Collision risk assessment method and device based on lane and surrounding vehicle motion prediction

Abstract

The invention discloses a collision risk assessment method and device based on lane and surrounding vehicle motion prediction, and belongs to the field of automatic driving. The method comprises the following steps: calculating position information of a vehicle and surrounding vehicles in a road coordinate system; based on a quintic polynomial, generating all local candidate paths of the vehicle; based on the lane state transition matrix and a lane transition model, calculating dynamic real-time probability distribution of each surrounding vehicle driving to each lane; based on the position information of the surrounding vehicles and the dynamic real-time probability distribution, using a cubic polynomial to generate future driving tracks of the surrounding vehicles on different lanes; and based on the dynamic real-time probability distribution and the time required for collision between the local candidate path of the own vehicle and the future driving track of each surrounding vehicle on different lanes, calculating a total collision risk value when the own vehicle runs along each local candidate path. According to the method, probability motion prediction and a lane structure are combined, the motion uncertainty in a real driving environment is effectively reflected, and the risk assessment accuracy is improved.

Description

Technical Field

[0001] The present invention relates to the field of autonomous driving technology, and in particular to a collision risk assessment method and device based on lane and surrounding vehicle motion prediction. Background Art

[0002] With the promotion of the application of autonomous driving technology, the environment and traffic conditions faced by vehicles during driving have become increasingly complex. In order to ensure the safety of vehicles, especially in changeable and unpredictable driving scenarios and complex road environments, the autonomous driving system must be able to effectively predict potential collision risks.

[0003] Existing collision risk assessment methods can be divided into two categories: deterministic methods and uncertain methods. Deterministic methods calculate indicators such as time to collision (TTC) or time to brake (TTB) through simplified future motion prediction models (such as constant yaw rate and acceleration models) to quantify collision risks. However, such methods cannot reflect the uncertainty of future motion and may completely ignore potential risks when the vehicle behavior is different from the predicted trajectory. Uncertain methods, also known as probabilistic methods, predict the future behavior of the vehicle by utilizing its historical data, current state, and environmental information, and calculate the collision probability through probability distribution. However, the modeling of motion uncertainty in these methods fails to fully consider road geometry and lane characteristics.

[0004] The patent with publication number CN111731283A and invention name "Method, device and electronic device for identifying collision risk of autonomous vehicles" mainly provides a method for judging whether there is a collision risk between vehicles by comparing multiple predicted trajectories of the vehicle with the trajectories of surrounding vehicles. This method is a deterministic method. Since the invention mainly relies on the generation and comparison of trajectories, it does not fully consider the uncertainty of vehicle behavior and the probability of lane changes, and lacks quantitative assessment of collision risks. Therefore, when facing high-dynamic scenarios such as high-speed driving on curves, it is impossible to accurately and intuitively assess the collision risk of the vehicle.

[0005] The patent with publication number CN114662967B and invention name "Unmanned Driving Collision Risk Assessment Method and System Based on Dynamic Bayesian Network" mainly provides a method for evaluating the multi-period collision risk probability of autonomous driving vehicles through a dynamic Bayesian network model combined with network-level risk impact. This method is an uncertainty method that uses parameters such as vehicle motion risk, surrounding vehicle risk, and network collision risk to comprehensively calculate the vehicle's collision probability. However, this method lacks consideration of the actual road structure and does not use information such as the number of lanes and road curvature as independent modeling parameters, resulting in certain limitations in the actual application of this method and the inability to meet the requirements for risk assessment accuracy under complex road structures.

[0006] Therefore, the existing assessment methods do not simultaneously consider the lane change characteristics and the uncertainty of vehicle behavior, resulting in insufficient accuracy in collision risk assessment in complex dynamic scenarios such as vehicle lane changes. Therefore, there is an urgent need to provide a collision risk assessment method based on the prediction of lane and surrounding vehicle motion. Summary of the invention

[0007] In order to solve the problem that the existing assessment methods fail to simultaneously consider the lane change characteristics and the uncertainty of vehicle behavior, resulting in insufficient accuracy in collision risk assessment in complex dynamic scenarios such as vehicle lane changes, an embodiment of the present invention provides a collision risk assessment method and device based on lane and surrounding vehicle motion prediction.

[0008] On the one hand, a collision risk assessment method based on lane and surrounding vehicle motion prediction is provided, the method comprising:

[0009] Based on the Cartesian coordinates of the vehicle and surrounding vehicles, calculate the position information of each vehicle in the road coordinate system; wherein the horizontal and vertical coordinates of the road coordinate system represent the arc length along the road curve and the lateral offset between the vehicle and the center line of the road respectively;

[0010] Based on the quintic polynomial and the current position information of the ego vehicle, all local candidate paths of the ego vehicle are generated;

[0011] The lateral speed of the vehicle is introduced to determine the lane state transfer matrix, and based on the lane state transfer matrix and the lane switching model, the dynamic real-time probability distribution of each surrounding vehicle driving into each lane is calculated;

[0012] Based on the position information of the surrounding vehicles and the dynamic real-time probability distribution of each surrounding vehicle driving into each lane, a cubic polynomial is used to generate future driving trajectories of the surrounding vehicles in different lanes;

[0013] Based on the dynamic real-time probability distribution and the time required for collision between the local candidate path of the vehicle and the future driving trajectory of each surrounding vehicle on different lanes, the total collision risk value of the vehicle when traveling along each local candidate path is calculated.

[0014] On the other hand, a collision risk assessment device based on lane and surrounding vehicle motion prediction is provided, which is used to implement the steps described in any method embodiment of the specification, and the device includes:

[0015] A conversion unit, used to calculate the position information of each vehicle in a road coordinate system based on the Cartesian coordinates of the vehicle and surrounding vehicles; wherein the horizontal and vertical coordinates of the road coordinate system represent the arc length along the road curve and the lateral offset between the vehicle and the center line of the road respectively;

[0016] A candidate path generation unit for the ego vehicle, for generating all local candidate paths for the ego vehicle based on the quintic polynomial and the current position information of the ego vehicle;

[0017] A calculation unit, used to introduce the lateral speed of the vehicle to determine the lane state transfer matrix, and calculate the dynamic real-time probability distribution of each surrounding vehicle driving into each lane based on the lane state transfer matrix and the lane switching model;

[0018] A surrounding vehicle trajectory prediction unit, for generating future driving trajectories of surrounding vehicles in different lanes using a cubic polynomial based on the position information of surrounding vehicles and the dynamic real-time probability distribution of each surrounding vehicle driving toward each lane;

[0019] An evaluation unit is used to calculate a total collision risk value when the vehicle is traveling along each local candidate path based on the dynamic real-time probability distribution and the time required for collision between the local candidate path of the vehicle and the future driving trajectory of each surrounding vehicle on different lanes.

[0020] On the other hand, a computer device is provided, which includes a memory and a processor, wherein the memory is used to store a computer program, and the processor is used to execute the computer program stored in the memory to implement the steps of the above-mentioned method.

[0021] On the other hand, a computer-readable storage medium is provided, wherein a computer program is stored in the storage medium, and when the computer program is executed by a processor, the steps of the above-mentioned method are implemented.

[0022] On the other hand, a computer program product is provided, comprising a computer program, wherein the computer program implements the steps of the above method when executed by a processor.

[0023] The technical solution provided by the present invention can at least bring the following beneficial effects:

[0024] The uncertainty of vehicle motion is modeled based on lane changes, and constraints such as road geometry, number of lanes and road width are introduced. The dynamic real-time probability distribution of each surrounding vehicle moving to each lane is calculated to further determine the future driving trajectory of each surrounding vehicle in different lanes, which can more effectively reflect the motion uncertainty in the real driving environment. Finally, based on the indicator of time required for collision (TTC), the intersection time of different vehicle trajectories is quantified as a collision risk value to improve the accuracy of vehicle collision risk assessment in complex driving scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0026] Figure 1 It is a flow chart of a collision risk assessment method based on lane and surrounding vehicle motion prediction provided by an embodiment of the present invention;

[0027] Figure 2 It is a structural diagram of a collision risk assessment device based on lane and surrounding vehicle motion prediction provided by an embodiment of the present invention;

[0028] Figure 3 It is a hardware architecture diagram of a computer device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0029] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0030] The specific implementation of the above concept is described below.

[0031] Please refer to Figure 1 , an embodiment of the present invention provides a collision risk assessment method based on lane and surrounding vehicle motion prediction, the method comprising:

[0032] Step 100: Calculate the position information of each vehicle in the road coordinate system based on the Cartesian coordinates of the vehicle and surrounding vehicles; wherein the horizontal and vertical coordinates of the road coordinate system represent the arc length along the road curve and the lateral offset between the vehicle and the center line of the road respectively;

[0033] Step 102: Generate all local candidate paths of the vehicle based on the quintic polynomial and the current position information of the vehicle;

[0034] Step 104: Introduce the lateral speed of the vehicle to determine the lane state transfer matrix, and calculate the dynamic real-time probability distribution of each surrounding vehicle driving into each lane based on the lane state transfer matrix and the lane switching model;

[0035] Step 106: Based on the position information of surrounding vehicles and the dynamic real-time probability distribution of each surrounding vehicle driving into each lane, a cubic polynomial is used to generate future driving trajectories of surrounding vehicles in different lanes;

[0036] Step 108: Based on the dynamic real-time probability distribution and the time required for collision between the local candidate path of the ego vehicle and the future driving trajectory of each surrounding vehicle in different lanes, calculate the total collision risk value when the ego vehicle travels along each local candidate path.

[0037] In an embodiment of the present invention, uncertainty modeling of vehicle motion is performed based on lane switching, and constraints such as road geometry, number of lanes and road width are introduced to calculate the dynamic real-time probability distribution of each surrounding vehicle moving to each lane, so as to further determine the future driving trajectory of each surrounding vehicle in different lanes, which can more effectively reflect the motion uncertainty in a real driving environment; finally, based on the indicator of time required for collision (TTC), the intersection time of different vehicle trajectories is quantified as a collision risk value, so as to improve the accuracy of vehicle collision risk assessment in complex driving scenarios.

[0038] Described below Figure 1 How the various steps are performed.

[0039] For step 100:

[0040] In this embodiment, the geometric information of the road curve is first collected by the sensor, the points on the curve are recorded in the form of Cartesian coordinate system (x, y), and all the points on the curve are fitted with a polynomial, and the expression of the road curve is obtained as follows:

[0041] y=c 2 x 2 +c 1 x+c 0

[0042] In the formula, c 2 、c 1 and c 0 are the polynomial coefficients.

[0043] In order to effectively process the motion state of a vehicle traveling on a curved road, this embodiment converts the Cartesian coordinate system (x, y) into a road coordinate system (s, q) to represent the vehicle's position information. The horizontal axis s in the road coordinate system represents the arc length of the vehicle moving along the road curve, that is, the distance traveled along the center line of the road, and the vertical axis q represents the lateral offset between the vehicle and the center line of the road.

[0044] The calculation method of the road coordinate system (s,q) is:

[0045]

[0046] Where (x, y) is the current Cartesian coordinate of the vehicle, (x * ,y * ) is the point on the road centerline closest to the vehicle.

[0047] The geometry of a curved road is the curvature K of the centerline of the road. c To characterize, the calculation formula is as follows:

[0048]

[0049] Therefore, the road coordinate system can process the lateral and tangential motion of the vehicle separately, thereby simplifying the difficulty of modeling and predicting vehicle motion on curved roads. Compared with the traditional Cartesian coordinate system, the road coordinate system can more intuitively reflect the relative position between the vehicle and the road, facilitate real-time calculation and control, and therefore has greater application value in the planning and decision-making of autonomous driving systems.

[0050] Regarding step 102:

[0051] In some implementations, step 102 may include:

[0052] Determine several final arc lengths of the local candidate paths of the ego vehicle based on the tangential velocity of the ego vehicle and several preset tangential accelerations;

[0053] Based on the current position information of the ego vehicle and the final arc length, a quintic polynomial is used to calculate several local candidate paths that are used to characterize the variation of the lateral offset of the ego vehicle with the arc length.

[0054] In this embodiment, assuming that the vehicle maintains a constant acceleration during the entire motion process, the final arc length can be calculated by the following formula:

[0055]

[0056] Among them, s f is the final arc length of a local candidate path, v 0 is the tangential velocity, and a is a preset tangential acceleration.

[0057] Next, a quintic polynomial is used to represent the variation of the lateral offset q of the local candidate path with the arc length s, which is used as the basis for generating the ego vehicle curve trajectory to achieve accurate control of the ego vehicle's motion trajectory on the curved road. The formula is as follows:

[0058]

[0059] Among them, β i are the coefficients of the polynomial, used to generate different local candidate paths, q i and q fThey represent the initial and final lateral offsets from the road centerline relative to the arc length s, and are the lateral velocity and lateral acceleration in the road coordinate system, s f is the final arc length of the local candidate path.

[0060] In this implementation, the local candidate path of the vehicle is generated based on a quintic polynomial to ensure the smoothness and dynamic continuity of the generated path. The quintic polynomial can provide sufficient degrees of freedom to ensure that the velocity and acceleration at the end of the path drop smoothly to zero while satisfying the initial conditions (position, velocity, acceleration) of the vehicle, and better control the acceleration change of the vehicle on the path.

[0061] In addition, in order to provide a more intuitive collision risk assessment result, the present invention projects all generated local candidate paths onto a two-dimensional collision risk trajectory plane S T The collision risk trajectory plane S T The coordinate axes of are the tangential acceleration of the vehicle and the final lateral offset of the local candidate path. Each local candidate path of the vehicle can be projected onto the collision risk trajectory plane as a single point. For a given tangential acceleration, each trajectory has a unique final lateral offset. Its expression is as follows:

[0062] S T =(a i ,q f,j )for i∈[1,…,N acc ],j∈[1,…,N path ]

[0063] Among them, a i is the tangential acceleration, q f,j is the final lateral offset of the local candidate path; N acc is the amount of available tangential acceleration, N path is the number of local candidate paths.

[0064] Regarding step 104:

[0065] In some implementations, step 104 may include:

[0066] S1, determine the initial probability value of switching from lane i to lane j when the lateral speed is 0, to define the initial probability matrix;

[0067] S2, defines the mean and standard deviation of the Gaussian distribution of the lateral speed in the single-lane lane change and double-lane lane change cases, respectively, so as to determine the modified probability matrix using the Gaussian cumulative distribution function;

[0068] S3, add the initial probability matrix and the modified probability matrix to obtain the state transition probability value of the surrounding vehicles switching from lane i to lane j, and integrate the lane state transition matrix;

[0069] S4, based on the lane state transfer matrix and using the lane change model, calculates the initial dynamic real-time probability distribution of each surrounding vehicle driving into each lane.

[0070] In this embodiment, a road coordinate system related to road lanes is used to replace the traditional Cartesian coordinate system, so as to better handle the road geometry and lane constraints in actual driving scenarios.

[0071] First, the present invention introduces lateral speed into the lane state transfer matrix to quickly detect the vehicle's lane change intention. The vehicle's lateral speed reflects its lane change tendency. When the vehicle's lateral speed is high, the system will give a higher lane state change probability, otherwise it will maintain a low lane state change probability.

[0072] The lane state transfer matrix based on lateral velocity consists of two parts: the initial probability matrix and the modified probability matrix.

[0073] In step S1, the initial probability is given by the initial probability matrix Integration, the value of the element (i, j) in the matrix represents the lateral velocity The initial probability value of switching from lane i to lane j when . Based on prior information and experience, this embodiment defines the initial probability matrix as follows:

[0074]

[0075] In step S2, based on prior information and experience, the lateral speed is calculated for the single lane change and double lane change situations. The mean and standard deviation of a Gaussian distribution are defined as:

[0076]

[0077] Where i,j≤N l is the lane number counting from the left side of the road to the right side; sgn(·) is the sign function, ρ ij is the Gaussian distribution mean of the lateral velocity of the vehicle when switching from lane i to lane j, σ ij is the standard deviation of the Gaussian distribution, representing the uncertainty of the lateral velocity.

[0078] Correction probability Calculated by Gaussian cumulative distribution function:

[0079]

[0080] in, represents the lateral velocity of each surrounding vehicle, ρ ij is the Gaussian distribution mean of the lateral velocity of the vehicle when switching from lane i to lane j, σ ij is the standard deviation of the Gaussian distribution, which represents the uncertainty of the lateral velocity.

[0081] In step S3, the calculated corrected probability value is added to the initial probability value to obtain the lane state transition probability value based on the lateral speed, and then the lane state transition matrix is ​​integrated. The expression is as follows:

[0082]

[0083] in, is the unnormalized transition probability of switching from lane i to lane j.

[0084] In step S4, the lane change model is introduced to calculate the probability of each surrounding vehicle driving into a different target lane. The formula is as follows:

[0085]

[0086]

[0087] in, represents the probability of switching to lane j at time k-1, is the lane state transfer matrix, is the probability that the surrounding vehicles are located in the i-th lane at time k-1. When k = 0, represents the probability of the vehicle in its initial state, assuming that the probability of each lane-changing behavior is the same at the beginning, that is, uniformly distributed; is the normalized initial dynamic real-time probability of switching from lane i to lane j.

[0088] In some implementations, in order to capture the uncertainty in vehicle motion and improve the prediction accuracy of vehicle dynamic behavior in complex driving scenarios, after step S4, the following may also be included:

[0089] S5, performing weighted summation of the actual lateral offset of each surrounding vehicle according to the initial dynamic real-time probability of the corresponding lane to obtain an estimated mixed lateral offset of the vehicle relative to each lane;

[0090] S6, calculating an outer product by multiplying the residual of the actual lateral offset of each surrounding vehicle and the estimated mixed lateral offset by its transpose, and performing a weighted summation of the outer product and the initial dynamic real-time probability distribution to combine the lateral offset uncertainties of multiple lanes into a prediction covariance matrix;

[0091] S7, determining an observation residual and an observation covariance matrix of the ego vehicle sensor based on the estimated mixed lateral offset and the predicted covariance matrix;

[0092] S8, based on the observation residual and the observation covariance matrix, calculates the likelihood function used to quantify the degree of match between the predicted value and the observed value, and uses the likelihood function to combine the historical observation value and the current observation value to update the initial dynamic real-time probability distribution of each surrounding vehicle driving into each lane to obtain the final dynamic real-time probability distribution.

[0093] In step S5, the most likely lateral position of the vehicle under uncertainty is represented by an estimated hybrid lateral offset, which is calculated as follows:

[0094]

[0095] in, The actual lateral offset of the surrounding vehicles in each lane According to the initial dynamic real-time probability of the corresponding lane change Perform weighted summation to obtain the estimated mixed lateral offset of the vehicle relative to the jth lane.

[0096] In step S6, the uncertainty of the lateral position of the surrounding vehicles is calculated by the predicted covariance of the lateral offset of the vehicle in the jth lane The outer product is calculated by multiplying the residual of the actual lateral offset of each surrounding vehicle and the estimated mixed lateral offset by its transpose, and the outer product and the initial dynamic real-time probability distribution are A weighted sum is performed to provide uncertainty information about the lateral position of surrounding vehicles at future times. The prediction covariance calculation formula is as follows:

[0097]

[0098] in, represents the difference between the actual lateral offset of the surrounding vehicles and the estimated mixed lateral offset, i.e., the residual; W is the width of a single lane; Q j is the noise covariance matrix in the system, which obeys a zero-mean normal distribution with a standard deviation of (W / 4) to represent the basic noise of the lateral motion of the surrounding vehicles.

[0099] In step S7, the observation residual and observation covariance matrix of the vehicle sensor are calculated by the following formula:

[0100]

[0101] in, is the observed residual, z k is the observed value, is the estimated mixed lateral offset for prediction; is the observation covariance, To predict the covariance, R is the measured noise covariance matrix, which obeys the zero-mean normal distribution and reflects the random noise and error generated by the sensor during measurement.

[0102] In step S8, by calculating the likelihood function Quantify the degree of match between the predicted value and the actual observed value. The smaller the residual, the larger the likelihood value, indicating that the predicted value is closer to the observed value and the model is more accurate. The calculation formula is as follows:

[0103]

[0104] In step S9, the historical observations and the current observations are combined based on the likelihood function to update the initial dynamic real-time probability distribution of each surrounding vehicle driving toward each lane, and the final dynamic real-time probability distribution is obtained, which is calculated as follows:

[0105]

[0106] Therefore, this embodiment fully considers the uncertainty of motion and updates the model probability based on the observation residuals and covariance of the ego-vehicle sensor, which can further improve the accuracy of the prediction results.

[0107] Regarding step 106:

[0108] In some implementations, step 106 may include:

[0109]

[0110] Where q(s) is the function of the lateral offset of the surrounding vehicles relative to the center line of the lane as the arc length s changes, and a i are the coefficients of the polynomial, s f is the target arc length of the surrounding vehicles traveling along the lane, q f is the distance between the centerline of the target lane and the centerline of the initial lane, q i is the lateral offset at the initial moment, Ψ is the heading angle of the surrounding vehicles relative to the tangent of the road centerline, tanΨ represents the lateral velocity of the surrounding vehicles at the initial moment, dq(s f ) / ds represents the lateral speed of the vehicle at the last moment. A value of 0 means that the vehicle is traveling completely along the center line of the target lane and will not deviate from it.

[0111] In this embodiment, a cubic polynomial is used to avoid overshoot and oscillation caused by a quintic polynomial. Therefore, q in the cubic polynomial is f is the distance between the center line of the target lane and the center line of the initial lane, that is, there are as many q as there are lanes f value.

[0112] Regarding step 108:

[0113] In some implementations, step 108 may include:

[0114] B1, based on the current speed of the vehicle, calculate the minimum safe distance between vehicles;

[0115] B2, for each local candidate path of the vehicle, execute:

[0116] The two vehicles are considered to collide when they are at the minimum safe distance, and the time required for the collision between the local candidate path of the vehicle and the future driving trajectory of each surrounding vehicle in different lanes is calculated to determine the collision risk probability;

[0117] B3, based on the collision risk probability and the dynamic real-time probability distribution of each surrounding vehicle driving into each lane, calculate the collision risk value of the local candidate path and each surrounding vehicle, so as to sum up and calculate the total collision risk value when the vehicle travels along each local candidate path.

[0118] In step B1, the minimum safe distance between the vehicle and the preceding vehicle is defined as follows:

[0119] S safe =s 0 +v·T head

[0120] Among them, s 0 is the minimum safe distance, which is usually determined by the braking distance and other parameters of the vehicle; v is the current speed of the vehicle; T head The head distance that the vehicle needs to maintain for safety.

[0121] Secondly, the local candidate path of the ego vehicle and the predicted trajectory of the other vehicle are represented by two trajectory curves, and the body of the ego vehicle and the other vehicle are represented by two rectangles corresponding to the specific vehicle size. The rectangle represented by the ego vehicle needs to be extended forward by the safe distance S between the vehicles. safe Therefore, by judging whether the two rectangles overlap while moving along the trajectory curve, it is possible to determine whether the self-vehicle and the other vehicle have a collision possibility, and then calculate the time required for collision (TTC) between the two as a specific indicator for quantifying the collision risk.

[0122] In step B2, the time from the current position of the two vehicles to the intersection minus the minimum safe distance is taken as the time required for the collision between the two vehicles t c , in order to determine the collision risk probability

[0123]

[0124] Among them, t cis the time required for collision; α is the adjustment coefficient for calculating collision risk, with a value range of (0,1), which controls the speed at which risk changes over time. A larger value indicates a more aggressive assessment of collision risk; The exponential function represents the collision risk as a function of the collision time t c exponential decay of represents the geometric shapes of the ego vehicle and surrounding vehicles i at t c Whether it overlaps. E is the local candidate path of the vehicle, T Si is a future driving trajectory of the surrounding i-th vehicle.

[0125] In step B3, the collision risk between the local candidate path of the vehicle and the surrounding i-th vehicle is calculated based on the probability distribution of the surrounding vehicles changing lanes to different lanes. The calculation formula is as follows:

[0126]

[0127] Where j is the lane number; N l is the total number of lanes on the road; It represents the dynamic real-time probability of the surrounding i-th vehicle being in the j-th lane calculated in step 104.

[0128] Thus, in a multi-lane and multi-surrounding vehicle scenario, assuming that surrounding vehicles move independently while following their predicted trajectories, the local candidate path T of the ego vehicle can be calculated. E The total collision risk value P(C|T E ):

[0129]

[0130] Among them, N S is the total number of surrounding vehicles.

[0131] In some embodiments, after calculating the overall collision risk value when the vehicle travels along different local candidate paths, the risk value can be mapped to a two-dimensional collision risk trajectory plane. For each point on the collision risk trajectory plane, it represents a local candidate path of the vehicle, and the horizontal and vertical coordinates corresponding to the point are the tangential acceleration of the vehicle and the final lateral offset of the local candidate path, respectively, and the color depth of each point corresponds to the size of the collision risk value. Thus, the collision risk values ​​of all local candidate paths of the vehicle are mapped to the collision risk trajectory plane in the form of points, together forming a collision risk map, which intuitively displays the collision risk levels and changing trends of different local candidate paths of the vehicle, and can intuitively reflect the effects of different combinations of the two variables of tangential acceleration and the final lateral offset of the local candidate path on the collision risk.

[0132] Please refer to Figure 2 The embodiment of the present invention provides a collision risk assessment device based on lane and surrounding vehicle motion prediction, which is used to implement the steps of any method embodiment in the specification, and the device includes:

[0133] The conversion unit 201 is used to calculate the position information of each vehicle in the road coordinate system based on the Cartesian coordinates of the vehicle and the surrounding vehicles; wherein the horizontal and vertical coordinates of the road coordinate system represent the arc length along the road curve and the lateral offset between the vehicle and the center line of the road respectively;

[0134] The ego vehicle candidate path generation unit 202 is used to generate all local candidate paths of the ego vehicle based on the quintic polynomial and the current position information of the ego vehicle;

[0135] The calculation unit 203 is used to introduce the lateral speed of the vehicle to determine the lane state transfer matrix, and calculate the dynamic real-time probability distribution of each surrounding vehicle driving into each lane based on the lane state transfer matrix and the lane switching model;

[0136] The surrounding vehicle trajectory prediction unit 204 is used to generate the future driving trajectory of the surrounding vehicles in different lanes by using a cubic polynomial based on the position information of the surrounding vehicles and the dynamic real-time probability distribution of each surrounding vehicle driving to each lane;

[0137] The evaluation unit 205 is used to calculate the total collision risk value of the vehicle when traveling along each local candidate path based on the dynamic real-time probability distribution and the collision time between the local candidate path of the vehicle and the future driving trajectory of each surrounding vehicle in different lanes.

[0138] In one embodiment of the present invention, the vehicle candidate path generation unit 202 is used to execute:

[0139] Determine several final arc lengths of the local candidate paths of the ego vehicle based on the tangential velocity of the ego vehicle and several preset tangential accelerations;

[0140] Based on the current position information of the ego vehicle and the final arc length, a quintic polynomial is used to calculate several local candidate paths that are used to characterize the variation of the lateral offset of the ego vehicle with the arc length.

[0141] In one embodiment of the present invention, the computing unit 203 is used to execute:

[0142] Determine the initial probability value of switching from lane i to lane j when the lateral speed is 0 to define an initial probability matrix;

[0143] The mean and standard deviation of the Gaussian distribution of the lateral velocity in the single-lane lane change and double-lane lane change cases are defined respectively, so as to determine the modified probability matrix using the Gaussian cumulative distribution function;

[0144] Add the initial probability matrix to the modified probability matrix to obtain the state transition probability value of the surrounding vehicles switching from lane i to lane j, and integrate the lane state transition matrix;

[0145] Based on the lane state transfer matrix and using the lane change model, the initial dynamic real-time probability distribution of each surrounding vehicle driving into each lane is calculated.

[0146] In one embodiment of the present invention, after calculating the initial dynamic real-time probability distribution of each surrounding vehicle driving toward each lane based on the lane state transfer matrix and using the lane switching model, the calculation unit 203 is further used to:

[0147] The actual lateral offset of each surrounding vehicle is weighted and summed according to the initial dynamic real-time probability of the corresponding lane to obtain the estimated mixed lateral offset of the vehicle relative to each lane;

[0148] The outer product is calculated by multiplying the residual of the actual lateral offset of each surrounding vehicle and the estimated mixed lateral offset by its transpose, and the outer product and the initial dynamic real-time probability distribution are weighted summed to combine the lateral offset uncertainties of multiple lanes into a prediction covariance matrix;

[0149] Based on the estimated mixed lateral offset and the predicted covariance matrix, the observation residual and the observation covariance matrix of the ego vehicle sensor are determined;

[0150] Based on the observation residuals and the observation covariance matrix, the likelihood function used to quantify the degree of match between the predicted values ​​and the observed values ​​is calculated, and the historical observations and the current observations are combined using the likelihood function to update the initial dynamic real-time probability distribution of each surrounding vehicle driving into each lane to obtain the final dynamic real-time probability distribution.

[0151] In one embodiment of the present invention, the surrounding vehicle trajectory prediction unit 204 is used to perform:

[0152]

[0153] Where q(s) is the function of the lateral offset of the surrounding vehicles relative to the center line of the lane as the arc length s changes, and a i are the coefficients of the polynomial, s f is the target arc length of the surrounding vehicles traveling along the lane, q f is the distance between the centerline of the target lane and the centerline of the initial lane, q i is the lateral offset at the initial moment, Ψ is the heading angle of the surrounding vehicles relative to the tangent of the road centerline, tanΨ represents the lateral velocity of the surrounding vehicles at the initial moment, dq(s f ) / ds represents the lateral speed of the vehicle at the last moment. A value of 0 means that the vehicle is traveling completely along the center line of the target lane and will not deviate from it.

[0154] In one embodiment of the present invention, the evaluation unit 205 is used to perform:

[0155] Calculate the minimum safe distance between vehicles based on the current speed of the vehicle;

[0156] For each local candidate path of the ego vehicle, execute:

[0157] The two vehicles are considered to collide when they are at the minimum safe distance, and the time required for the collision between the local candidate path of the vehicle and the future driving trajectory of each surrounding vehicle in different lanes is calculated to determine the collision risk probability;

[0158] Based on the collision risk probability and the dynamic real-time probability distribution of each surrounding vehicle driving into each lane, the collision risk value of the local candidate path and each surrounding vehicle is calculated to sum up and calculate the total collision risk value when the vehicle travels along each local candidate path.

[0159] It should be noted that the collision risk assessment device based on lane and surrounding vehicle motion prediction provided in the above embodiment is only illustrated by the division of the above functional units. In actual applications, the above functions can be assigned to different functional units as needed, that is, the internal structure of the device can be divided into different functional units to complete all or part of the functions described above. In addition, the above device embodiment and the method embodiment belong to the same concept, and the specific implementation process is detailed in the method embodiment, which will not be repeated here.

[0160] The embodiment of the present application also provides a computer device, please refer to Figure 3 The computer device includes a processor and a memory, wherein at least one instruction, at least one program, a code set or an instruction set is stored in the memory, and the at least one instruction, at least one program, a code set or an instruction set is loaded and executed by the processor to implement the collision risk assessment method based on lane and surrounding vehicle motion prediction provided by the above-mentioned method embodiments.

[0161] An embodiment of the present application also provides a computer-readable storage medium, on which is stored at least one instruction, at least one program, code set or instruction set, and the at least one instruction, at least one program, code set or instruction set is loaded and executed by a processor to implement the collision risk assessment method based on lane and surrounding vehicle motion prediction provided by the above-mentioned method embodiments.

[0162] An embodiment of the present application also provides a computer program product, which includes a computer program. A processor of a computer device reads the computer program from a computer-readable storage medium, and the processor executes the computer program, so that the computer device executes any of the collision risk assessment methods based on lane and surrounding vehicle motion prediction in the above-mentioned embodiments.

[0163] For the convenience of description, the above system or device is described by dividing it into various modules or units according to its functions. Of course, when implementing the present application, the functions of each unit can be implemented in the same or multiple software and / or hardware.

[0164] It can be known from the description of the above implementation methods that those skilled in the art can clearly understand that the present application can be implemented by means of software plus a necessary general hardware platform. Based on such an understanding, the technical solution of the present application can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product can be stored in a storage medium such as ROM / RAM, a magnetic disk, an optical disk, etc., and includes several instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods of each embodiment of the present application or some parts of the embodiments.

[0165] Finally, it should be noted that, in this article, relational terms such as first, second, third and fourth are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the presence of other identical elements in the process, method, article or device including the elements.

[0166] The above are only preferred implementations of the present application. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present application. These improvements and modifications should also be regarded as the scope of protection of the present application.

Claims

1. A collision risk assessment method based on lane and surrounding vehicle motion prediction, characterized in that: The method comprises: Based on the Cartesian coordinates of the vehicle and surrounding vehicles, calculate the position information of each vehicle in the road coordinate system; wherein the horizontal and vertical coordinates of the road coordinate system represent the arc length along the road curve and the lateral offset between the vehicle and the center line of the road respectively; Based on the quintic polynomial and the current position information of the ego vehicle, all local candidate paths of the ego vehicle are generated; The lateral speed of the vehicle is introduced to determine the lane state transfer matrix, and based on the lane state transfer matrix and the lane switching model, the dynamic real-time probability distribution of each surrounding vehicle driving into each lane is calculated; Based on the position information of the surrounding vehicles and the dynamic real-time probability distribution of each surrounding vehicle driving into each lane, a cubic polynomial is used to generate future driving trajectories of the surrounding vehicles in different lanes; Based on the dynamic real-time probability distribution and the time required for collision between the local candidate path of the vehicle and the future driving trajectory of each surrounding vehicle on different lanes, the total collision risk value of the vehicle when traveling along each local candidate path is calculated.

2. The method according to claim 1, characterized in that The method of generating all local candidate paths of the vehicle based on the quintic polynomial and the current position information of the vehicle includes: Determine several final arc lengths of the local candidate paths of the ego vehicle based on the tangential velocity of the ego vehicle and several preset tangential accelerations; Based on the current position information of the ego vehicle and the final arc length, a quintic polynomial is used to calculate a number of local candidate paths for characterizing the variation of the lateral offset of the ego vehicle with the arc length.

3. The method according to claim 1, characterized in that The lateral speed of the introduced vehicle determines the lane state transfer matrix, and based on the lane state transfer matrix and the lane switching model, calculates the dynamic real-time probability distribution of each surrounding vehicle driving into each lane, including: Determine the initial probability value of switching from lane i to lane j when the lateral speed is 0 to define an initial probability matrix; The mean and standard deviation of the Gaussian distribution of the lateral velocity in the single-lane lane change and double-lane lane change cases are defined respectively, so as to determine the modified probability matrix using the Gaussian cumulative distribution function; Add the initial probability matrix to the modified probability matrix to obtain the state transition probability value of the surrounding vehicles switching from lane i to lane j, and integrate the lane state transition matrix; Based on the lane state transfer matrix, the lane switching model is used to calculate the initial dynamic real-time probability distribution of each surrounding vehicle driving into each lane.

4. The method according to claim 3, characterized in that After calculating the initial dynamic real-time probability distribution of each surrounding vehicle driving toward each lane based on the lane state transfer matrix and using the lane switching model, the method further includes: The actual lateral offset of each surrounding vehicle is weighted and summed according to the initial dynamic real-time probability of the corresponding lane to obtain the estimated mixed lateral offset of the vehicle relative to each lane; The outer product is calculated by multiplying the residual of the actual lateral offset of each surrounding vehicle measured by the sensor and the estimated mixed lateral offset by its transpose, and the outer product and the initial dynamic real-time probability distribution are weighted summed to combine the lateral offset uncertainties of multiple lanes into a prediction covariance matrix; Determining an observation residual and an observation covariance matrix of an ego vehicle sensor based on the estimated mixed lateral offset and the predicted covariance matrix; Based on the observation residual and the observation covariance matrix, a likelihood function for quantifying the degree of match between the predicted value and the observed value is calculated, and the historical observation value and the current observation value are combined using the likelihood function to update the initial dynamic real-time probability distribution of each surrounding vehicle driving into each lane to obtain the final dynamic real-time probability distribution.

5. The method according to claim 1, characterized in that The method of generating future driving trajectories of surrounding vehicles in different lanes by using a cubic polynomial based on the position information of surrounding vehicles and the dynamic real-time probability distribution of each surrounding vehicle driving toward each lane includes: Where q(s) is the function of the lateral offset of the surrounding vehicles relative to the center line of the lane as the arc length s changes, and a i are the coefficients of the polynomial, s f is the target arc length of the surrounding vehicles traveling along the lane, q f is the distance between the centerline of the target lane and the centerline of the initial lane, q i is the lateral offset at the initial moment, Ψ is the heading angle of the surrounding vehicles relative to the tangent of the road centerline, tanΨ represents the lateral velocity of the surrounding vehicles at the initial moment, dq(s f ) / ds represents the lateral speed of the vehicle at the last moment. A value of 0 means that the vehicle is traveling completely along the center line of the target lane and will not deviate from it.

6. The method according to claim 1, characterized in that The calculating of the total collision risk value when the vehicle is traveling along each local candidate path based on the dynamic real-time probability distribution and the time required for collision between the local candidate path of the vehicle and the future driving trajectory of each surrounding vehicle in different lanes comprises: Calculate the minimum safe distance between vehicles based on the current speed of the vehicle; For each local candidate path of the ego vehicle, execute: The two vehicles are considered to collide when they are at the minimum safe distance, and the time required for the collision between the local candidate path of the vehicle and the future driving trajectory of each surrounding vehicle in different lanes is calculated to determine the collision risk probability; Based on the collision risk probability and the dynamic real-time probability distribution of each surrounding vehicle traveling toward each lane, the collision risk value of the local candidate path and each surrounding vehicle is calculated to sum up and calculate the total collision risk value when the vehicle travels along each of the local candidate paths.

7. A collision risk assessment device based on lane and surrounding vehicle motion prediction, used to implement the steps of any of the methods described in claims 1-6 above, characterized in that: The device comprises: A conversion unit, used to calculate the position information of each vehicle in a road coordinate system based on the Cartesian coordinates of the vehicle and surrounding vehicles; wherein the horizontal and vertical coordinates of the road coordinate system represent the arc length along the road curve and the lateral offset between the vehicle and the center line of the road respectively; A candidate path generation unit for the ego vehicle, for generating all local candidate paths for the ego vehicle based on the quintic polynomial and the current position information of the ego vehicle; A calculation unit, used to introduce the lateral speed of the vehicle to determine the lane state transfer matrix, and calculate the dynamic real-time probability distribution of each surrounding vehicle driving into each lane based on the lane state transfer matrix and the lane switching model; A surrounding vehicle trajectory prediction unit, for generating future driving trajectories of surrounding vehicles in different lanes using a cubic polynomial based on the position information of surrounding vehicles and the dynamic real-time probability distribution of each surrounding vehicle driving toward each lane; An evaluation unit is used to calculate a total collision risk value when the vehicle is traveling along each local candidate path based on the dynamic real-time probability distribution and the time required for collision between the local candidate path of the vehicle and the future driving trajectory of each surrounding vehicle on different lanes.

8. A computer device, characterized in that: The computer device includes a memory and a processor, the memory is used to store a computer program, and the processor is used to execute the computer program stored in the memory to implement the steps of any one of the methods described in claims 1-6.

9. A computer-readable storage medium, characterized in that: The storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the method described in any one of claims 1 to 6 are implemented.

10. A computer program product, characterized in that The method comprises a computer program, wherein when the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images