A lane-changing method and system for unmanned vehicles based on a DRF model and adaptive preview

Abstract

The application belongs to the technical field of automatic driving, and particularly relates to a kind of human-like lane changing method and system of unmanned vehicle based on DRF model and adaptive preview, comprising: vehicle perceives surrounding environment, obtains perception information;Obtain the uncertainty of driver perception information and driver behavior, and build driver risk field model;The perception information is input into driver risk field model, and the driving risk quantitative value is obtained;According to the kinematics and two-degree-of-freedom dynamics model of vehicle, an adaptive preview method is established;A human-like lane changing decision model is constructed, the risk threshold is calculated according to the perception information, the calculated driving risk quantitative value is compared with the perception risk threshold, and the lane changing decision of unmanned vehicle is made;The application can effectively solve the problems of difficult expansion of automatic driving decision planning scene, not smooth switching of different scenes and low degree of personification by establishing risk field model and combining adaptive preview method.

Description

Technical Field

[0001] This invention belongs to the field of autonomous driving technology for unmanned vehicles, specifically relating to a human-like lane-changing method and system for unmanned vehicles based on DRF model and adaptive prediction. Background Technology

[0002] Vehicle intelligence is one of the three transformative technologies in the automotive industry, playing a crucial role in improving traffic safety and efficiency. Currently, autonomous vehicles still exhibit significant behavioral differences from human-driven vehicles and lack information communication, potentially impacting normal traffic order and efficiency. Learning from human driving intelligence to achieve human-like decision-making in intelligent vehicles is the core of autonomous driving. Lane changing is not only one of the most complex behaviors during driving but also has strong practical significance for autonomous driving technology, making it a current research focus. Current lane changing technologies mainly employ two approaches: rule-based lane changing decision-making and neural network-based lane changing decision-making. Neural network methods struggle to handle complex multi-task scenarios, while rule-based lane changing decision-making typically requires considering multiple scenarios to develop a decision tree and make decisions accordingly, but it cannot cover all driving scenarios and has poor adaptability. Summary of the Invention

[0003] To address the problems existing in the prior art, this invention proposes a human-like lane-changing method for unmanned vehicles based on a DRF model and adaptive pre-aiming. The method includes: the vehicle perceiving the surrounding environment to obtain perception information; acquiring driver perception information and the uncertainty of driver behavior, and constructing a driver risk field model; inputting the perception information into the driver risk field model to obtain a quantitative value of driving risk; establishing an adaptive pre-aiming method based on the vehicle's kinematics and two-degree-of-freedom dynamics model, setting an evaluation function to be optimized using the adaptive pre-aiming method, and extracting the optimal pre-aiming time based on the evaluation function to be optimized; constructing a human-like lane-changing decision model, calculating a perception risk threshold based on historical perception information, comparing the calculated quantitative value of driving risk with the perception risk threshold, and using the human-like lane-changing decision model to perform human-like lane changing for the unmanned vehicle based on the comparison result and the optimal pre-aiming time.

[0004] Preferably, the vehicle's perception of the surrounding environment includes: using sensors to perceive environmental information, wherein the perceived information includes surrounding vehicle information and environmental information, and the environmental information includes road boundary information and surrounding road vegetation location information; and constructing a cost map based on the perceived information.

[0005] Preferably, the expression for the driver risk field model is:

[0006]

[0007] Where z(x,y) represents the driver risk field model, p represents the Gaussian parabola steepness, (x c ,yc R is the center of rotation of the vehicle at the previous moment. car Let s represent the radius of rotation of the vehicle, s represent the arc length of rotation, v represent the instantaneous velocity of the vehicle, and t represent the radius of rotation of the vehicle. la For the driver's forward-looking time, v·t la The forward sight distance is represented by σ, which represents the parabolic width of the Gaussian section.

[0008] Preferably, the process of establishing an adaptive anti-aiming method for unmanned vehicles includes: establishing a vehicle kinematic model and updating the model's parameters; updating the kinematic model based on the updated parameters; updating the vehicle's motion state at the next moment based on the new kinematic model; establishing a two-degree-of-freedom dynamic model of the vehicle and updating its parameters; and establishing a humanized anti-aiming time evaluation function based on the updated model parameters, wherein the expression of the humanized anti-aiming time evaluation function is:

[0009]

[0010] Where α1 is the driving accuracy style coefficient, and α2 is the driving comfort style coefficient; J E The evaluation function is based on the transverse error. This is an evaluation function based on lateral acceleration.

[0011] Preferably, based on risk threshold theory, the model only attempts to keep driving risks below a risk threshold; therefore, the model does not provide a specific trajectory. The process of using a driver decision model to guide an autonomous vehicle in lane changing includes:

[0012] Establish a steering angle update method: by quantifying the driving risk C k With risk threshold C t The process involves comparing the current speed with the expected speed and building a decision tree based on the comparison results. The process of translating perceived risk into control actions includes:

[0013] To make lane changes smoother, the influence of the heading angle on the steering angle is considered, and the steering angle is updated using the following formula:

[0014] δ k+1 =δ k +k φ (φ r -φ car )

[0015] The decision tree is constructed as follows:

[0016] When C k <C t And v k <v des Or C k <C t And v k >vdes At that time, δ k+1 =δ k +k φ (φ r -φ car ), and v k+1 =v k +k v (s ved -v k );

[0017] When C k >C t And v k <v des When, obtain the optimal δ k+1 This makes C k+1 <C t Or minimize C k+1 , where v k+1 =v k +k v (v des -v k );

[0018] When C k >C t And v k >v des When, obtain the optimal δ k+1 Minimize C k+1 ,in

[0019] v k+1 =v k +k vc (C t -C k )+k v (v des -v k ).

[0020] Where, φ r， φ car They are the aiming time t la The heading angle of the road and vehicle after seconds, v des For the desired speed, k φ This is the gain for heading control.

[0021] The optimal parameters are determined through grid search.

[0022] Finally, the velocity v at the next time step is obtained based on the decision-making method. k+1 With steering angle δ k+1 Control the car to change lanes.

[0023] A human-like lane-changing method and system for unmanned vehicles based on DRF model and adaptive preview includes: a data acquisition module, a driving risk assessment module, an adaptive preview module, and a human-like decision-making module;

[0024] The data acquisition module is used to acquire the vehicle's environmental information, wherein the perception information includes obstacle information and environmental information, wherein the obstacle information includes surrounding vehicle information; and the environmental information includes road boundary information and surrounding road vegetation location information.

[0025] The driving risk assessment module is used to calculate the current vehicle quantitative risk, that is, to construct a driver risk field model based on the information obtained by the data acquisition module, and to calculate the driver driving risk quantitative value based on the driver risk field.

[0026] The adaptive anti-aiming module calculates the current optimal anti-aiming time using an anti-aiming time evaluation function and updates the vehicle's state for the next moment based on the anti-aiming time.

[0027] The human-like decision-making module makes lane-changing decisions for the vehicle based on the driving risk quantification value, current speed, expected speed, and risk threshold.

[0028] The beneficial effects of this invention are:

[0029] This invention establishes a driver risk field model that can adapt to various scenarios, solving the problems of difficulty in expanding autonomous driving decision-making and planning scenarios and the lack of smooth transitions between different scenarios. An adaptive anticipation mechanism is established in this model, taking into account driving style and human decision-making characteristics, so that the autonomous vehicle driving actions generated by the decision are closer to human driving actions, thus meeting the needs of human-like driving in autonomous vehicles. Attached Figure Description

[0030] Figure 1 This is a diagram of the unmanned vehicle lane-changing architecture in an embodiment of the present invention;

[0031] Figure 2 This is a flowchart illustrating the lane-changing process for unmanned vehicles in an embodiment of the present invention.

[0032] Figure 3 This is a risk field result diagram in an embodiment of the present invention;

[0033] Figure 4 This is a schematic diagram of the vehicle kinematics model in an embodiment of the present invention;

[0034] Figure 5 This is a diagram of a two-degree-of-freedom vehicle dynamics model under dynamic conditions in an embodiment of the present invention.

[0035] Figure 6 This is a diagram showing the vehicle lane-changing response under dynamic operating conditions in an embodiment of the present invention. Detailed Implementation

[0036] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0037] A human-like lane-changing method and system for unmanned vehicles based on a DRF model and adaptive pre-aiming is disclosed. The method includes: the vehicle perceiving the surrounding environment to obtain perception information; acquiring the driver's perception information and the uncertainty of the driver's behavior, and constructing a driver risk field model; inputting the perception information into the driver risk field model to obtain the driving risk; establishing an adaptive pre-aiming method for the unmanned vehicle based on the vehicle's kinematics and two-degree-of-freedom dynamics model; constructing a human-like decision-making model, calculating the perception risk threshold based on the perception information, comparing the calculated driving risk with the risk threshold, and making lane-changing decisions for the unmanned vehicle.

[0038] A specific implementation of a lane-changing method and system for autonomous vehicles based on DRF model and adaptive anticipation, such as... Figure 1 As shown, the method includes:

[0039] Step 1: The vehicle senses its surroundings.

[0040] In this embodiment, the autonomous vehicle needs to complete the collection and processing of multi-source information, specifically including:

[0041] Step 11: Environmental Perception: Autonomous vehicles perceive their environment by installing various sensors, including active and passive sensors such as radar sensors, lidar sensors, and camera sensors. These sensors are used to perceive environmental information surrounding the autonomous vehicle; this environmental information includes road boundary information (width and location) and the location of surrounding road vegetation.

[0042] Step 12: Obstacle Information: Since various sensors collect different environmental information, this embodiment needs to process the collected environmental information to obtain information about surrounding vehicles (position, direction, and rate of change of speed).

[0043] Step 2: Establish the driver risk field.

[0044] In this embodiment, establishing a driver risk field mainly involves creating a risk field model and generating moving grid points to shift the field. A cost map is generated based on the driver risk assessment, and the product of this map and the risk field is summed over all grid points to obtain the quantified risk. The specific process includes:

[0045] Step 21: Driver risk field model: Establish a risk field model based on the information transmitted from the sensors.

[0046] Step 22: Moving Grid Point Generation: Grid points are created and moved based on velocity and aiming time to make the field dynamic.

[0047] Step 23: Cost Map Generation: Assign risk levels to corresponding obstacles based on the driver's risk assessment, and finally generate a cost map. This cost map can be multiplied by the risk field and summed over all grid points to obtain the quantified risk.

[0048] Step 3: Establish an adaptive preview method.

[0049] In this embodiment, the adaptive pre-aiming method is mainly established by creating a vehicle kinematic model and a two-degree-of-freedom dynamic model to obtain lateral deviation and lateral acceleration deviation, and then establishing an evaluation function to be optimized, thereby extracting the optimal pre-aiming time. The vehicle kinematic model is used to obtain the first derivative information of the vehicle displacement, thereby obtaining some control variables and lateral deviation; the two-degree-of-freedom dynamic model is used to obtain the second derivative information of the vehicle displacement, i.e., stability information, thereby obtaining some control variables and lateral acceleration deviation; the established evaluation function needs to be minimized, and it includes lateral deviation and lateral acceleration deviation information.

[0050] Step 4: Make human-like lane-changing decisions for the vehicle based on the parameters obtained from the adaptive preview method.

[0051] In this embodiment, by inputting the calculated quantified risk and speed information into the information sensing unit, a decision tree is used to calculate the vehicle's required lane-changing risk threshold based on the perceived information, and a lane-changing decision is made based on the driver's satisfaction. This includes: quantifying the driving risk (C...) k ) and risk threshold (C t The system compares the current speed with the expected speed, builds a decision tree, and translates perceived risk into control actions (speed and steering).

[0052] A specific implementation of a lane-changing method and system for autonomous vehicles based on DRF model and adaptive anticipation, such as... Figure 2 As shown, the method includes:

[0053] S1. The autonomous vehicle uses active or passive sensors to perceive the environment and collect obstacle information. These sensors include camera sensors, radar sensors, and lidar sensors. These sensors perceive the environmental information surrounding the autonomous vehicle; this environmental information includes road boundary information (width, position) and the location of surrounding road vegetation. This embodiment requires processing the collected environmental information to obtain surrounding vehicle direction information (front vehicle, rear vehicle, left vehicle, right vehicle), width information, and position information.

[0054] S2. Establish a driver risk field model based on the uncertainty of driver perception and action, and obtain the quantitative risk of vehicle driving by combining perception information; wherein the driver risk field model is expressed as:

[0055]

[0056] Where z(x,y) represents the driver risk field model; p represents the Gaussian parabolic steepness; (x c ,y c R is the vehicle rotation center determined in the previous step. car Let s represent the radius of rotation of the vehicle, s represent the arc length of rotation, v represent the instantaneous velocity of the vehicle, and t represent the radius of rotation of the vehicle. la For the driver's forward-looking time, v·t la The forward sight distance is represented by σ. σ represents the parabolic width of the Gaussian section, which is a function of the rotational arc length s. It allows for asymmetry between the inner and outer edges of the parabola and is related to the vehicle steering angle and vehicle width.

[0057] In this embodiment, the Gaussian cross-section parabolic width function σ of the risk field model is modeled as being related to the vehicle steering angle and vehicle width, and can be expressed as:

[0058] σ(s)=(m+k i |δ|)s+c

[0059] Where, k i The term describes the widening or narrowing of the inner and outer edges of the field when turning, where i=1 represents the inner edge and i=2 represents the outer edge; m is the slope of the field widening or narrowing when going straight; and c represents the field width at the vehicle's location, which is proportional to the vehicle width.

[0060] The field is described using grid points, moves with the grid, and is used in conjunction with a cost map, including:

[0061]

[0062]

[0063] It can be seen that the grid region is 10·v·t wide. la A square region. x b1 x buThese respectively represent the start and end positions of the horizontal grid; y b1 y bu These respectively represent the starting and ending positions of the grid in the vertical direction.

[0064] The rotation angle θ in the kinematics is obtained, thus enabling the movement of the mesh. Its expression is:

[0065]

[0066] Where x represents the starting point x b1 The endpoint is x bu A horizontal grid value array generated at 0.5 intervals, where y represents the grid value starting from the y-th node. b1 The endpoint is y bu A vertical grid value array generated at 0.5 intervals.

[0067] like Figure 3 As shown, the size of the risk field is directly proportional to the speed, and the bias of the risk field is related to the steering wheel angle.

[0068] S3. Establish an adaptive anti-aiming method. This specifically includes: establishing an adaptive anti-aiming method for the vehicle based on its kinematics and two-degree-of-freedom dynamics model.

[0069] Establish vehicle kinematics model and update parameters: vehicle at position (x v ,y v ) around the center of rotation (x c ,y c With radius R car Rotate by an angle θ, rotate by an arc length s, and reach the next position (x). v1 ,y v1 ), where θ is obtained in the moving grid.

[0070] The simplified kinematic model yields the following expression:

[0071] s = R car ·θ

[0072] Where L is the wheelbase of the vehicle; δ is the steering wheel angle.

[0073] The vehicle's motion state at the next moment is updated based on a simplified kinematic model, expressed as follows:

[0074]

[0075] Where φ is the vehicle steering angle (at the current moment), φ k Let φ be the steering angle at the current moment. k+1 dt is the turning angle at the next moment; dt is the time step.

[0076] Construct a two-degree-of-freedom dynamic model of the vehicle and update the model parameters. The expression is as follows:

[0077]

[0078]

[0079] Where u represents the longitudinal velocity of the vehicle; v lat Indicates the lateral speed of the vehicle; I z The z-axis represents the moment of inertia about the z-axis; k1 and k2 represent the lateral stiffness of the front and rear tires, respectively; l f l r Indicates the distance from the front and rear axles to the center of gravity; L is the vehicle wheelbase; i sw The transmission ratio of the vehicle's rotating system; a y This is lateral acceleration.

[0080] Based on the two-degree-of-freedom dynamic model of the vehicle, a humanized evaluation function for the aiming time is established, and its expression is as follows:

[0081]

[0082] Where α1 is the driving accuracy style coefficient, and α2 is the driving comfort style coefficient; J E The evaluation function is based on the transverse error. This is an evaluation function based on lateral acceleration.

[0083] The optimal pre-aiming time is calculated based on the anthropomorphic evaluation function of the pre-aiming time when J is minimized.

[0084] S4. Establish a human-like lane-changing decision-making method, calculate the perceived risk threshold based on perceived information, and make lane-changing decisions based on driver satisfaction.

[0085] The lane-changing decision-making method is as follows: by quantifying the driving risk C k With the calculated risk threshold C t Compare the current speed v k With the expected speed v des By comparing data and building a decision tree, driving risks are transformed into control actions (speed and steering). That is:

[0086] When C k <C t And v k <v des Or C k <C t And v k >v desAt the same time, obtain the road heading angle and vehicle heading angle at the optimal preview time; update the steering angle of the current unmanned vehicle based on the road heading angle and vehicle heading angle at the optimal preview time; predict the speed of the unmanned vehicle in the next time period based on the current speed and the expected speed; control the unmanned vehicle to change lanes based on the updated steering angle and the speed of the unmanned vehicle in the next time period; the formula for updating the steering angle of the current unmanned vehicle and predicting the speed of the unmanned vehicle in the next time period is:

[0087] δ k+1 =δ k +k φ (φ r -φ car )

[0088] v k+1 =v k +k v (v des -v k )

[0089] When C k >C t And v k <v des At that time, the optimal steering angle of the autonomous vehicle is obtained based on the current driving risk and driving speed, and the speed of the autonomous vehicle in the next time period is predicted, where C k+1 <C t Or minimize C k+1 At that time, the optimal steering angle for the autonomous vehicle is determined; the autonomous vehicle is controlled to change lanes based on the optimal steering angle and its speed in the next time period; the formula for predicting the speed of the autonomous vehicle in the next time period is:

[0090] v k+1 =v k +k v (v des -v k )

[0091] C k >C t And v k >v des At that time, the optimal steering angle of the autonomous vehicle is obtained based on the current driving risk and driving speed, and the speed of the autonomous vehicle in the next time period is predicted, where C k+1 When minimized, the optimal steering angle for the autonomous vehicle is obtained; the autonomous vehicle's lane-changing is controlled based on the optimal steering angle and the vehicle's speed in the next time interval; the formula for predicting the vehicle's speed in the next time interval is:

[0092] v k+1 =v k +k vc (C t -C k )+kv (v des -v k )

[0093] in, Pre-aiming time t la The heading angle of the road after seconds, Pre-aiming time t la The vehicle's heading angle, v, after seconds des For the desired speed, k φ This is the gain for heading control.

[0094] In one specific embodiment, the simulation parameters shown in Table 1 are used. Based on these model parameters, the anthropomorphic lane-changing effect of the vehicle can be obtained.

[0095] Table 1 Simulation Parameter Settings

[0096]

[0097] The vehicle kinematic model proposed in this invention is as follows: Figure 4 As shown, the established two-degree-of-freedom dynamic model of the vehicle is as follows: Figure 5 As shown, the calculation of the parameters of the above potential field model yields the following results: Figure 6 The simulation results of lane changing are shown in the figure.

[0098] A human-like lane-changing method and system for unmanned vehicles based on DRF model and adaptive preview, the system includes: a data acquisition module, a driving risk assessment module, an adaptive preview module, and a human-like decision-making module;

[0099] The data acquisition module is used to acquire the vehicle's environmental information, wherein the perception information includes obstacle information and environmental information, wherein the obstacle information includes surrounding vehicle information; and the environmental information includes road boundary information and surrounding road vegetation location information.

[0100] The driving risk assessment module is used to calculate the current vehicle quantitative risk, that is, to construct a driver risk field model based on the information obtained by the data acquisition module, and to calculate the driver driving risk quantitative value based on the driver risk field.

[0101] The adaptive anti-aiming module calculates the current optimal anti-aiming time using an anti-aiming time evaluation function and updates the vehicle's state for the next moment based on the anti-aiming time.

[0102] The human-like decision-making module makes lane-changing decisions for the vehicle based on the driving risk quantification value, current speed, expected speed, and risk threshold.

[0103] The system implementation method of the present invention is the same as the method implementation method.

[0104] The above-described embodiments further illustrate the purpose, technical solution, and advantages of the present invention. It should be understood that the above-described embodiments are merely preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made to the present invention within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A lane-changing method for unmanned vehicles based on a DRF model and adaptive prediction, characterized in that, include: The vehicle senses its surroundings and obtains sensory information; Acquire driver perception information and driver behavior uncertainties, and construct a driver risk field model; The perceived information is input into the driver risk field model to obtain a quantitative value of driving risk; An adaptive anti-aiming method is established based on the vehicle's kinematics and two-degree-of-freedom dynamics model. The adaptive anti-aiming method is used to set the evaluation function to be optimized, and the optimal anti-aiming time is extracted based on the evaluation function to be optimized. A human-like lane-changing decision model is constructed. The perceived risk threshold is calculated based on historical perception information. The calculated driving risk quantification value is compared with the perceived risk threshold. Based on the comparison results and the optimal aiming time, the human-like lane-changing decision model is used to perform human-like lane changing for the unmanned vehicle.

2. The lane-changing method for unmanned vehicles based on DRF model and adaptive preview as described in claim 1, characterized in that, Vehicles perceive their surroundings by using sensors to sense environmental information, including information about surrounding vehicles and the environment, such as road boundary information and the location of vegetation along the road. A cost map is then constructed based on the perceived information.

3. The lane-changing method for unmanned vehicles based on DRF model and adaptive preview as described in claim 1, characterized in that, The expression for the driver risk field model is: Where z(x,y) represents the driver risk field model, p represents the Gaussian parabola steepness, (x c ,y c R is the center of rotation of the vehicle at the previous moment. car Let s represent the radius of rotation of the vehicle, s represent the arc length of rotation, v represent the instantaneous velocity of the vehicle, and t represent the radius of rotation of the vehicle. la For the driver's forward-looking time, v·t la The forward sight distance is represented by σ, which represents the parabolic width of the Gaussian section.

4. The lane-changing method for unmanned vehicles based on DRF model and adaptive preview as described in claim 3, characterized in that, The expression for the parabolic width of a Gaussian section is: σ(s)=(m+k i |δ|)s+c Where, k i The parameters represent the changes in the inner and outer edges of the field when turning, where i=1 represents the inner edge and i=2 represents the outer edge; m is the slope of the field change when driving straight; and c represents the field width at the vehicle's location.

5. The lane-changing method for unmanned vehicles based on DRF model and adaptive preview as described in claim 1, characterized in that, The process of establishing an adaptive anti-aiming method for unmanned vehicles includes: establishing a vehicle kinematic model and updating its parameters; updating the kinematic model based on the updated parameters; updating the vehicle's motion state at the next moment based on the new kinematic model; establishing a two-degree-of-freedom dynamic model of the vehicle and updating its parameters; and establishing a humanized anti-aiming time evaluation function based on the updated model parameters, where the expression for the humanized anti-aiming time evaluation function is: Where α1 is the driving accuracy style coefficient, and α2 is the driving comfort style coefficient; J E The evaluation function is based on the transverse error. This is an evaluation function based on lateral acceleration.

6. The lane-changing method for unmanned vehicles based on DRF model and adaptive preview as described in claim 5, characterized in that, The formula for updating the parameters of the two-degree-of-freedom dynamics model of a vehicle is as follows: Where u represents the longitudinal velocity of the vehicle; v lat Indicates the lateral speed of the vehicle; I z The z-axis represents the moment of inertia about the z-axis; k1 and k2 represent the lateral stiffness of the front and rear tires, respectively; l f l r Indicates the distance from the front and rear axles to the center of gravity; L is the vehicle wheelbase; i sw The transmission ratio of the vehicle's rotating system; a y This is lateral acceleration.

7. The lane-changing method for unmanned vehicles based on DRF model and adaptive preview as described in claim 1, characterized in that, The process of using a human-like lane-changing decision-making model to perform human-like lane-changing for autonomous vehicles includes: quantifying the driving risk C k With risk threshold C t A comparison is made between the current speed and the desired speed, and a decision tree is constructed based on the comparison results. The autonomous vehicle then performs human-like lane changing based on the decision tree. The decision tree includes: When C k <C t And v k <v des Or C k <C t And v k >v des At the same time, obtain the road heading angle and vehicle heading angle at the optimal preview time; update the steering angle of the current unmanned vehicle based on the road heading angle and vehicle heading angle at the optimal preview time; predict the speed of the unmanned vehicle in the next time period based on the current speed and the expected speed; control the unmanned vehicle to change lanes based on the updated steering angle and the speed of the unmanned vehicle in the next time period; the formula for updating the steering angle of the current unmanned vehicle and predicting the speed of the unmanned vehicle in the next time period is: d k+1 =d k +k φ (f r -f car ) v k+1 =v k +k v (v des -v k ) When C k >C t And v k <v des At that time, the optimal steering angle of the autonomous vehicle is obtained based on the current driving risk and driving speed, and the speed of the autonomous vehicle in the next time period is predicted, where C k+1 <C t Or minimize C k+1 At that time, the optimal steering angle for the autonomous vehicle is determined; the autonomous vehicle is controlled to change lanes based on the optimal steering angle and its speed in the next time period; the formula for predicting the speed of the autonomous vehicle in the next time period is: v k+1 =v k +k v (v des -v k ) C k >C t And v k >v des At that time, the optimal steering angle of the autonomous vehicle is obtained based on the current driving risk and driving speed, and the speed of the autonomous vehicle in the next time period is predicted, where C k+1 When minimized, the optimal steering angle for the autonomous vehicle is obtained; the autonomous vehicle's lane-changing is controlled based on the optimal steering angle and the vehicle's speed in the next time interval; the formula for predicting the vehicle's speed in the next time interval is: v k+1 =v k +k vc (C t -C k )+k v (v des -v k ) in, Pre-aiming time t la The heading angle of the road after seconds, Pre-aiming time t la The vehicle's heading angle, v, after seconds des For the desired speed, k φ This is the gain for heading control.

8. A lane-changing system for unmanned vehicles based on DRF model and adaptive prediction, characterized in that... The system includes: a data acquisition module, a driving risk assessment module, an adaptive anti-aiming module, and a human-like decision-making module; The data acquisition module is used to acquire the vehicle's environmental information, wherein the perception information includes obstacle information and environmental information, wherein the obstacle information includes surrounding vehicle information; and the environmental information includes road boundary information and surrounding road vegetation location information. The driving risk assessment module is used to calculate the current vehicle quantitative risk, that is, to construct a driver risk field model based on the information obtained by the data acquisition module, and to calculate the driver's driving quantitative risk based on the driver risk field. The adaptive anti-aiming module calculates the current optimal anti-aiming time using an anti-aiming time evaluation function and updates the vehicle's state for the next moment based on the anti-aiming time. The human-like decision-making module makes lane-changing decisions for the vehicle based on the driving risk quantification value, current speed, expected speed, and risk threshold.

Images