Intelligent vehicle lane changing trajectory planning method based on model predictive control

Abstract

The invention provides an intelligent vehicle lane changing trajectory planning method based on model predictive control, which comprises the following steps: firstly, sensing environment information in real time, calculating an emergency braking distance and an expected safety distance, and deciding behaviors of an intelligent vehicle; secondly, on the basis that a vehicle point quality model serves as a prediction model, an improved artificial potential field method and a model are combined to predict, control and plan a lane changing track, and the current optimal lane changing reference track selected from the sine lane changing reference track cluster is included; and finally, tracking the trajectory by adopting a trajectory tracking controller based on model predictive control. The improved artificial potential field method is adopted to fully adapt to the position, speed and acceleration changes of surrounding obstacle vehicles, the dynamic obstacle avoidance effect is remarkable, and the lane changing safety of the intelligent vehicle is guaranteed; the situations of following the front vehicle, canceling lane change and the like are also considered, and the adaptive scene is wider.

Description

Technical Field

[0001] The invention relates to a lane-changing trajectory planning method for an intelligent vehicle based on model predictive control, and belongs to the technical field of lane-changing trajectory planning for intelligent vehicles. Background Art

[0002] In recent years, with the rapid development of computer technology and perception technology, autonomous driving technology is moving from partial autonomous driving to fully autonomous driving. Technologies such as adaptive cruise control, driver assistance, and lane change assistance have been widely used in vehicles, effectively improving the safety, comfort, and efficiency of driving. Changing lanes is the most common driving behavior in the driving process. How to change lanes reasonably and safely has become the focus of current autonomous driving technology. Reasonable and safe lane changes can improve driving efficiency, while unreasonable lane changes will lead to safety accidents. In order to enable intelligent vehicles to complete lane change tasks more safely and efficiently, it is necessary to plan lane change trajectories that meet safety, feasibility, and comfort.

[0003] Artificial Potential Field (APF) and Model Predictive Control (MPC) have been widely used in trajectory planning for smart cars. MPC predicts the vehicle state in the future limited time domain based on the existing model and current state, and then designs appropriate objective functions and constraints according to the needs to construct an optimization problem. The optimal control variables are obtained by solving the optimization problem to achieve the purpose of planning and control. However, complex models will increase the amount of calculation of MPC, which will greatly affect the real-time performance of smart car planning and control. APF is a classic obstacle avoidance method for smart cars. Under the action of the attraction designed based on the target point and the repulsion designed based on the obstacles and road boundaries, the smart car can avoid obstacles and reach the target point. However, few existing studies consider the acceleration information of surrounding obstacles and road boundary information in the potential field, and there are few strategies for when the lane change conditions are not met and the safety of the smart car is threatened during the lane change process, resulting in the safety and feasibility of the smart car lane change process to be improved. Summary of the invention

[0004] The technical problem to be solved by the present invention is to overcome the problem in the prior art that the APF only considers the position and speed of the surrounding obstacles but not the acceleration and road boundaries, and proposes a lane change trajectory planning method for an intelligent vehicle based on model predictive control, which uses a sinusoidal curve as a reference trajectory when changing lanes, and obtains a smoother and more efficient lane change trajectory while meeting the conditions of safety and feasibility.

[0005] The present invention provides a lane change trajectory planning method for an intelligent vehicle based on model predictive control. The system involved in the method includes an environment perception module, a behavior decision module, a lane change trajectory planning module and a trajectory tracking module. The method includes the following steps:

[0006] Step 1: Obtain environmental perception information based on laser radar, GPS sensor and vehicle speed sensor. The environmental perception information includes the status information of the smart car and surrounding dynamic obstacles and the information of the current road of the smart car. The status information of the smart car and obstacles includes posture, speed, acceleration, etc. The road information includes the center line and boundary line of each lane. The status information of the smart car and surrounding dynamic obstacles is input into the emergency braking distance model and the expected safety distance model, and the behavior decision information is output according to the decision scheme;

[0007] Step 2: Construct the corresponding objective function of trajectory planning based on improved artificial potential field method and model predictive control according to the environmental perception information and behavior decision information, and construct the optimization problem of trajectory planning module in combination with constraint conditions;

[0008] Step 3: Select the optimal sinusoidal lane change reference trajectory under current conditions according to the lane change trajectory evaluation function to ensure the safety, efficiency and comfort of the lane change process.

[0009] The present invention provides real-time acquisition of relevant information based on the environment perception module to make decisions, plan and control. The present invention first perceives the environment information in real time and calculates the emergency braking distance and the expected safety distance to decide the behavior of the smart car; then, based on the vehicle point mass model as a prediction model, the improved artificial potential field method and model predictive control are combined to plan the lane change trajectory, which includes the current optimal lane change reference trajectory selected from the sinusoidal lane change reference trajectory cluster; finally, a trajectory tracking controller based on model predictive control is used to track the trajectory.

[0010] The technical solution further optimized as the present invention is as follows:

[0011] In the step 1, the state information of the dynamic obstacle vehicles and the self-vehicle around the smart car is considered in real time, and the information is input into the emergency braking distance model and the expected safety distance model. Further, according to the emergency braking distance, the expected safety distance and the longitudinal distance between the self-vehicle and the surrounding obstacle vehicles, it is judged whether there is an intention to change lanes and whether the feasibility of changing lanes is met, and finally the behavior decision information of the smart car is output;

[0012] The emergency braking distance refers to the distance between the rear vehicle and the front vehicle that can prevent the rear vehicle from colliding with the front vehicle and ensure driving safety when the front vehicle brakes urgently. The emergency braking distance model is:

[0013]

[0014] Where S is the emergency braking distance of the vehicle in the current state, v0 is the speed of the vehicle before emergency braking, t1 is the time for the deceleration to increase linearly, S ε is the correction distance, b max is the maximum deceleration and b max =μg, μ is the road adhesion coefficient, g is the acceleration due to gravity and g=9.8m / s 2 ;

[0015] The expected safety distance is mainly related to the speed of the front and rear vehicles, the road adhesion coefficient, the gravity acceleration and the emergency braking distance, and measures the minimum distance that the rear vehicle is expected to maintain with the front vehicle when traveling at the expected speed. The expected safety distance model is:

[0016]

[0017] Where D fs is the expected safety distance under the current state of the vehicle, τ r is the sensor delay time, v M is the longitudinal speed of the rear vehicle; v F is the longitudinal speed of the preceding vehicle, S M is the emergency braking distance of the rear vehicle, a F is the acceleration of the front vehicle.

[0018] In step 1, the specific method for detecting the generation of lane change intention and the feasibility of lane change is:

[0019] (1) Generation of lane change intention - judging whether the longitudinal distance between the smart car and the vehicle ahead in the current lane is less than the expected safety distance of the smart car. If the longitudinal distance between the smart car and the vehicle ahead in the current lane is less than the expected safety distance of the smart car, it means that the safety of the smart car cannot be guaranteed when driving at the expected speed, so the lane change intention is generated; otherwise, it means that the safety performance of the smart car is guaranteed when driving at the expected speed, so the lane change intention is not generated;

[0020] (2) Lane change feasibility condition: The longitudinal distance between the smart car and the vehicle behind the target lane is used to determine whether the lane change feasibility condition is met. When the longitudinal distance between the smart car and the vehicle behind the target lane is less than the emergency braking distance of the vehicle behind the target lane, it means that the lane change feasibility condition is not met. Otherwise, it means that the lane change feasibility condition is met.

[0021] In step 1, a method for constructing intelligent vehicle behavior decision planning according to lane change intention and lane change feasibility is specifically implemented as follows:

[0022] Step 1.1, determine whether there is an intention to change lanes, and if so, further determine whether the feasibility conditions for lane change are met; if there is no intention to change lanes, maintain the desired speed;

[0023] Step 1.2, judging whether the feasibility condition of lane change is met, if the feasibility condition of lane change is met, executing lane change; if not, slowing down and following the vehicle ahead;

[0024] Step 1.3: During the lane change process, the longitudinal distance between the smart car and the vehicle behind the target lane is detected in real time, and whether the lane change safety condition is met is determined based on the detection result, that is, whether the longitudinal distance between the smart car and the vehicle behind the target lane detected in real time is greater than the emergency braking distance of the vehicle behind the target lane. If the lane change process safety is not met, the lane change is canceled and the vehicle continues to slow down and follow the vehicle ahead.

[0025] Step 1.4: Output the behavior decision information to the trajectory planning module.

[0026] The specific decision-making process is as follows: when the distance between the ego vehicle and the vehicle in front is less than the desired safety distance of the ego vehicle, the ego vehicle will have the intention to change lanes, otherwise it will maintain the desired state. Whether the feasibility of the lane change is established depends on the longitudinal distance between the ego vehicle and the vehicle behind in the target lane. The present invention only considers the case where there is only one vehicle in the target lane. After the intention to change lanes is generated, if the longitudinal distance to the vehicle in the target lane is greater than the emergency braking distance of the vehicle behind the target lane, the feasibility of the lane change is met and the lane change is executed, otherwise the vehicle in front should be followed. During the lane change process, if the distance between the ego vehicle and the surrounding obstacle vehicles is less than the emergency safety distance, the lane change should be cancelled.

[0027] In step 2, after receiving the behavior decision information, the trajectory planning module will construct the objective function and optimization problem of the trajectory planning module based on the improved artificial potential field method and model predictive control according to the behavior decision information and environmental perception information. The improved artificial potential field method takes into account the position, speed, and acceleration of the vehicle and the position, speed, and acceleration of the surrounding obstacle vehicles, as well as the road information of the environment, and specifically includes the road potential field function and the obstacle potential field function, and its calculation method is as follows:

[0028] (1) The calculation method of the road potential field function is:

[0029] U road =k r (yy l ) 2 (yy r ) 2

[0030] Where U road is the road potential field function, k r is the adjustment coefficient of the road potential field, y is the lateral position of the vehicle, and y l ,y r They are the left and right centerline positions of a two-lane road, respectively;

[0031] (2) The calculation method of the obstacle potential field function is:

[0032] U obs =w1k o f1(x|μ1,∑1)+w2k o f2(x|μ2,∑2)

[0033] Where μ1=(x1,y1) T ,μ2=(x2,y2) T

[0034]

[0035] Among them, U obs is the obstacle potential field function, w1 and w2 are weights, and w1+w2=1, x is the position information of the vehicle and x=(x av ,y av ) T , k o is the comprehensive coefficient to measure the impact of obstacles. f1 and f2 are two-dimensional joint probability density functions, μ1 and μ2 are means, x1, y1, ∑1, To calculate the parameters related to the mean and covariance of f1, x2, y2, ∑2, To calculate the parameters related to the mean and covariance of f2;

[0036] Furthermore, the mean and covariance matrix of f1 and f2 are defined as:

[0037]

[0038] In the formula, (x obs ,y obs ) is the position coordinate of the obstacle vehicle, (v x_obs ,v y_obs ) are the longitudinal and lateral velocities of the obstacle vehicle, (a x_obs ,a y_obs ) is the longitudinal acceleration and lateral acceleration of the obstacle vehicle, S x To measure the longitudinal safety distance, S y To measure the horizontal safety distance, S o is the minimum longitudinal distance determined by the vehicle size, D o is the minimum lateral distance determined by the vehicle size, Δx and Δy represent the adjustable distances in the longitudinal and lateral directions respectively, and T0 is the time interval; k m is a positive parameter related to the vehicle weight and is defined as:

[0039]

[0040] m is the mass of the obstacle vehicle.

[0041] In step 2, the objective function and optimization problem of the trajectory planning module can be switched arbitrarily under different decision conditions, thereby ensuring the flexibility of the driving trajectory of the smart car. The optimization problem of the trajectory planning module is specifically expressed as:

[0042]

[0043]

[0044] U min ≤U i ≤U max

[0045] y rr ≤y t ≤y rl

[0046] |u(t)|<μg

[0047] Where N p is the prediction time domain, N c is the control time domain, η((t+i|t) is the prediction value in the prediction time domain, η ref (t+i|t) is the reference quantity in the prediction time domain, t is the current sampling time, i is the i-th time in the current prediction time domain, U i is the control quantity, U all is the total function of the constructed artificial potential field, Q and R are weight matrices, is the derivative of the state quantity in the vehicle point mass model, ξ(t) is the state quantity in the vehicle point mass model, u(t) is the control quantity in the vehicle point mass model, U min , U max are the upper and lower bounds of the controlled quantity, y t is the lateral position of the vehicle at time t, y rr ,y rl are lane boundaries, and |u(t)|<μg is the constraint condition of the control quantity.

[0048] In step 3, the reference quantity in the prediction time domain in the trajectory planning module optimization problem is the optimal sinusoidal lane change reference trajectory when executing the lane change decision. When a combination of lane change time and longitudinal acceleration is determined, a sinusoidal lane change reference trajectory can be determined. The calculation formula of the sinusoidal lane change reference trajectory is as follows:

[0049] The lateral acceleration at any time t is:

[0050]

[0051] The lateral velocity at any time t is:

[0052]

[0053] The lateral displacement model at any time t is:

[0054]

[0055] In the formula, t s is the lane change start time, t e is the lane change end time, and W is the lane width.

[0056] In step 3, the method for solving the optimal sinusoidal lane change reference trajectory is as follows:

[0057] Step 3.1, after receiving the lane change instruction from the decision module, based on the state information of surrounding vehicles acquired in real time, assuming that the acceleration of surrounding vehicles remains unchanged during the lane change process, predict their motion state within the lane change time;

[0058] Step 3.2: Sample the lane change time interval ΔT in [2s, 6s] and 2 ,3m / s 2 ] to sample the longitudinal acceleration interval Δa, and any combination of lane change time and longitudinal acceleration can determine a lane change reference trajectory;

[0059] Step 3.3, evaluate the lane change trajectory, consider the safety, riding comfort and efficiency of the smart car's lane change process, linearly weight the three indicators to obtain the lane change reference trajectory evaluation function, and determine the optimal lane change time and longitudinal acceleration by finding the maximum value of the function.

[0060] The evaluation function for measuring the safety, efficiency and comfort of the lane change reference trajectory and the final linear weighted objective function are:

[0061] (1) Lane-changing process safety evaluation function

[0062]

[0063] Where N is the total number of obstacle vehicles, k s is the safety factor, x(t) and y(t) are the longitudinal and lateral positions of the smart car at time t, respectively. are the longitudinal position and lateral position of the i-th obstacle vehicle at time t respectively;

[0064] (2) Lane-changing process efficiency evaluation function

[0065]

[0066] In the formula, k et , k ev are the efficiency coefficients related to time and speed, v x(t), v y (t) are the longitudinal speed and lateral speed of the smart car at time t, respectively, and T is the lane change time of the current sampling;

[0067] (3) The comfort index of the lane change process is usually measured by acceleration, and the comfort evaluation function

[0068]

[0069] In the formula, k cx , k cy are the longitudinal and lateral comfort coefficients, a x (t), a y (t) are the longitudinal acceleration and lateral acceleration of the smart car at time t;

[0070] The comprehensive evaluation function is obtained by linearly weighting the safety, comfort and efficiency evaluation functions, as shown below:

[0071] j=w s j s +w c j c +w e j e

[0072] In the formula, w s 、w c 、w e are the weight coefficients of safety, comfort and efficiency respectively.

[0073] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0074] (1) The present invention adopts an improved artificial potential field function, which not only considers the position and speed of dynamic obstacles, but also adds acceleration information, and constructs a road potential field function based on road information; thus, the smart car is more sensitive to the dynamic changes of dynamic obstacles and tries to drive on the center line of the lane, thereby increasing the safety of the smart car during driving.

[0075] (2) The present invention designs an evaluation function for selecting the optimal sinusoidal lane change reference trajectory, and selects the optimal lane change reference trajectory from a series of lane change reference trajectory clusters, thereby improving the safety, efficiency and comfort of the lane change process.

[0076] (3) The present invention can not only plan lane change trajectories, but also construct different trajectory planning optimization problems based on decision information, thereby planning a reasonable trajectory that meets the current driving environment, so that the flexibility of the smart car in the lane change process is effectively improved. Specifically, using the method of the present invention, the smart car can achieve situations such as canceling lane changes, slowing down to follow the vehicle in front, etc., which improves the flexibility of the smart car in the face of emergencies during the lane change process.

[0077] In summary, the improved artificial potential field method adopted in the present invention can fully adapt to the changes in the position, speed and acceleration of the surrounding obstacle vehicles, and the dynamic obstacle avoidance effect is remarkable, which ensures the safety of the smart car changing lanes. The present invention also takes into account situations such as following the preceding vehicle and canceling the lane change, and is adaptable to a wider range of scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0078] Figure 1 This is a structural diagram of the lane change trajectory planning system of the present invention.

[0079] Figure 2 This is a flow chart of the lane change trajectory planning method proposed in the present invention.

[0080] Figure 3 The figure is a behavioral decision flow chart of the present invention.

[0081] Figure 4 This is a vehicle-road simulation diagram of the road condition of the present invention.

[0082] Figure 5 This is a diagram of the lateral controller tracking simulation results when the speed is 60km / h.

[0083] Figure 6 This is a diagram of the simulation results of the longitudinal controller tracking of the present invention.

[0084] Figure 7 The figure is a simulation diagram of a lane change trajectory according to an embodiment of the present invention.

[0085] Figure 8 The figure is a vehicle position diagram at different times according to an embodiment of the present invention. DETAILED DESCRIPTION

[0086] The technical solution of the present invention is further described in detail below in conjunction with the accompanying drawings: This embodiment is implemented on the premise of the technical solution of the present invention, and a detailed implementation method and specific operation process are given, but the protection authority of the present invention is not limited to the following embodiments.

[0087] like Figure 1 As shown, the intelligent vehicle lane change trajectory planning system based on artificial potential field method and model predictive control includes an environment perception module, a behavior decision module, a lane change trajectory planning module and a trajectory tracking module. Among them:

[0088] The environmental perception module includes laser radar, GPS sensor and vehicle speed sensor, etc., which are used to obtain information about the smart car, surrounding obstacles and roads;

[0089] The behavior decision module includes the emergency braking distance model and the expected safety distance model, as well as the decision-making process based on these two models;

[0090] The lane-changing trajectory module includes constructing a point mass model of the vehicle to predict the vehicle's future state in a limited time domain, designing a road potential field and an obstacle potential field based on a two-dimensional joint probability density function to ensure the safety of the smart car during driving, solving the optimal lane-changing time and longitudinal acceleration based on a sinusoidal lane-changing reference trajectory model, and finally designing an objective function that satisfies the safety, smoothness, and efficiency of the smart car's lane-changing, and solving the optimal lane-changing trajectory at the current moment;

[0091] The trajectory tracking module includes building a vehicle lateral tracking controller and a longitudinal tracking controller based on the model predictive control algorithm to achieve trajectory tracking.

[0092] like Figure 2 As shown, the intelligent vehicle lane change trajectory planning method based on artificial potential field method and model predictive control includes:

[0093] S1. Acquire the status information of the smart car, the status information of the surrounding dynamic obstacles and the information of the road the smart car is currently traveling on in real time. The status information of the smart car and obstacles includes position, speed, acceleration, etc. The road information includes the center line and boundary line of each lane.

[0094] The position and posture of the smart car are obtained by the GPS sensor; the speed information and acceleration information of the smart car are obtained through the vehicle speed sensor; the road information is obtained by the high-precision map, including road boundary information, lane line information, etc.; the status information and speed of the surrounding obstacle vehicles are obtained through the lidar.

[0095] S2. Input the state information of the smart car and the obstacle obtained in step S1 into the emergency braking distance model and the expected safety distance model to determine whether there is a need to change lanes. If there is a need to change lanes, determine whether the lane change conditions are met. If the lane change conditions are not met, follow the vehicle in front, otherwise change lanes. The decision results include maintaining the state (when there is no need to change lanes), following the vehicle in front (when there is a need to change lanes but the lane change conditions are not met), and changing lanes (when there is a need to change lanes and the lane change conditions are met). Figure 3 Provide behavioral decision-making plans.

[0096] Specifically, the emergency braking distance refers to the distance between the rear vehicle and the front vehicle when the front vehicle brakes urgently, which is just enough to prevent the rear vehicle from colliding with the front vehicle, thus ensuring driving safety. The emergency braking distance model is:

[0097]

[0098] Where S is the emergency braking distance of the vehicle in the current state, v0 is the speed of the vehicle before emergency braking, t1 is the time for the deceleration to increase linearly, S ε is the correction distance, the maximum deceleration is b max =μg, μ is the road adhesion coefficient, g=9.8m / s 2is the acceleration due to gravity.

[0099] The expected safety distance is mainly related to the speed of the front and rear vehicles, the road adhesion coefficient, the gravitational acceleration and the emergency braking distance. It measures the minimum distance that the rear vehicle is expected to maintain with the front vehicle when driving at the expected speed. The expected safety distance model is:

[0100]

[0101] In the formula, τ r is the sensor delay time; v M is the longitudinal speed of the rear vehicle; v F is the longitudinal speed of the preceding vehicle; S M is the emergency braking distance of the rear vehicle, a F is the acceleration of the front vehicle.

[0102] Based on the emergency braking distance, expected safety distance and the longitudinal distance between the vehicle and surrounding obstacle vehicles, it is judged whether the smart car has the intention to change lanes and whether the feasibility of lane change is met, and finally the behavior decision information of the smart car is output.

[0103] Among them, the specific method for detecting the generation of lane change intention and the feasibility of lane change is:

[0104] (1) Generation of lane change intention - judging whether the longitudinal distance between the smart car and the vehicle ahead in the current lane is less than the expected safety distance of the smart car. If the longitudinal distance between the smart car and the vehicle ahead in the current lane is less than the expected safety distance of the smart car, it means that the safety of the smart car cannot be guaranteed when driving at the expected speed, so the lane change intention is generated; otherwise, it means that the safety performance of the smart car is guaranteed when driving at the expected speed, so the lane change intention is not generated;

[0105] (2) Lane change feasibility condition: The longitudinal distance between the smart car and the vehicle behind the target lane is used to determine whether the lane change feasibility condition is met. When the longitudinal distance between the smart car and the vehicle behind the target lane is less than the emergency braking distance of the vehicle behind the target lane, it means that the lane change feasibility condition is not met. Otherwise, it means that the lane change feasibility condition is met.

[0106] The method of constructing intelligent vehicle behavior decision planning based on lane change intention and lane change feasibility is as follows:

[0107] Step 1.1, determine whether there is an intention to change lanes, and if so, further determine whether the feasibility conditions for lane change are met; if there is no intention to change lanes, maintain the desired speed;

[0108] Step 1.2, judging whether the feasibility condition of lane change is met, if the feasibility condition of lane change is met, executing lane change; if not, slowing down and following the vehicle ahead;

[0109] Step 1.3: During the lane change process, the longitudinal distance between the smart car and the vehicle behind the target lane is detected in real time, and a determination is made based on the detection result whether the lane change safety condition is met (i.e., whether the longitudinal distance between the smart car and the vehicle behind the target lane detected in real time is greater than the emergency braking distance of the vehicle behind the target lane). If the lane change process safety condition is not met, the lane change is canceled and the vehicle continues to decelerate and follow the vehicle ahead.

[0110] Step 1.4: Output the behavior decision information to the trajectory planning module.

[0111] S3. Construct different objective functions according to the decision information of step S2. The objective function of local trajectory planning should include the reference state under the decision. In particular, when the decision is to change lanes, an evaluation function is established based on the sinusoidal lane change curve cluster to select the current optimal lane change reference trajectory from the sinusoidal lane change reference trajectory, and determine the lane change time and longitudinal acceleration. During the lane change process, it is detected in real time whether the smart car is under a safety threat. If so, cancel the lane change and follow the front vehicle, otherwise continue to execute the lane change operation. According to the decision information output by step S2, construct the objective function of trajectory planning based on improved APF and MPC.

[0112] The specific method of constructing the objective function of trajectory planning based on improved APF and MPC is as follows:

[0113] (1) Establish a point mass model of the vehicle, which can be abbreviated as:

[0114]

[0115] In the formula, φ, Y, X are the lateral speed, longitudinal speed, vehicle heading angle, and the horizontal and vertical coordinates of the vehicle in the Cartesian coordinate system; u is the control variable a y .

[0116] (2) The improved artificial potential field method includes road potential field function and obstacle potential field function.

[0117] The road potential field function is constructed as:

[0118] U road =k r (yy l ) 2 (yy r ) 2

[0119] Where U road is the road potential field function, k r is the adjustment coefficient of the road potential field, y is the lateral position of the vehicle, and y l ,y rThey are the left and right centerline positions of a two-lane road respectively.

[0120] The specific construction method of the obstacle potential field based on the two-dimensional joint probability density function is as follows:

[0121] The two-dimensional joint probability density function is:

[0122]

[0123] Where x = (x1, x2) T , x i are all one-dimensional random variables, x i The expectation is μ i , ∑ is the covariance matrix.

[0124] The obstacle potential field function is defined as:

[0125] U obs =w1k o f1(x|μ1,∑1)+w2k o f2(x|μ2,∑2)

[0126] Where μ1=(x1,y1) T ,μ2=(x2,y2) T

[0127]

[0128] Among them, U obs is the obstacle potential field function, w1 and w2 are weights, and w1+w2=1, x is the position information of the vehicle and x=(x av ,y av ) T , k o is the comprehensive coefficient to measure the impact of obstacles. f1 and f2 are two-dimensional joint probability density functions, μ1 and μ2 are means, x1, y1, ∑1, To calculate the parameters related to the mean and covariance of f1, x2, y2, ∑2, Parameters related to calculating the mean and covariance of f2.

[0129] Furthermore, the mean and covariance matrix of f1 and f2 are defined as:

[0130]

[0131] In the formula, (x obs ,y obs ) is the position coordinate of the obstacle vehicle, (v x_obs ,v y_obs ) are the longitudinal and lateral velocities of the obstacle vehicle, (a x_obs ,ay_obs ) is the longitudinal acceleration and lateral acceleration of the obstacle vehicle, S x , S y Measures the longitudinal and lateral safety distance, S o , D o is the minimum longitudinal and lateral distance determined by the vehicle size, Δx and Δy represent the adjustable longitudinal and lateral distances, and T0 is the time interval; k m is a positive parameter related to the vehicle weight and is defined as:

[0132]

[0133] m is the mass of the obstacle vehicle.

[0134] Finally, the total potential field function is the sum of the road potential field and the obstacle potential field, defined as:

[0135] U all =U road +U obs

[0136] The objective function of local trajectory planning should include the constructed artificial potential field function in order to ensure the safety of smart car driving.

[0137] (3) If the decision result in step S2 is to change lanes, in order to ensure safety, efficiency and comfort during the lane change process, the optimal lane change reference trajectory is selected from the sinusoidal curve cluster as the reference trajectory during the lane change process. The sinusoidal curve lane change reference trajectory model can be divided into lateral acceleration, lateral velocity and lateral displacement models. The specific form is as follows:

[0138] The lateral acceleration a at any time t y (t) is:

[0139]

[0140] The lateral velocity v at any time t y (t) is:

[0141]

[0142] The lateral displacement model y at any time t y (t) is:

[0143]

[0144] In the formula, t s is the lane change start time, t e is the lane change end time, A y is the peak value of lateral acceleration, and W is the lane width.

[0145] The optimal lane change reference trajectory is obtained based on the following three principles:

[0146] (1) After receiving the lane change command from the decision module, based on the real-time status information of surrounding vehicles, assuming that the acceleration of surrounding vehicles remains unchanged during the lane change process, predict their motion status during the lane change time.

[0147] (2) Sample the lane change time interval ΔT within [2s, 6s] and 2 ,3m / s 2 ]The longitudinal acceleration interval Δa is sampled, and any combination of lane change time and longitudinal acceleration can determine a lane change reference trajectory.

[0148] (3) Evaluate the lane change trajectory, taking into account the safety, riding comfort and efficiency of the smart car's lane change process. The three indicators are linearly weighted to obtain the lane change reference trajectory evaluation function, and the lane change time and longitudinal acceleration are determined by finding the minimum value of the function.

[0149] Furthermore, the safety evaluation function j in the lane change process is constructed s for:

[0150]

[0151] In the formula, k s is the safety factor, x(t) and y(t) are the longitudinal and lateral positions of the smart car at time t, respectively. are the longitudinal position and lateral position of the i-th obstacle vehicle at time t, respectively, and N is the total number of obstacle vehicles.

[0152] Furthermore, the efficiency evaluation function j in the lane change process is constructed e for:

[0153]

[0154] In the formula, k et , k ev are the efficiency coefficients related to time and speed, v x (t), v y (t) are the longitudinal speed and lateral speed of the smart car at time t, respectively, and T is the lane change time of the current sampling.

[0155] Furthermore, the comfort index is often measured by acceleration, and the comfort evaluation function j is constructed. c for:

[0156]

[0157] In the formula, k cx , k cyare the longitudinal and lateral comfort coefficients, a x (t), a y (t) are the longitudinal acceleration and lateral acceleration of the smart car at time t.

[0158] The comprehensive evaluation function is obtained by linearly weighting the safety, comfort and efficiency evaluation functions, as shown below:

[0159] j=w s j s +w c j c +w e j e

[0160] In the formula, w s 、w c 、w e are the weight coefficients of safety, comfort and efficiency respectively. By evaluating the objective function of the optimal sinusoidal lane change reference trajectory and solving the minimum value of the objective function, the optimal lane change time and longitudinal acceleration can be obtained, and then the optimal sinusoidal lane change reference trajectory can be obtained.

[0161] (4) By finding the minimum value of j, the optimal lane change time and longitudinal acceleration can be determined, and then the optimal sinusoidal lane change reference trajectory during the lane change process can be determined. The vehicle point mass model is constructed as a prediction model to predict the state information of the vehicle in a limited time domain. Combined with the artificial potential field function and the reference trajectory, the objective function of the trajectory planning module can be constructed.

[0162] Furthermore, the objective function of constructing the local trajectory of the smart car based on APF and MPC is:

[0163]

[0164] Where N p 、N c are the prediction time domain and the control time domain, η((t+i|t), η ref (t+i|t) is the predicted value and reference value in the prediction time domain, U i is the control quantity, U all is the total function of the constructed artificial potential field.

[0165] S4. Construct an optimization problem according to the objective function of step S3 and the construction constraints.

[0166] Based on the objective function of the trajectory planning module and considering the relevant constraints of the smart car, the optimization problem of the trajectory planning module is constructed as follows:

[0167]

[0168]

[0169] U min ≤U i ≤U max

[0170] y rr ≤y t ≤y rl

[0171] |u(t)|<μg

[0172] Where N p 、N c are the prediction time domain and the control time domain, η((t+i|t), η ref (t+i|t) is the predicted value and reference value in the prediction time domain, t is the current sampling time, i is the i-th time in the current prediction time domain, U i is the control quantity, U all is the total function of the constructed artificial potential field, Q and R are weight matrices, is the derivative of the state quantity in the vehicle point mass model, ξ(t) is the state quantity in the vehicle point mass model, u(t) is the control quantity in the vehicle point mass model, U min , U max are the upper and lower bounds of the controlled quantity, y t is the lateral position of the vehicle at time t, y rr ,y rl is the lane boundary, |u(t)|<μg is the constraint condition of the control quantity.

[0173] In particular, η in the optimization problem ref (t+i|t) will change based on the decision information, such as when there is a lane change demand but no lane change conditions, η ref (t+i|t) At this time, the current lane and the speed of the front vehicle should be used as the target trajectory to make the smart car follow the vehicle. When the lane change condition is met, the calculated optimal sinusoidal lane change reference trajectory is used as η ref (t+i|t).

[0174] S5. Solve the optimization problem constructed in step S4 to obtain Nc-dimensional discrete local trajectory points.

[0175] S6. Smooth the discrete local trajectory output in step S5 by a polynomial fitting method, and send the fitting parameters to the trajectory tracking module.

[0176] The local trajectory solved in step S5 is a discrete trajectory point. If it is directly input into the trajectory tracking control layer, it is not conducive to tracking. Therefore, the discrete local trajectory is smoothed by the polynomial fitting method, and the fitting parameters are sent to the trajectory tracking module. The fifth-order polynomial expression is:

[0177] f(t)=a0+a1t+a2t 2 +a3t 3 +a4t 4 +a5t 5

[0178] In the formula, a=[a0 a1 a2 a3 a4 a5] is the coefficient of the polynomial to be determined, and a0, a1, a2, a3, a4, and a5 are the coefficients of the constant term to the fifth-order term in the polynomial respectively.

[0179] S7. Design a lateral and longitudinal trajectory tracking controller based on model predictive control, and track the trajectory according to the parameters of the local trajectory output in step S6.

[0180] (1) Construct an MPC lateral trajectory tracking controller based on the vehicle's dynamic model. The vehicle's dynamic model is:

[0181]

[0182] It can be simplified as:

[0183]

[0184] In the formula, are the derivatives of the state variables and the control variables in the vehicle dynamics model, respectively. is the state quantity of the system, Y, X are the lateral speed, longitudinal speed, vehicle yaw angle, vehicle yaw angle change rate, vehicle lateral position, vehicle longitudinal position, front wheel steering angle u dyn =δ f is the control quantity of the system.

[0185] The kinetic model is linearized as:

[0186]

[0187] Furthermore, it is discretized into:

[0188] ξ dyn (k+1)=A dyn (k)ξ dyn (k)+B dyn (k)u dyn (k)

[0189] In the formula, A dyn (k) = I + T dyn A dyn (t), B dyn (k) = T dyn B dyn (t), I is the unit matrix, Tdyn is the sampling period. The discretized system state equation is:

[0190]

[0191] In the formula, A dyn (k),ξ dyn (k), B dyn (k),u dyn (k), η dyn (k), C dyn (k) are state quantity coefficient matrix, state quantity, control quantity coefficient matrix, control quantity, output quantity coefficient matrix, and output quantity respectively.

[0192] (2) In order to improve the stability of the intelligent vehicle's motion, the control quantity increment of the system is used as the controlled quantity, so a new state vector is constructed as follows:

[0193] ξ dyn (k|t)=[x dyn u dyn (k-1)] T

[0194] In the formula, x dyn is the state variable in the dynamic model.

[0195] The new state space expression is:

[0196]

[0197] In the formula, Δu dyn (k) is the control quantity increment.

[0198] The output of the new system in the prediction time domain is as follows:

[0199] Y dyn (t) = Ψ dyn ξ dyn (t|t)+Θ dyn ΔU dyn (t)

[0200] In the formula, Ψ dyn is the state quantity coefficient matrix, ξ dyn (t|t) is the state matrix, Θ dyn is the control increment coefficient matrix, ΔU dyn (t) is the control quantity increment matrix.

[0201] Considering the complexity of solving the problem, a relaxation factor is added to the objective function, and the objective function of the lateral trajectory tracking controller is constructed as follows:

[0202]

[0203] Where η dyn (t+i|t) is the predicted value in the prediction time domain at time t, η dyn,ref (t+i|t) is the reference quantity in the prediction time domain at time t, Δu dyn (t+i|t) is the control quantity increment in the control time domain at time t, ρ is a constant, and ε is the relaxation factor.

[0204] The optimization problem to be solved in each control cycle is described as

[0205]

[0206] stΔu dyn,min ≤Δu dyn,t ≤Δu dyn,max

[0207] u dyn,min ≤AΔu dyn,t +u dyn,t ≤u dyn,max

[0208] y hc,min ≤y hc ≤y hc，max

[0209] y sc,min -ε≤y sc ≤y sc,max +ε

[0210] ε>0

[0211] Where N p 、N c are the prediction time domain and the control time domain respectively, y hc is a hard constraint output. A hard constraint output means that the output cannot increase the constraint range. sc is a soft constraint output. Soft constraint output means that the relaxation factor can dynamically change the output range of the output. hc,min and hc,max is the hard constraint limit, y sc,min and sc,max is the soft constraint limit.

[0212] (3) Solving this problem, we can obtain a series of control increments in the control time domain, as shown below:

[0213]

[0214] In the formula, are the control quantity increments in the control domain solved at time t respectively.

[0215] The first element in the control sequence is applied to the system as the actual control increment, as shown below:

[0216] u dyn (t) = u dyn (t-1)+Δu dyn,t

[0217] Through rolling optimization and feedback correction, the intelligent vehicle can achieve lateral tracking of the trajectory.

[0218] Furthermore, the longitudinal control of the vehicle can be expressed using a first-order inertial system as follows:

[0219]

[0220] Where K is the system gain; τ d is the time constant, a des is the reference acceleration, and a is the actual acceleration. The continuous system state equation of the vehicle longitudinal motion can be expressed as:

[0221]

[0222] Where x = [va] T is the state quantity of the system, u=a des is the control input of the system, τ d is the time constant, a des is the reference acceleration.

[0223] Furthermore, it is discretized as follows:

[0224] x(k+1)=A k x(k)+B k u(k)

[0225] In the formula,

[0226]

[0227] Among them, k is the current sampling time, T s is the sampling period.

[0228] The control target of longitudinal control is to track the given speed, so the speed v is used as the system output, and the output equation is:

[0229]

[0230] Taking into account the smoothness of the control process, the change of the control amount is taken as the control amount, and the cost function shown below is constructed:

[0231]

[0232] yp (t+i|t) is the predicted value in the prediction time domain at time t, y ref (t+i|t) is the reference quantity in the prediction time domain at time t.

[0233] (4) Considering the relevant constraints, the optimization problem is constructed as:

[0234]

[0235] Stu min ≤u(t+i)≤u max

[0236] Δu min ≤Δu(t+i)≤Δu max

[0237] Where, N p is the prediction time domain, N c is the prediction time domain, u min 、u max is the maximum value of acceleration, Δu min , Δu max is the maximum value of the acceleration increment.

[0238] Solving this optimization problem will obtain a series of control increments in the control time domain. The first output value plus the control amount at the previous moment will give the optimal control amount at that moment. Inputting it into the controlled vehicle will realize the speed tracking function.

[0239] In summary, the model predictive control lateral and longitudinal trajectory tracking controllers are designed based on the vehicle dynamics model and the first-order inertial system. The optimization problem of the lateral tracking controller is as follows:

[0240]

[0241] stΔu dyn,min ≤Δu dyn,t ≤Δu dyn,max

[0242] u dyn,min ≤AΔu dyn,t +u dyn,t ≤u dyn,max

[0243] y hc,min ≤y hc ≤y hc，max

[0244] y sc,min -ε≤y sc ≤y sc,max +ε

[0245] ε>0

[0246] The optimization problem of the longitudinal tracking controller is as follows:

[0247]

[0248] Stu min ≤u(t+i)≤u max

[0249] Δu min ≤Δu(t+i)≤Δu max

[0250] Solve the above two optimization problems and input the first control variable in the solution into the vehicle, so that the vehicle can be controlled to move along the planned trajectory.

[0251] Example 1

[0252] like Figure 4 As shown, the driving environment of the smart car in this embodiment is a straight road, there is an obstacle vehicle in front of the current road, and there is an obstacle vehicle behind the left road. In order to obtain environmental information in real time, the present invention builds a Carsim and Simulink joint simulation platform to verify the feasibility of the method.

[0253] Firstly, the effectiveness of the trajectory tracking module is verified. Figure 5 , Figure 6 shown.

[0254] Set the driving trajectories of the two obstacle vehicles. The initial position of the vehicle in the current lane is (50, -2), and it moves forward at a constant speed of 45 km / h; the initial position of the vehicle in the target lane is (10, 2), and it moves forward at a constant speed of 45 km / h; the expected speed of the vehicle is 60 km / h.

[0255] First, the environment information is obtained in real time. When the distance between the smart car and the vehicle in front is greater than the expected safety distance of the smart car, there is no intention to change lanes and the expected speed is maintained. The expected safety distance calculation method is as follows:

[0256]

[0257] When the smart car intends to change lanes, it will calculate whether the distance to the vehicle behind it in the target lane is greater than the emergency braking distance of the vehicle behind it. If the distance is greater than this, it means that the feasibility condition of lane change is met. The calculation method of emergency braking distance is as follows:

[0258]

[0259] After the feasibility conditions for lane change are met, the optimal sinusoidal lane change reference trajectory under the current state is calculated, and the objective function is as follows:

[0260] j=w s j s +w c j c +w e j e

[0261] After solving the final lane change time and longitudinal acceleration, we can bring them into the sinusoidal lane change reference trajectory model to obtain the optimal lane change reference trajectory. The calculation of the sinusoidal lane change reference trajectory is as follows:

[0262] The lateral acceleration at any time is:

[0263]

[0264] The lateral velocity at any time is:

[0265]

[0266] The lateral displacement model at any time is:

[0267]

[0268] The trajectory planning optimization problem is constructed based on the artificial potential field method and model predictive control, which is in the following form:

[0269]

[0270]

[0271] U min ≤U i ≤U max

[0272] y rr ≤y t ≤y rl

[0273] |u(t)|<μg

[0274] Solve this problem and fit the discrete trajectory through a fifth-order polynomial to obtain a smooth trajectory. The fitting equation is:

[0275] f(t)=a0+a1t+a2t 2 +a3t 3 +a4t 4 +a5t 5

[0276] The parameters are input into the trajectory tracking layer for easy tracking. In this embodiment, the lane change trajectory of the smart car is shown in Figure 7 and Figure 8 . In particular, Figure 8 The unfilled black boxes in the figure represent two obstacle vehicles. In order to distinguish different vehicles at the same time, the simulation results are presented with letters and numbers near the boxes. The letter O represents the ego vehicle (the controlled vehicle), the letter L represents the vehicle in the left lane, and the letter R represents the vehicle in the right lane.

[0277] The above description is only a specific implementation of the present invention, but the protection scope of the present invention is not limited thereto. Any person familiar with the technology can understand and think of any changes or substitutions within the technical scope disclosed by the present invention, which should be included in the scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. A lane change trajectory planning method for an intelligent vehicle based on model predictive control, characterized in that: The following steps are involved: Step 1: Obtain environmental perception information based on laser radar, GPS sensor and vehicle speed sensor. The environmental perception information includes the status information of the smart car and surrounding dynamic obstacles and the information of the current road of the smart car. The status information of the smart car and surrounding dynamic obstacles is input into the emergency braking distance model and the expected safety distance model, and the behavior decision information is output according to the decision scheme. Step 2: Construct the corresponding objective function of trajectory planning based on improved artificial potential field method and model predictive control according to the environmental perception information and behavior decision information, and construct the optimization problem of trajectory planning module in combination with constraint conditions; Step 3: Select the optimal sinusoidal lane change reference trajectory under current conditions according to the lane change trajectory evaluation function to ensure the safety, efficiency and comfort of the lane change process.

2. According to claim 1, a smart vehicle lane change trajectory planning method based on model predictive control is characterized in that: In step 1, the emergency braking distance model is: Where S is the emergency braking distance of the vehicle in the current state, v0 is the speed of the vehicle before emergency braking, t1 is the time for the deceleration to increase linearly, S ε is the correction distance, b max is the maximum deceleration and b max =μg, μ is the road adhesion coefficient, g is the acceleration due to gravity and g=9.8m / s 2 ; The expected safety distance model is: Where D fs is the expected safety distance under the current state of the vehicle, τ r is the sensor delay time, v M is the longitudinal velocity of the rear vehicle, v F is the longitudinal speed of the preceding vehicle, S M is the emergency braking distance of the rear vehicle, a F <is the acceleration of the vehicle in front.

3. According to claim 2, a smart vehicle lane change trajectory planning method based on model predictive control is characterized in that: In step 1, the specific method for detecting the generation of lane change intention and the feasibility of lane change is: (1) Generation of lane change intention - judging whether the longitudinal distance between the smart car and the vehicle ahead in the current lane is less than the expected safety distance of the smart car. If the longitudinal distance between the smart car and the vehicle ahead in the current lane is less than the expected safety distance of the smart car, it means that the safety of the smart car cannot be guaranteed when driving at the expected speed, so the lane change intention is generated; Otherwise, it means that the safety performance of the smart car is guaranteed when driving at the expected speed, so there is no intention to change lanes; (2) Lane change feasibility condition: The longitudinal distance between the smart car and the vehicle behind the target lane is used to determine whether the lane change feasibility condition is met. When the longitudinal distance between the smart car and the vehicle behind the target lane is less than the emergency braking distance of the vehicle behind the target lane, it means that the lane change feasibility condition is not met. Otherwise, it means that the lane change feasibility condition is met.

4. According to claim 3, a smart vehicle lane change trajectory planning method based on model predictive control is characterized in that: In step 1, a method for constructing intelligent vehicle behavior decision planning according to lane change intention and lane change feasibility is specifically implemented as follows: Step 1.1, determine whether there is an intention to change lanes, and if so, further determine whether the feasibility conditions for lane change are met; if there is no intention to change lanes, maintain the desired speed; Step 1.2, judging whether the feasibility condition of lane change is met, if the feasibility condition of lane change is met, executing lane change; if not, slowing down and following the vehicle ahead; Step 1.3: During the lane change process, the longitudinal distance between the smart car and the vehicle behind the target lane is detected in real time, and whether the lane change safety condition is met is determined based on the detection result. If the lane change process safety condition is not met, the lane change is canceled and the vehicle continues to slow down and follow the vehicle ahead; Step 1.4: Output the behavior decision information to the trajectory planning module.

5. The method for planning lane change trajectory of an intelligent vehicle based on model predictive control according to claim 4, characterized in that: In step 2, the improved artificial potential field method includes a road potential field function and an obstacle potential field function, and the calculation method thereof is as follows: (1) The calculation method of the road potential field function is: OR road =k r (yy l ) 2 (yy r ) 2 Where U road is the road potential field function, k r is the adjustment coefficient of the road potential field, y is the lateral position of the vehicle, and y l ,y r They are the left and right centerline positions of a two-lane road, respectively; (2) The calculation method of the obstacle potential field function is: U obs =w1k o f1(x|μ1,∑1)+w2k o f2(x|μ2,∑2) Where μ1=(x1,y1) T ,μ2=(x2,y2) T Among them, U obs is the obstacle potential field function, w1 and w2 are weights, and w1+w2=1, x is the position information of the vehicle and x=(x av ,y av ) T , k o is the comprehensive coefficient to measure the impact of obstacles. f1 and f2 are two-dimensional joint probability density functions, μ1 and μ2 are means, x1, y1, ∑1, To calculate the parameters related to the mean and covariance of f1, x2, y2, ∑2, Parameters related to calculating the mean and covariance of f2.

6. The method for planning lane change trajectory of an intelligent vehicle based on model predictive control according to claim 5, characterized in that: In step 2, the optimization problem of the trajectory planning module is specifically expressed as: IN min ≤U i ≤U max and rr ≤y t ≤y rl |u(t)|<μg Where N p is the prediction time domain, N c is the control time domain, η((t+i|t) is the prediction value in the prediction time domain, η ref (t+i|t) is the reference quantity in the prediction time domain, t is the current sampling time, i is the i-th time in the current prediction time domain, U i is the control quantity, U all is the total function of the constructed artificial potential field, Q and R are weight matrices, is the derivative of the state quantity in the vehicle point mass model, ξ(t) is the state quantity in the vehicle point mass model, u(t) is the control quantity in the vehicle point mass model, U min , U max are the upper and lower bounds of the controlled quantity, y t is the lateral position of the vehicle at time t, y rr ,y rl are lane boundaries, and |u(t)|<μg is the constraint condition of the control quantity.

7. The method for planning lane change trajectory of an intelligent vehicle based on model predictive control according to claim 6, characterized in that: In step 3, the calculation formula of the sinusoidal lane change reference trajectory is as follows: The lateral acceleration at any time t is: The lateral velocity at any time t is: The lateral displacement model at any time t is: Where, t s is the lane change start time, t e is the lane change end time, and W is the lane width.

8. The method for planning lane change trajectory of an intelligent vehicle based on model predictive control according to claim 7, characterized in that: In step 3, the method for solving the optimal sinusoidal lane change reference trajectory is as follows: Step 3.1, based on the state information of surrounding vehicles obtained in real time, assuming that the acceleration of surrounding vehicles remains unchanged during the lane change process, predict their motion state during the lane change time; Step 3.2: Sample the lane change time interval ΔT in [2s, 6s] and 2 ,3m / s 2 ] to sample the longitudinal acceleration interval Δa, and any combination of lane change time and longitudinal acceleration can determine a lane change reference trajectory; Step 3.3, evaluate the lane change trajectory, consider the safety, riding comfort and efficiency of the smart car's lane change process, linearly weight the three indicators to obtain the lane change reference trajectory evaluation function, and determine the optimal lane change time and longitudinal acceleration by finding the maximum value of the function.

9. The method for planning lane change trajectory of an intelligent vehicle based on model predictive control according to claim 8, characterized in that: The evaluation function for measuring the safety, efficiency and comfort of the lane change reference trajectory and the final linear weighted objective function are: (1) Lane-changing process safety evaluation function Where N is the total number of obstacle vehicles, k s is the safety factor, x(t) and y(t) are the longitudinal and lateral positions of the smart car at time t, respectively. are the longitudinal position and lateral position of the i-th obstacle vehicle at time t respectively; (2) Lane-changing process efficiency evaluation function In the formula, k et , k ev are the efficiency coefficients related to time and speed, v x (t), v y (t) are the longitudinal speed and lateral speed of the smart car at time t, respectively, and T is the lane change time of the current sampling; (3) Comfort evaluation function In the formula, k cx , k cy are the longitudinal and lateral comfort coefficients, a x (t), a y (t) are the longitudinal acceleration and lateral acceleration of the smart car at time t; The comprehensive evaluation function is obtained by linearly weighting the safety, comfort and efficiency evaluation functions, as shown below: j=w s j s +w c j c +w e j e In the formula, w s 、w c 、w e are the weight coefficients of safety, comfort and efficiency respectively.

10. The method for lane change trajectory planning of an intelligent vehicle based on model predictive control according to claim 5, characterized in that: The mean and covariance matrix of f1 and f2 are defined as: In the formula, (x obs ,y obs ) is the position coordinate of the obstacle vehicle, (v x_obs ,v y_obs ) are the longitudinal and lateral velocities of the obstacle vehicle, (a x_obs ,a y_obs ) is the longitudinal acceleration and lateral acceleration of the obstacle vehicle, S x To measure the longitudinal safety distance, S y To measure the horizontal safety distance, S o is the minimum longitudinal distance determined by the vehicle size, D o is the minimum lateral distance determined by the vehicle size, Δx and Δy represent the adjustable distances in the longitudinal and lateral directions respectively, and T0 is the time interval; k m is a positive parameter related to the vehicle weight and is defined as: m is the mass of the obstacle vehicle.

Images