel angle.

4. A control method for a multi-vehicle collision mitigation system for an autonomous driving vehicle according to any one of claims 1 to 3, characterized in that: The following steps are involved: Step 1: In a collision scenario, the autonomous vehicle establishes a self-vehicle coordinate system with the self-vehicle as the center; Step 2: Based on the information collection unit in the information perception module, the vehicle camera is used to determine the type of surrounding objects, and the vehicle radar is used to collect the movement direction, movement speed, and coordinates of the surrounding objects in the vehicle coordinate system centered on the vehicle; Step 3, based on the calculation unit in the information perception module, the relative motion direction θ, relative motion speed V, and relative distance P between the vehicle and the surrounding objects are calculated according to the information collected by the information collection unit in step 2; Step 4: Compare the vehicle state information at the current time t with the state information of the surrounding objects and the control amount u t , the movement direction and speed of the surrounding objects are input into the state prediction unit of the vehicle dynamics module, and the state s of the vehicle at the next time step is predicted in the prediction time domain Np t+1 The state of the surrounding objects is as shown below: s t+1 =fs t ,u t Step 5: The predicted vehicle state s at the next time step is t+1 The collision detection unit in the vehicle dynamics module inputs the state of the surrounding objects, and the coordinates of the upper left, lower left, upper right and lower right boundary points of the ego vehicle in the ego vehicle coordinate system form the ego vehicle rectangular frame. Similarly, the coordinates of the upper left, lower left, upper right and lower right boundary points of the surrounding objects in the ego vehicle coordinate system form the object rectangular frame. Step 6, determine whether the vehicle rectangle and the object rectangle intersect, if so, output a collision signal, if not, repeat steps 4 and 5; Step 7: Output the collision signal to the human body injury prediction module. Based on the state of the vehicle at the time of the collision and the state of the object that collided, the human body injury prediction model h(·) is used to predict the injury condition J of the traffic participant, as shown in the following formula: J=h(V,θ,P) Step 8: Generate different vehicle state control quantities k steps forward from the current time t based on the discretization unit The generated control volume Input it into the state prediction unit of the vehicle dynamics module, repeat steps 4-7, and obtain the damage status of different traffic participants at the current time t and k steps forward under the current vehicle state control quantity Step 9: Based on the different vehicle state control quantities obtained in step 8 Injury of traffic participants The cost function unit in the collision mitigation planning module establishes a collision mitigation cost function J, as shown in the following formula. The cost function includes the control amount constraint of the ego vehicle: Vehicle state quantity With reference state Constraints and the amount of traffic participants injured Step 10: Obtain the optimal vehicle state control value by minimizing the collision mitigation cost function minJ Step 11: The optimal vehicle state control quantity obtained in step 10 is The input is sent to the motion tracking module to control the vehicle’s brake pedal and steering wheel angle; Step 12: In a scenario where a collision is about to occur, all autonomous vehicles repeat steps 1-11 to achieve multi-vehicle collision mitigation.

Images