Driving behavior unified model construction method based on psychological safety space

Abstract

The invention provides a novel driving behavior unified model based on a psychological safety space. A psychological safety space is defined for the first time by deeply discussing a relationship between a scene awareness theory and a personal space theory, and an asymmetric factor is introduced on the basis of the psychological safety space so as to perfect an existing risk field model. Meanwhile, in combination with the risk field theory, the boundary of the psychological safety space is quantitatively analyzed. On the basis, a plurality of theories are integrated, and a comprehensive driving behavior unified model is constructed. The model not only comprises a space trajectory planning unified algorithm, but also covers a motion speed adjustment algorithm, so that more accurate driving behavior prediction is realized. The model aims to provide a solid theoretical basis and practical guidance for traffic prediction and improvement of traffic safety. By applying the model, potential risks in driving behaviors can be effectively identified and analyzed, and scientific support is provided for optimizing traffic management and improving driving safety.

Description

Technical Field

[0001] The present invention relates to the fields of traffic behavior modeling and autonomous driving, and particularly relates to a method for constructing a unified driving behavior model based on a psychological safety space Background Art

[0002] Driving behavior plays an important role in the research of traffic systems and forms the basis of traffic behavior research. Phenomena such as traffic accidents and traffic jams all stem from subtle changes in driving behavior. By deeply studying driving behavior, it helps to understand the behavior patterns and decision-making processes of drivers on the road, and can further reveal the complex interactions and the internal mechanisms of traffic flow that affect the operation of the overall traffic system. It not only has theoretical significance for traffic engineering and traffic management, but also has great practical significance for the optimization and development of the actual traffic system

[0003] To improve the generalization ability of the model and achieve a unified description of driving behaviors in different scenarios, scholars have begun to attempt modeling methods based on field theory. By introducing continuous field variables, the field theory model can describe the behavioral responses of drivers in different scenarios, which can not only handle complex environmental changes but also improve the adaptability and unity of the model in different driving scenarios. Li et al. proposed a dynamic driving risk potential field to characterize the degree of safety risk during the driving process and applied this dynamic driving risk potential field model to car-following and lane-changing models. Ma et al. generated interaction forces based on the potential field to guide vehicle trajectories and avoid collisions. Wang et al. constructed a driving safety field model from the perspective of field theory and defined a driving safety index and a relative driving safety index based on the constructed model to provide different types of warnings for vehicles to avoid collisions. They mainly tend to provide the optimal trajectory, and the model is more suitable for planning the trajectories of autonomous vehicles. Tan et al. quantified the potential danger levels in different regions around moving objects by describing the risk distribution around the vehicle, proposed the risk field theory, and modeled car-following and lane-changing behaviors based on the risk dynamic balance and preview theory. The work of Tan et al. has attracted wide attention. Hua and Wang et al. further established an approach lane model and an intersection passing model based on the risk field theory, demonstrating the application potential of this theory in complex traffic environments. However, the risk field model established by Tan et al. assumes that the risk distribution is symmetric and does not consider the risk differences caused by the speed direction. There are certain limitations in the way of adjusting vehicle motion through symmetric risk distribution and perceiving objective risks. For example, in the car-following scenario, excellent performance is obtained by perceiving the risk behind the leading vehicle to control the motion of the following vehicle. However, for the vehicle that passes through the conflict point first during intersection passing, the acceleration behavior of the yielding vehicle will be delayed due to the slow attenuation of the risk behind. Therefore, based on the definitions of situation awareness theory and personal space theory, this paper proposes a psychological safety space (PSS) model. The driver adjusts operations based on subjective sense of security rather than objective risks, providing a new perspective and method for understanding and simulating drivers' behaviors and decisions.

[0004] The personal space theory was proposed by Robert Sommer in the 1960s to describe the spatial distance required by an individual to maintain their psychological and physiological safety during interactions with others. People need to be able to occupy and control a certain spatial domain to obtain relative subjective sense of security, and the magnitude of people's risk perception is closely related to the sense of security. The situation awareness theory states that people will predict possible future changes and results based on their understanding of the current situation and make decisions.

[0005] This study is based on the risk field theory. Considering the influence of the speed direction on the risk distribution, an asymmetry factor is introduced to achieve an asymmetric description of the risk distribution. We organically combine the situation awareness theory and the personal space theory through the risk distribution, and further define the upper and lower bounds of the psychological safety space based on this distribution. Combining the situation awareness theory, this study calculates the distance between the maximum risk value that the vehicle may encounter within the prediction time period under the current state of the vehicle and the psychological safety boundary, so as to quantify the driver's sense of safety in different situations. On this basis, combined with vehicle path planning, a unified driving behavior model based on the psychological safety space is realized. This model provides a unified safety perception framework for the driver by integrating situation awareness and risk perception, and realizes precise adjustment of vehicle motion control. Summary of the Invention

[0006] In view of the deficiencies of existing vehicle motion planning methods, the present invention proposes a unified driving behavior model based on the psychological safety space. The method specifically includes the following steps:

[0007] Step 1: Define the personal space area that the driver wants to occupy and control to maintain a sense of driving security during the driving process as the psychological safety space. Quantify the risk distribution around the host vehicle based on the risk field theory, and set the boundary of the psychological safety space;

[0008] Step 2: Make a judgment according to the current traffic environment, use the unified algorithm of spatial trajectory planning to determine the parameters of the control points of the Bezier curve, and construct the vehicle driving path based on the Bezier curve to provide a spatial trajectory basis for subsequent risk prediction and speed adjustment;

[0009] Step 3: Predict the basic trajectory of the vehicle during the prediction time period based on the current motion state of the target vehicle, determine the risk of the vehicle within the prediction time through the risk field theory, and obtain the risk borne by the target vehicle at the current moment and the maximum value of the risk borne within the prediction time period;

[0010] Step 4: Synthesize the current state of the host vehicle and the environmental change results obtained by previewing, calculate the distance between the evaluated risk and the boundary of the psychological safety space, and then obtain the behavior decision that the driver needs to take to maintain a sense of driving security, and complete the driving behavior modeling.

[0011] Further, the setting of the boundary of the psychological safety space in Step 1 includes the following steps:

[0012] Step 11: Introduce an asymmetry factor to improve the existing risk field model; among them, the function of the asymmetry factor is as follows:

[0013]

[0014] γ asym，x and γasym，y The asymmetry factors in the longitudinal and lateral directions respectively, η determines the attenuation rate of the risk around the vehicle with the changes in speed and distance. Figure 2 The risk distributions before and after introducing the asymmetry factor are shown. The attenuation rate of the risk in the opposite direction of the speed increases with the increase in speed and distance. The risk field function after introducing the asymmetry factor is as follows:

[0015]

[0016] Step 12: Based on the perceived risk levels of the driver for the surrounding objects in different spaces, the space around the vehicle is divided into a safe area, a dangerous area, and a steady state area, as Figure 3 shown, and the risks corresponding to the boundaries of the steady state area are defined as the upper and lower bounds of the boundary of the psychological safety space, denoted by S upper and S lower respectively.

[0017] Furthermore, the establishment of the space trajectory in Step 2 includes the following steps:

[0018] Step 21: Taking the approach lane of the intersection as the reference, the midpoint of the stop line of the approach lane of the intersection is the origin of the coordinate system, the straight line where the stop line is located is the X-axis of the coordinate system, and the perpendicular bisector of the stop line is the Y-axis of the coordinate system to establish the coordinate system of the approach lane of the intersection, as Figure 4 shown.

[0019] Step 22: Use a B-spline curve to describe the path of the vehicle passing through the intersection. This curve is controlled by control points P o , P d , P 1 and P 2 . Among them,

[0020] P o represents the starting point of the vehicle path, with the coordinates (0, 0); P d represents the ending point of the vehicle path, with the coordinates (w, h); P 1 controls the shape of the path after the vehicle enters the intersection, located on the extension line of the center of the approach lane of the intersection, with the coordinates (0, L 1 ); P2 controls the shape of the path when the vehicle is about to leave the intersection, located on the extension line of the center of the departure lane of the intersection, with the coordinates (w - L 2 sinα, h - L 2 cosα);

[0021] L 1 represents the distance between the control points P o and P 1 ; L 2 represents the distance between the control points P d and P 2The distance between; corresponding to each group of intersection approach and exit lanes, where w and h respectively represent the horizontal and vertical distances between the midpoints of the approach lane exit and the entrance of the exit lane, and α represents the supplementary angle of the angle between the lines where the approach lane and the exit lane are located, that is, the angle between the line where the exit lane is located and the Y-axis; the optimal control parameter L of the vehicle path 1 、L 2 and the calculation formula of the intersection passing path based on the Bezier curve is as follows:

[0022] L 1 =0.407h + 0.523

[0023] h=7.33v y -14.46

[0024] L 2 =0.709(|w| + h)-(7.052 + L 1 )

[0025] B(t)=(1 - t) 3 P o + 3t(1 - t) 2 P 1 + 3t 2 (1 - t)P 2 + t 3 P d

[0026] Step 23, decouple the speed and the lane-changing path, and use the Bezier curve dynamic programming for the lane-changing trajectory based on the starting point, ending point and heading angle of the lane change.

[0027] The nearest lane-changing position can be calculated using the following equation:

[0028] x min =max(x rollover ,x decel )

[0029]

[0030] where v(t + T p ) is the speed of the target vehicle at the end of the lane change, T P is the prediction time of the trajectory planning, y p is the lateral distance at which the lane change ends at time T P ,a rollover is the rollover acceleration threshold, set to 1.4m / s 2 。v(t) is the current speed of the target vehicle, a min is the maximum deceleration of the vehicle, set to -8m / s 2The farthest lane change position constraint mainly considers the maximum acceleration ability of the vehicle to calculate the maximum distance that the vehicle can cover within the prediction time. This maximum distance affected by the acceleration performance can be simply calculated by the following equation:

[0031]

[0032] where a max is the maximum acceleration of the vehicle, which is set to 1.5 m / s 2 .

[0033] Generate a lane change path that conforms to the psychological safety boundary based on the lane change end point. The lane change path of the vehicle from the current position to the lane change end position uses a cubic Bézier curve, and the control point coordinates of this curve are as follows:

[0034] d x = x end

[0035] d y = y p

[0036] P o = (0, 0)

[0037] P d = (d x , d y )

[0038]

[0039] where θ is the heading angle of the target vehicle at the current moment, and d x and d y are the longitudinal and lateral distances of the lane change end position relative to the lane change start position, respectively, determined according to the traversable reach range. Based on the generated lane change curve, a lane change path is initially generated. Assume that the driver expects to reach the lane change end point along the lane change curve with uniform acceleration after the preview time to complete the lane change. Therefore, the acceleration corresponding to this trajectory can be calculated by the following equation:

[0040]

[0041] where S path and y′(x) are the length and slope of the path curve, respectively. After obtaining the acceleration, the corresponding lane change trajectory can be derived, and the risk of this trajectory can be calculated. If the risk of the planned path meets the driver's psychological safety boundary, then this path is the lane change path expected by the driver. If no path that meets the driver's psychological safety boundary is found after traversing all possible lane change paths, then the path with the minimum risk among all lane change paths is selected as the expected lane change path.

[0042] If the vehicle is already in the process of changing lanes, to avoid frequent path planning during lane changes, update the reachable position range of the desired path and extend the planned path along the road direction according to the calculated reachable position range. Select the lane change end point from far to near within the reachable position range. Assume that the target vehicle will travel along the planned path with uniform acceleration and reach the predicted lane change end point after the prediction time. Calculate the risk corresponding to the planned trajectory. If the risk of the planned path meets the driver's psychological safety boundary, then this path is the lane change path desired by the driver. If it does not meet the driver's psychological safety boundary, determine whether the path with the minimum risk is less than the driver's maximum acceptable risk. If this constraint is not met, it means that the existing planned path no longer meets the conditions for the driver to change lanes to the target lane, and the lane change path needs to be replanned.

[0043] Furthermore, risk prediction is performed in step 3, including the following steps:

[0044] Step 31, the position and speed of the vehicle within the preview time period can be calculated by the following formulas:

[0045] x p,i (t 0 , t 0 +t j ) = x i (t 0 ) + v i (t 0 )t j

[0046] v p，i (t 0 , t 0 +t j ) = v i (t 0 )

[0047] At the current moment t 0 , the initial position and speed of the i-th vehicle are x i (t 0 ) and v i (t 0 )t. Within the prediction time T, the position and speed sequences of the predicted vehicle are x p,i (t 0 , t 0 +t j ) and v p,i (t 0 , t 0 +t j ). The time t j is the time series from 0 to the preview time T. The prediction time T is taken as 1.5 s.

[0048] Step 32: Substitute the predicted vehicle position and speed sequences into the risk field function to obtain the perceived risk level of the host vehicle with respect to surrounding vehicles within the prediction time. The calculation method is as follows:

[0049]

[0050] risk p,max (t) is the maximum value of the perceived risk level of the host vehicle within the prediction time at time t. R(x p,i,j , y p,i,j ) is the risk value of the i-th surrounding vehicle at the prediction time t j . n is the number of surrounding vehicles of the host vehicle, and step_t is the step size into which the prediction time T is divided. By taking the maximum value of these individuals within the prediction time period, the predicted risk value within the prediction time can be obtained.

[0051] Furthermore, in Step 4, based on the perceived risk level at the current moment and the maximum value of the perceived risk level of the host vehicle within the prediction time, calculate the distance from this value to the boundary of the psychological safety space to make a decision control. The control method is as follows:

[0052] risk cur (t) = max(R(x p,i,j , y p,i,j )) for i = 0, 1,..., n and j = 0

[0053] risk est (t) = ρ · risk cur (t) + (1 - ρ) · risk p,max (t)

[0054]

[0055] a min ≤ a(t) ≤ a max

[0056] where ρ is the weight that balances the influence degrees of the perceived risk level at the current moment and the maximum value of the perceived risk level of the host vehicle within the prediction time on the decision. risk cur (t) is the perceived risk level at the current moment. S upper and S lower represent the risk values corresponding to the upper and lower bounds of the psychological safety space. Here, recommended values of the model parameters are provided. Select S upper = 0.382, S lower = 0.327, ω 1 = 15.37, ω 2 = 18.10. a(t) is the acceleration at the current time t and needs to satisfy the maximum acceleration and minimum acceleration constraints, a minFor the minimum acceleration, in this paper, a min = -8 m / s 2 is set. If a(t) ≤ a min , then a(t) = a min . Here, a max is the maximum acceleration. In this paper, a max = 1.5 m / s 2 is set. If a(t) ≥ a max , then a(t) = a max .

[0057] So far, the modeling part has been completed. Description of the Drawings

[0058] Figure 1 is the flow method for establishing the model of the present invention;

[0059] Figure 2 is the risk distribution diagram before and after introducing the asymmetry factor;

[0060] Figure 3 is the schematic diagram for dividing the psychological safety boundary;

[0061] Figure 4 is the road coordinate system diagram;

[0062] Figure 5 is the predicted lane - change trajectory diagram;

[0063] Figure 6 is the speed simulated and predicted by each car - following model;

[0064] Figure 7 is the relative spacing simulated and predicted by each car - following model. Detailed Embodiments

[0065] The present invention will be described in detail below in conjunction with the drawings and embodiments.

[0066] This part is based on the improved risk field introducing the asymmetry factor to prove the feasibility of the unified driving behavior model based on the psychological safety space, including the following steps:

[0067] Step 1: Define the personal space area that the driver wants to occupy and control to maintain driving security during the driving process as the psychological safety space. Quantify the risk distribution around the host vehicle based on the risk field theory and set the boundary of the psychological safety space;

[0068] Step 2: Make a judgment according to the current traffic environment, use the unified algorithm for spatial trajectory planning to determine the parameters of the control points of the Bezier curve, construct the vehicle driving path based on the Bezier curve, and provide a spatial trajectory basis for subsequent risk prediction and speed adjustment;

[0069] Step 3: Based on the current motion state of the target vehicle, predict the basic trajectory of the vehicle during the prediction time period. Determine the risk of the vehicle within the prediction time through the risk field theory, and obtain the risk borne by the target vehicle at the current moment and the maximum value of the risk borne during the prediction time period.

[0070] Step 4: Integrate the current state of the host vehicle and the environmental change results obtained from preview, calculate the distance between the evaluated risk and the boundary of the psychological safety space, and then obtain the behavioral decision that the driver needs to take to maintain driving security, thus completing the driving behavior modeling.

[0071] In one embodiment, the setting of the boundary of the psychological safety space in Step 1 includes the following steps:

[0072] Step 11: Introduce an asymmetry factor to improve the existing risk field model. The comparison diagrams before and after are as Figure 2 shown; among them, the function of the asymmetry factor is as follows:

[0073]

[0074] γ asym，x and γ asym，y are the longitudinal and lateral asymmetry factors respectively, η determines the attenuation rate of the risk around the vehicle with the change of speed and distance, Figure 1 is the risk distribution before and after introducing the asymmetry factor. The attenuation rate of the risk in the opposite direction of the speed increases with the increase of speed and distance. The risk field function after introducing the asymmetry factor is as follows:

[0075]

[0076]

[0077] Step 12: Based on the perceived risk levels of the driver for surrounding objects in different spaces, divide the space around the vehicle into a safe area, a dangerous area, and a steady state area, as Figure 3 shown. Define the risk corresponding to the boundary of the steady state area as the upper and lower bounds of the boundary of the psychological safety space, and use S upper and S lower to represent them respectively.

[0078] In one embodiment, the establishment of the spatial trajectory in Step 2 includes the following steps:

[0079] Step 21: Take the approach lane of the intersection as the reference, the midpoint of the stop line of the approach lane of the intersection as the origin of the coordinate system, the straight line where the stop line is located as the X-axis of the coordinate system, and the perpendicular bisector of the stop line as the Y-axis of the coordinate system to establish the coordinate system of the approach lane of the intersection, as Figure 4 shown.

[0080] Step 22, use a Bezier curve to describe the path of the vehicle passing through the intersection. The curve is controlled by control points P o , P d , P 1 and P 2 . Among them,

[0081] P o represents the starting point of the vehicle path, with coordinates (0, 0); P d represents the ending point of the vehicle path, with coordinates (w, h); P 1 controls the shape of the path after the vehicle enters the intersection, located on the extension line of the center of the approach lane of the intersection, with coordinates (0, L 1 ); P 2 controls the shape of the path when the vehicle is about to leave the intersection, located on the extension line of the center of the departure lane of the intersection, with coordinates (w - L 2 sinα, h - L 2 Cosα);

[0082] L 1 represents the distance between control points P o and P 1 ; L 2 represents the distance between control points P d and P 2 . For each group of approach lanes and departure lanes of the intersection, where w and h respectively represent the horizontal and vertical distances between the midpoints of the exit of the approach lane and the entrance of the departure lane, and α represents the supplementary angle of the included angle between the lines where the approach lane and the departure lane are located, that is, the included angle between the line where the departure lane is located and the Y-axis; the optimal control parameters L 1 , L 2 and the calculation formula for the intersection passing path based on the Bezier curve are as follows:

[0083] L 1 = 0.407h + 0.523

[0084] L 2 = 0.709(|w| + h) - (7.052 + L 1 )

[0085] B(t) = (1 - t)3P o + 3t(1 - t) 2 P 1 + 3t 2 (1 - t)P 2 + t 3 P d

[0086] When the vehicle changes lanes, assuming α = 0 and w ≠ 0, that is, the vehicle's "lane change" is similar to going straight within an intersection. Some scholars use Bezier curves to describe the lane-changing path of manually driven vehicles or plan the lane-changing path for autonomous vehicles. Different from the models in the intersection scenario, the starting and ending points of the path in the intersection scenario are known. In the lane-changing scenario, the only thing that can be determined is that the starting point is on the center line of the lane before the lane change, and the ending point is on the center line of the lane after the lane change, that is, w is known, but h is unknown. To solve this problem, the longitudinal speed (along the Y-axis) during the lane change is introduced as a measure of h. And the average longitudinal speed v y and the longitudinal distance h are related as follows:

[0087] h = 7.33v y -14.46

[0088] Step 23, decouple the speed from the lane-changing path, and based on the starting point, ending point, and heading angle of the lane change, use Bezier curves to dynamically plan the lane-changing trajectory, as Figure 5 shown.

[0089] Considering vehicle dynamics, calculate the reachable range of the host vehicle on the target lane. Establish an x-axis parallel to the center line of the current lane and a y-axis perpendicular to the center line of the current lane, with the vehicle center as the origin. Calculate the lateral distance between the current position of the vehicle and the ending point. Due to the limitations of vehicle dynamics, the executable position of the target vehicle when executing the lane-changing strategy is restricted. The nearest lane-changing position constraint is mainly to avoid the vehicle rolling over during the lane change and to consider the maximum deceleration ability of the vehicle. Therefore, the nearest lane-changing position can be calculated using the following equation:

[0090] x min = max(x rollover , x decel )

[0091]

[0092] where v(t + T p ) is the speed of the target vehicle at the end of the lane change, T P is the prediction time for trajectory planning, y p is the lateral distance at the end of the lane change at time T P , a rollover is the rollover acceleration threshold, set to 1.4 m / s 2 . v(t) is the current speed of the target vehicle, a min is the maximum deceleration of the vehicle, set to -8 m / s 2 . The farthest lane-changing position constraint mainly considers the maximum acceleration ability of the vehicle to calculate the maximum distance that the vehicle can cover within the prediction time. This maximum distance affected by the acceleration performance can be simply calculated using the following equation:

[0093]

[0094] Among them, a max is the maximum acceleration of the vehicle, which is set to 1.5 m / s 2 .

[0095] Generate a lane change path based on the lane change end point. This process involves traversing from the lower limit x end to the upper limit x max to determine the lane change end point until a path that meets the psychological safety boundary is generated. The lane change path of the vehicle from the current position to the lane change end position also uses a cubic Bézier curve. The control point coordinates of this cubic Bézier curve are as follows:

[0096] d x = x end

[0097] d y = y p

[0098] P o = (0, 0)

[0099] P d = (d x , d y )

[0100]

[0101] Among them, θ is the heading angle of the target vehicle at the current moment, and d x and d y are the longitudinal and lateral distances of the lane change end position relative to the lane change start position respectively. Determined according to the traversed reachable range. Based on the generated lane change curve, a lane change path is initially generated. Assume that the driver expects to reach the lane change end point along the lane change curve with uniform acceleration after the preview time to complete the lane change. Therefore, the acceleration corresponding to this trajectory can be calculated by the following equation:

[0102]

[0103] Among them, S pathy(x) and y′(x) are the length and slope of the path curve respectively. After obtaining the acceleration, the corresponding lane-changing trajectory can be easily derived, and the risk of this trajectory can be calculated. If the risk of the planned path meets the driver's psychological safety boundary, this path is the lane-changing path desired by the driver. If no path that meets the driver's psychological safety boundary is found after traversing all possible lane-changing paths, the path with the minimum risk among all lane-changing paths is selected as the desired lane-changing path. Referring to the research of Tan et al., the maximum acceptable risk level is set to 0.8. In this case, the minimum risk of the desired lane-changing path still needs to be lower than the driver's maximum acceptable risk. If this constraint is not met, it indicates that the current situation does not meet the conditions for the driver to change lanes to the target lane, and the target lane of the path planning needs to be adjusted and the path needs to be re-planned.

[0104] If the vehicle is already in the process of changing lanes, to avoid frequent path planning during lane change, update the reachable position range of the desired path, and extend the planned path along the road direction according to the calculated reachable position range. Select the lane-changing end point from far to near within the reachable position range. Assume that the target vehicle will travel at a uniform acceleration along the planned path and reach the predicted lane-changing end point after the prediction time. Calculate the risk corresponding to the planned trajectory. If the risk of the planned path meets the driver's psychological safety boundary, this path is the lane-changing path desired by the driver. If it does not meet the driver's psychological safety boundary, determine whether the path with the minimum risk is less than the driver's maximum acceptable risk. If this constraint is not met, it means that the existing planned path no longer meets the conditions for the driver to change lanes to the target lane, and the lane-changing path needs to be re-planned.

[0105] In one embodiment, the risk prediction in step 3 includes the following steps:

[0106] Step 31, the position and speed of the vehicle within the preview time period can be calculated by the following formulas:

[0107] x p,i (t 0 , t 0 + t j ) = x i (t 0 ) + v i (t 0 )t j

[0108] v p,i (t 0 , t 0 + t j ) = v i (t 0 )

[0109] At the current moment t 0Under the condition, the initial position and speed of the i-th vehicle are x i (t 0 ) and v i (t 0 )t. During the prediction time T, the predicted position and speed sequences of the predicted vehicle are x p,i (t 0 , t 0 + t j ) and v p,i (t 0 , t 0 + t j ). The time t j is a time series from 0 to the preview time T. The prediction time T is taken as 1.5 s.

[0110] Step 32: Substitute the predicted vehicle position and speed sequences into the risk field function to obtain the perceived risk level of the ego vehicle to surrounding vehicles within the prediction time. The calculation method is as follows:

[0111]

[0112] riskp,max(t) is the maximum value of the perceived risk level of the ego vehicle within the prediction time at time t, R(xp,i,jjyp,i,j) is the risk value of the i-th surrounding vehicle at the prediction time t, n is the number of surrounding vehicles of the ego vehicle, and step-t is the step size into which the prediction time T is divided. By taking the maximum value of these individuals within the prediction time period, the predicted risk value within the prediction time can be obtained.

[0113] In one embodiment, in Step 4, based on the perceived risk level at the current moment and the maximum value of the perceived risk level of the ego vehicle within the prediction time, the distance from this value to the boundary of the psychological safety space is calculated to make a decision control. The control method is as follows:

[0114] risk cur (t) = max(R(x p,i,j , y p,i,j ))), i = 0, 1,..., n, j = 0

[0115] risk est (t) = ρ · risk cur (t) + (1 - ρ) · risk p,max (t)

[0116]

[0117] a min ≤ a(t) ≤ a max

[0118] Among them, ρ is the weight that balances the influence degree of the maximum value of the risk level perceived at the current moment and the risk level perceived by the host vehicle within the prediction time on the decision-making, risk cur (t) is the risk level perceived at the current moment, S upper and S lower represent the risk values corresponding to the upper and lower bounds of the psychological safety space. Here, recommended values of the model parameters are provided. Select S upper = 0.382, S lower = 0.327, ω 1 = 15.37, ω 2 = 18.10. a(t) is the acceleration at the current moment t, which needs to satisfy the maximum acceleration and minimum acceleration constraints. a min is the minimum acceleration. In this paper, a min = -8m / s 2 , if a(t) ≤ a min , then a(t) = a min , a max is the maximum acceleration. In this paper, a max = 1.5m / s 2 , if a(t) ≥ a max , then a(t) = a max .

[0119] In one embodiment, in the car-following scenario, we selected 9 cases from the Next Generation Simulation (NGSIM) for model verification, and compared them with the DR model, IDM model, and DSM model to verify the model based on the psychological safety space. These cases are the observation results of the US Highway 101 (Hollywood Freeway) in Los Angeles, and the measured leading vehicle trajectory is used as the initial data. The initial position of the following vehicle is set as the measured value. From the simulation results, the PSS model proposed in this paper has similar effects to the DR model and DSM model, and is significantly better than the IDM model. Figure 6 and Figure 7 show that the PSS model, DR model, and DSM produce similar results in terms of the speed and relative spacing of the following vehicle, and can all simulate the vehicle stop and start processes. The speed generated by the IDM is not much different from the results of the other three models, but over time, the error of the relative spacing gradually increases. The invention not only helps to understand and predict driving behavior, but also can be applied to fields such as vehicle control algorithm optimization and road safety assessment.

[0120] The above embodiments are only used to illustrate the present invention. Among them, each step of the method and so on can be changed. Any equivalent transformation and improvement made on the basis of the technical solution of the present invention should not be excluded from the protection scope of the present invention.

Claims

1. A unified model of driving behavior based on psychological safety space, characterized in that: The following steps are involved: Step 1: Define the personal space that the driver wants to occupy and control to maintain a sense of driving safety during driving as the psychological safety space, quantify the risk distribution around the vehicle based on the risk field theory, and set the boundary of the psychological safety space; Step 2: Make a judgment based on the current traffic environment, use the unified algorithm of spatial trajectory planning to determine the parameters of the Bezier curve control points, and construct the vehicle driving path based on the Bezier curve to provide a spatial trajectory basis for subsequent risk prediction and speed adjustment; Step 3: predict the basic trajectory of the vehicle within the prediction time period based on the current motion state of the target vehicle, determine the risk of the vehicle within the prediction time period through risk field theory, and obtain the risk borne by the target vehicle at the current moment and the maximum risk borne within the prediction time period; Step 4: Based on the current state of the vehicle and the environmental change results obtained by previewing, the distance between the risk and the boundary of the psychological safety space is calculated and assessed, and then the behavioral decision that the driver should take to maintain a sense of driving safety is obtained, completing the driving behavior modeling.

2. The method according to claim 1, characterized in that The psychological safety space boundary setting in step 1 includes the following steps: Step 11, introduce asymmetric factors to improve the existing risk field model; Step 12: Based on the driver's perceived risk level of surrounding objects in different spaces, the space around the vehicle is divided into a safe zone, a dangerous zone, and a steady-state zone. The risk corresponding to the boundary of the steady-state zone is defined as the upper and lower boundaries of the psychological safety space boundary, respectively expressed as S upper and S lower express.

3. The method according to claim 1, characterized in that The spatial trajectory establishment in step 2 includes the following steps: Step 21, establish the intersection entrance road coordinate system with the intersection entrance road as the reference, the midpoint of the intersection entrance road stop line as the origin of the coordinate system, the straight line where the stop line is located as the coordinate system X-axis, and the midpoint perpendicular line of the stop line as the coordinate system Y-axis. Step 22, use a Bezier curve to describe the path of the vehicle through the intersection. The curve is composed of control points P o , P d , P1 and P2 control, where P o Indicates the starting point of the vehicle path, with coordinates (0,0); P d Indicates the end point of the vehicle path, with coordinates (w, h); P1 controls the shape of the vehicle's path after entering the intersection, located on the extension line of the center of the intersection entrance road, with coordinates (0, L1); P2 controls the shape of the vehicle's path when it is about to leave the intersection, located on the extension line of the center of the intersection exit road, with coordinates (w-L2sinα,h-L2cosα); L1 represents the control point P o The distance between P1 and L2 represents the control point P d The distance between P1 and P2; corresponding to each group of intersection entrance and exit roads, where w and h represent the lateral and longitudinal distances between the midpoints of the entrance and exit roads respectively, and α represents the complementary angle of the straight line where the entrance and exit roads are located, that is, the angle between the straight line where the exit road is located and the Y axis; the optimal control parameters L1 and L2 of the vehicle path and the calculation formula of the intersection traffic path based on the Bezier curve are as follows: L1=0.407h+0.523 h=7.33v y -14.46 L2=0.709(|w|+h)-(7.052+L1) B(t)=(1-t) 3 P o +3t(1-t) 2 P1+3t 2 (1-t)P2+t 3 P d Step 23, decoupling the speed from the lane change path, based on the lane change start point, end point and heading angle, the nearest lane change position can be calculated using the following equation: x min =max(x rolllover ,x decel ) The furthest lane change position can be calculated using the following equation: The lane change path of the vehicle from the current position to the lane change end position uses a cubic Bezier curve, and the coordinates of the control points of the curve are as follows: d x =x end d y =y p P o ＝(0,0) P d =(d x ,d y ) 4. The method according to claim 1, characterized in that: Risk prediction is performed in step 3, including the following steps: Step 31, the position and speed of the vehicle during the preview period can be calculated by the following formula: x p,i (t0,t0+t j )=x i (t0)+v i (t0)t j v p,i (t0,t0+t j )=v i (t0) At the current time t0, the initial position and speed of the i-th vehicle are x i (t0) and v i (t0)t, within the prediction time T, the predicted vehicle position and speed sequences are x p,i (t0,t0+t j ) and v p,i (t0,t0+t j ), time t j is the time series from 0 to preview time T, Step 32, substitute the predicted vehicle position and speed sequence into the risk field function to obtain the perceived risk level of the vehicle to the surrounding vehicles within the prediction time, which is calculated as follows: risk p,max (t) is the maximum value of the risk level perceived by the vehicle within the prediction time at time t, R(x p,i,j ,y p,i,j ) is the surrounding i-th vehicle at the predicted time t j The risk value of the vehicle, n is the number of vehicles around the vehicle, step_t is the step size of the prediction time T, and by taking the maximum value of these individuals within the prediction time period, the predicted risk value within the prediction time can be obtained.

5. The method according to claim 1, characterized in that In step 4, based on the risk level perceived at the current moment and the maximum risk level perceived by the vehicle within the prediction time, the distance between the value and the boundary of the psychological safety space is calculated to make a decision and control. The control method is as follows: risk cur (t)=max(R(x p,i,j ,y p,i,j )),i=0,1,…n,j=0 risk est (t)=ρ·risk cur (t)+(1-ρ)·risk p,max (r) a min ≤a(t)≤a max Among them, ρ is the weight of balancing the risk level perceived at the current moment and the maximum risk level perceived by the vehicle within the prediction time on the decision-making degree, risk cur (t) is the perceived risk level at the current moment, S upper and S lower represents the risk value corresponding to the upper and lower bounds of the psychological safety space, a(t) is the acceleration at the current time t, which must meet the maximum acceleration and minimum acceleration constraints, a min is the minimum acceleration, and this paper sets a min =-8m / s 2 , if a(t)≤a min , then a(t)=a min , a max is the maximum acceleration, and this paper sets a max =1.5m / s 2 , if a(t)≥a max , then a(t)=a max .

Images