Intelligent driving obstacle avoidance trajectory planning method based on Internet of Vehicles

Abstract

The invention discloses an intelligent driving obstacle avoidance trajectory planning method based on the Internet of Vehicles, and relates to the technical field of intelligent driving, and the method comprises the steps: obtaining the vehicle information of a surrounding environment through a method of combining the Internet of Vehicles and a vehicle-mounted sensor, and selecting a threatened vehicle; carrying out data fusion on the related information of the threat vehicle through an LSTM network, and carrying out surrounding environment vehicle trajectory prediction on the fused data through a Transform network; the predicted surrounding environment vehicle track is sent to the planning layer, an S-L diagram and an S-T diagram are constructed through a path planning and speed planning decoupling method, and speed obstacle avoidance on the S-T diagram is completed through dynamic planning and quadratic planning. Therefore, by the adoption of the intelligent driving obstacle avoidance trajectory planning method based on the Internet of Vehicles, sensing information and a planning layer can be combined, the obstacle avoidance trajectory planning effect is improved, and a reasonable obstacle avoidance curve is provided for control.

Description

Technical Field

[0001] The present invention relates to the field of intelligent driving technology, and in particular to an intelligent driving obstacle avoidance trajectory planning method based on vehicle networking. Background Art

[0002] As cars become increasingly intelligent, the importance of intelligent driving becomes more and more apparent. During the autonomous driving process, vehicle trajectory prediction is a prerequisite for autonomous driving obstacle avoidance. The prediction of the future status and driving path of surrounding vehicles can greatly improve the obstacle avoidance capability of intelligent driving vehicles.

[0003] The existing autonomous driving obstacle avoidance technology has major problems such as limited information acquisition, inaccurate threat judgment, insufficient trajectory prediction accuracy, and lack of decoupling between path and speed planning. Therefore, it is necessary to propose an intelligent driving obstacle avoidance trajectory planning method based on the Internet of Vehicles. By combining the Internet of Vehicles with on-board sensors, the surrounding vehicle information can be fully obtained, and various information can be fully used to predict the trajectory of threatening vehicles, thereby improving the accuracy of the prediction. At the same time, a decoupled approach is adopted for obstacle avoidance path planning. By constructing the SL graph and the ST graph, the path planning and speed planning are independently optimized, thereby generating an efficient and reasonable obstacle avoidance trajectory, achieving effective obstacle avoidance in a dynamic environment, and providing a safer and more comfortable driving experience for the intelligent driving system. Summary of the invention

[0004] The purpose of the present invention is to provide an intelligent driving obstacle avoidance trajectory planning method based on vehicle networking, so as to improve the obstacle avoidance capability of intelligent vehicles and reduce the probability of collision of intelligent vehicles under vehicle networking conditions.

[0005] To achieve the above object, the present invention provides a method for intelligent driving obstacle avoidance trajectory planning based on Internet of Vehicles, comprising the following steps:

[0006] S1. Obtain information about surrounding vehicles through the Internet of Vehicles and on-board sensors;

[0007] S2: The acquired surrounding vehicle information is feature-fused through the LSTM network, and the fused features are passed to the Transformer to predict the trajectory of the surrounding vehicles, and then the predicted trajectory of the surrounding vehicles is sent to the planning layer;

[0008] S3. According to the predicted trajectory of the surrounding vehicles, the trajectory of the vehicle is planned by using dynamic programming and quadratic programming through path-speed planning decoupling, thus completing the trajectory obstacle avoidance operation of the intelligent driving vehicle.

[0009] Preferably, the acquired vehicle information includes the position, speed, and acceleration of surrounding threatening vehicles.

[0010] Preferably, the method comprises obtaining trajectory information including speed through lateral path planning and longitudinal speed planning;

[0011] Among them, lateral path planning is used to avoid static obstacles, including evenly spreading points in the driving range, connecting points with quintic polynomials, and optimizing the path through dynamic programming and quadratic programming;

[0012] Longitudinal speed planning is used to avoid surrounding vehicles. It includes constructing an ST map and projecting the surrounding vehicle predictions. Based on the predicted trajectory of the surrounding vehicles and the planned trajectory of the vehicle, it is determined whether to perform obstacle avoidance operations on the ST map.

[0013] Preferably, performing an obstacle avoidance operation on the ST graph includes: first, evenly scattering points on the ST graph; then, performing dynamic planning through obstacle distance cost, speed cost and destination cost;

[0014] Among them, the obstacle cost is used to measure the distance between the trajectory and the obstacle, expressed as:

[0015]

[0016] In the formula, Cost obs is the obstacle cost, s i is the longitudinal displacement, s obs is the minimum longitudinal displacement of the obstacle, s safe is the obstacle safety threshold, g is a piecewise function;

[0017] The speed cost is used for the smoothness and comfort of the lateral trajectory and is expressed as:

[0018]

[0019] In the formula, Cost speed is the longitudinal velocity penalty, v ref For road speed limit, Cost a Cost is the acceleration cost. jerk is the acceleration cost, W speed , W a , W jerk are the weight coefficients of velocity, acceleration and jerk cost terms respectively, is the longitudinal velocity of the sampling point;

[0020] The destination cost is used to measure the distance to the end of the path and is expressed as:

[0021] Cost des =W des (s n -s i ) 2 ;

[0022] In the formula, Cost des is the destination cost, W des is the weight coefficient of the destination cost, s n is the longitudinal displacement value of the path end point.

[0023] Preferably, solving the dynamic programming includes optimizing the target variable using position cost, reference speed cost and comfort cost function;

[0024] Among them, the position cost is the deviation between the current position of the vehicle and the optimal position in the planned trajectory, and the expression is:

[0025] Cost station =W station ∑(s i -s ref ) 2 ;

[0026] In the formula, Cost station is the location cost, W station The weight coefficient of the position cost, s ref is the longitudinal displacement of the initial path;

[0027] The reference speed cost is the deviation between the current vehicle speed and the reference speed calculated by speed planning, and the expression is:

[0028]

[0029] In the formula, Cost velocity is the reference speed cost, W velocity is the weight coefficient of the reference speed cost;

[0030] The comfort cost is used to control the vehicle acceleration and the rate of change of vehicle acceleration, expressed as:

[0031]

[0032] In the formula, Cost conf At the expense of comfort, W a , W jerk are respectively the acceleration weight coefficient and the acceleration change rate weight coefficient, is the longitudinal acceleration of the sampling point, and ΔT is the sampling interval in the time T direction.

[0033] Preferably, the solution of the dynamic programming requires constraints on the quadratic programming problem, including road boundary constraints, kinematic constraints, and starting point constraints;

[0034] Road boundary constraints are used to ensure that the vehicle path is within the road boundary conditions during planning and cannot exceed the road speed limit. The expression is:

[0035]

[0036] In the formula, s mini is the lower boundary of the road, s maxi is the boundary on the road, v limit Set speed limits for roads;

[0037] Kinematic constraints are used to ensure that the planned speed, acceleration, and acceleration rate of the vehicle meet the vehicle dynamics constraints. The expression is:

[0038]

[0039] In the formula, v min 、v max is the maximum speed limit and minimum speed limit of the vehicle, a min 、a max is the maximum acceleration limit and minimum acceleration limit of the vehicle, jerk min , jerk max The limit value of the acceleration change rate allowed for the vehicle;

[0040] The starting point constraint is used to ensure that the starting point of the initial trajectory does not change. The expression is:

[0041]

[0042] Where s begin , is the starting point state of the initial trajectory.

[0043] Therefore, the present invention adopts the above-mentioned intelligent driving obstacle avoidance trajectory planning method based on the Internet of Vehicles, which has the following technical effects:

[0044] (1) Combining the LSTM network and the Transformer neural network not only takes advantage of the LSTM's ability to process temporal features, but also gives full play to the benefits of the Transformer's multi-head attention mechanism. This allows for better use of the various information obtained and improves the accuracy of vehicle trajectory prediction in the surrounding environment.

[0045] (2) By decoupling the path-speed planning and combining dynamic programming with quadratic programming, a trajectory that avoids possible collisions in the future can be planned by projecting the predicted trajectory onto the ST graph, thereby improving the intelligent vehicle's obstacle avoidance capability and reducing the probability of collisions among intelligent vehicles under the conditions of the Internet of Vehicles.

[0046] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 It is a flow chart of an intelligent driving obstacle avoidance trajectory planning method based on the Internet of Vehicles;

[0048] Figure 2 It is a schematic diagram of a surrounding vehicle prediction network in an embodiment of an intelligent driving obstacle avoidance trajectory planning method based on vehicle networking;

[0049] Figure 3 It is a schematic diagram of obstacle avoidance trajectory planning SL in an embodiment of an intelligent driving obstacle avoidance trajectory planning method based on vehicle networking, wherein the ordinate axis L is a tangent direction perpendicular to the centerline of the lane, and the abscissa S is a direction along the centerline of the lane;

[0050] Figure 4 The present invention is a schematic diagram of obstacle avoidance trajectory planning ST in an embodiment of an intelligent driving obstacle avoidance trajectory planning method based on vehicle networking. DETAILED DESCRIPTION

[0051] The present invention can be explained in more detail by the following examples. The purpose of disclosing the present invention is to protect all changes and improvements within the scope of the present invention. The present invention is not limited to the following examples.

[0052] like Figure 1 As shown, the present invention provides an intelligent driving obstacle avoidance trajectory planning method based on the Internet of Vehicles, which mainly predicts the vehicle trajectory of the surrounding environment through deep learning and neural networks, and uses a combination of different networks to improve the real-time performance of trajectory prediction and optimize the accuracy of vehicle trajectory prediction. The specific steps are as follows:

[0053] S1. Obtain information about surrounding vehicles through the Internet of Vehicles and on-board sensors. For example, obtain the location, speed, and acceleration of threatening vehicles around through the Internet of Vehicles, and obtain road information of the current driving environment through on-board sensors, cameras, lidar, GPS, and IMU.

[0054] S2. In order to fully exploit the time series characteristics of trajectory data, the long short-term memory network (LSTM) is used to fuse the features of trajectory data. Among them, LSTM is an improved recurrent neural network, which effectively alleviates the gradient vanishing and gradient exploding problems of traditional RNN when processing long sequence data through its gating mechanism.

[0055] like Figure 2 As shown in the figure, based on the information collected by the Internet of Vehicles and the on-board sensors, the surrounding vehicle features are used as input, such as the position coordinates (x i ,y i), speed characteristic v i , acceleration characteristics a i . The horizontal coordinate X of the surrounding vehicle is obtained i =[x 1 ,x 2 ,x 3 ,x 4 ,...,x t ], vertical coordinate Y i =[y 1 ,y 2 ,y 3 ,y 4 , ..., y t ], speed V i =[v 1 , v 2 , v 3 , v 4 , ..., v t ] and acceleration a i =[a 1 , a 2 , a 3 , a 4 , ..., a t ] is used as input, and the historical state of the surrounding vehicles is obtained through the LSTM network, and the time series data is feature fused. Among them, the LSTM network outputs the hidden state h of each time step i , these hidden states are used to perform feature fusion, so that the fused features contain the trajectory information of the historical time steps, providing good data support for subsequent trajectory prediction.

[0056] Afterwards, the fused features are used as the input of the Transformer neural network, and the long-term dependencies between feature variables are captured through the self-attention mechanism. The trajectory is then predicted through the network, and the vehicle's trajectory prediction results are output, thereby fully utilizing the advantages of LSTM in processing time series data.

[0057] The Transformer attention mechanism enables the network to give higher weights to key information like humans, and can efficiently extract effective information to a certain extent. In addition, the Transformer neural network consists of an encoder and a decoder: the encoder uses summation and normalization operations to improve the network operation efficiency. The summation operation uses residual connections to solve the gradient vanishing problem, and the normalization operation can promote network convergence and accelerate the network training process; the decoder uses a masked multi-head attention mechanism. The added mask operation can avoid information leakage of the training truth value during the training process, facilitating the implementation of the training process.

[0058] Combining the LSTM network and the Transformer neural network can speed up trajectory prediction, while better acquiring the temporal features and global features in the data, and improving the accuracy of predicted trajectories of surrounding vehicles.

[0059] S3. By decoupling the path-speed planning, the predicted trajectory of the surrounding vehicles is projected, and the corresponding objective function and its constraints are set in combination with dynamic programming and quadratic programming to realize the obstacle avoidance path planning of intelligent driving, as follows:

[0060] First, construct the ST graph, project the predicted trajectories of the surrounding vehicles on the ST graph, and perform obstacle avoidance on the ST graph through dynamic programming, as follows:

[0061] (1) For lateral path planning, by evenly scattering points in the driving range, dynamic programming opens up a convex space (solution space) for path planning, and forms a rough solution in the solution space. The scattered points are connected by a quintic polynomial to ensure the smoothness and comfort of the initial path of dynamic planning, thereby ensuring the quality of the subsequent constructed path space and facilitating the solution of the optimal path. The expression of the quintic polynomial is:

[0062] l=f(s)=a 0 +a 1 s+a 2 s 2 +a 3 s 3 +a 4 s 4 +a 5 s 5 ;

[0063] Where l is the tangent direction perpendicular to the reference line, and s is the direction along the reference line. The reference line is the center line of the lane.

[0064] Please refer to Figure 3 , we can establish the boundary conditions of the quintic polynomial, and solve the coefficient a of the quintic polynomial by the six boundary conditions. 0 、a 1 、a 2 、a 3 、a 4 、a 5 ,as follows:

[0065]

[0066] Among them, f′(s) and f″(s) are the first-order derivative and second-order derivative of l with respect to s, respectively, θ h is the vehicle heading angle.

[0067] After being connected by a quintic polynomial, a cost function is constructed to evaluate candidate paths based on road environment, vehicle safety and other prerequisites. The cost function includes a linear combination of smoothness cost, obstacle cost and lane departure cost.

[0068] Among them, the smoothness cost is used to evaluate the smoothness of the planned path. The closer it is to a straight line, the smoother it is. The expression is as follows:

[0069] Cost smooth =W 1 ∑f′(s i ) 2 +W 2 ∑f″(s i ) 2 +W 3 ∑f″′(s i ) 2 ;

[0070] In the formula, Cost smooth is the smoothness cost, f′(s i ) is the slope of the curve, f″(s i ) is the curvature of the curve, f″′(s i ) is the curvature derivative, W 1 , W 2 , W 3 is the corresponding weight value.

[0071] The obstacle cost is used to evaluate the distance between the path and the obstacle, expressed as:

[0072]

[0073] Where, d 1 d 2 Represents the threshold of obstacle distance, C is a monotonically decreasing function.

[0074] After obtaining a rough solution of the path planning through dynamic programming, although the sampling points are connected by a quintic polynomial, they still do not meet the planning requirements. Therefore, quadratic programming is needed to further optimize the curve of the dynamic planning, so as to plan a complete drivable trajectory and send it to the control layer. The details are as follows:

[0075] Optimization variables: lateral displacement, lateral velocity, and lateral acceleration of each sampling point in the path space, expressed as x l =[l 0 ,l 0 ′,l 0 ″,l 1 ,l 1 ′,l 1 ″...,l n ,l n ′,ln ″] T .

[0076] Optimization goals: The path should be as close to the reference line as possible to ensure that the vehicle drives in the middle of the lane as much as possible; the path should be as continuous, stable and smooth as possible; the path should be in the center of the path space as much as possible, that is, as far away from obstacles as possible.

[0077] In the quadratic programming problem of the optimal path, the cost function includes reference line cost, smoothing cost and obstacle distance cost.

[0078] The reference line cost is:

[0079] Cost ref =W ref ∑l i 2 ;

[0080] Obstacle cost:

[0081]

[0082] Smoothing cost:

[0083] Cost smooth =W dl Σl i ' 2 +W ddl ∑l i ″ 2 +W dddl ∑(l i ″ +1 -l i ″) 2 .

[0084] In summary, the total cost of path quadratic programming can be expressed as:

[0085] Cost QP =Cost ref +Cost smooth +Cost obs1 ;

[0086] (2) Longitudinal speed planning:

[0087] According to the predicted trajectory projection of the surrounding vehicles on the ST map, combined with the set collision threshold W t , to determine whether the predicted trajectory of the surrounding vehicles and the planned trajectory of the self-vehicle are likely to collide. When the minimum distance between the obstacle vehicle and the self-vehicle under the vehicle network and radar is less than W t If there is a possibility of collision, the obstacle vehicle is projected on the ST graph; otherwise, the vehicle can pass through the area smoothly according to the current trajectory, and there is no need to perform speed obstacle avoidance operations.

[0088] Please refer to Figure 4 , obstacle avoidance operations are performed on the ST graph, which is similar to lateral path planning, as follows:

[0089] First, the ST graph is evenly distributed, and dynamic planning is performed using obstacle distance cost, speed cost, and destination cost.

[0090] Among them, the dynamic obstacle cost is used to measure the distance between the trajectory and the dynamic obstacle (i.e. the predicted trajectory of the surrounding vehicles), which can be expressed as:

[0091]

[0092] In the formula, Cost obs2 is the obstacle cost, s i is the longitudinal displacement, s obs is the minimum s value of the obstacle, s safe is the obstacle safety threshold, and g is a piecewise function.

[0093] The speed cost, for the smoothness and comfort of the lateral trajectory, is expressed as:

[0094] Cost ve =Cost speed +Cost a +Cost jerk ;

[0095]

[0096] In the formula, Cost speed is the longitudinal velocity penalty, v ref For road speed limit, Cost a Cost is the acceleration cost. jerk is the acceleration cost, W speed , W a , W jerk are the weight coefficients of velocity, acceleration and jerk cost terms respectively, is the longitudinal speed.

[0097] The destination cost is used to measure the distance to the end of the path and is expressed as:

[0098] Cost des =W des (s n -s i ) 2 ;

[0099] In the formula, Cost des is the destination cost, W desis the weight coefficient of the destination cost, s n is the longitudinal displacement value of the path end point.

[0100] Then, the quadratic programming is used to further optimize the dynamic programming results of the longitudinal velocity. Specifically, the optimization variables are the longitudinal displacement s, the longitudinal velocity Longitudinal acceleration The expression is The optimization objectives are: keep the initial trajectory of dynamic planning as close as possible; drive at the expected speed as much as possible; ensure the smoothness and comfort of vehicle driving, and the acceleration changes as smoothly as possible.

[0101] At this time, the cost functions of the quadratic programming are position cost, reference speed cost and comfort cost.

[0102] The position cost is the deviation between the current position of the vehicle and the optimal position in the planned trajectory, and the expression is:

[0103] Cost station =W station ∑(s i -s ref ) 2 ;

[0104] In the formula, Cost station is the location cost, W station The weight coefficient of the position cost, s ref is the longitudinal displacement of the initial path;

[0105] The reference speed cost is the deviation between the current vehicle speed and the reference speed calculated by speed planning, which ensures that the vehicle speed is within the speed limit allowed by the road. The expression is:

[0106]

[0107] In the formula, Cost velocity is the reference speed cost, W velocity is the weight coefficient of the reference speed cost;

[0108] The comfort cost is used to control the vehicle acceleration and the rate of change of vehicle acceleration, so that the vehicle does not accelerate or decelerate suddenly, and is expressed as:

[0109]

[0110] In the formula, Cost conf At the expense of comfort, W a , W jerk are respectively the acceleration weight coefficient and the acceleration change rate weight coefficient, The longitudinal acceleration is, and ΔT is the sampling interval in the time T direction.

[0111] In addition, when using quadratic programming to optimize the results of dynamic programming, corresponding constraints need to be established, including road boundary constraints, kinematic constraints, and starting point constraints.

[0112] Road boundary constraints are used to ensure that the vehicle path is within the road boundary conditions during planning and cannot exceed the road speed limit. The expression is:

[0113]

[0114] In the formula, s mini is the lower boundary of the road, s maxi is the boundary on the road, v limit Set speed limits for roads;

[0115] Kinematic constraints are used to ensure that the planned speed, acceleration, and acceleration rate of the vehicle meet the vehicle dynamics constraints. The expression is:

[0116]

[0117] In the formula, v min 、v max is the maximum speed limit and minimum speed limit of the vehicle, a min 、a max is the maximum acceleration limit and minimum acceleration limit of the vehicle, jerk min , jerk max The limit value of the acceleration change rate allowed for the vehicle;

[0118] The starting point constraint is used to ensure that the starting point of the initial trajectory does not change. The expression is:

[0119]

[0120] Where s begin , is the starting point state of the initial trajectory.

[0121] In summary, a planned curve that can avoid obstacles can be obtained, and the position, speed, and acceleration information of the curve are sent to the control layer, so that the control layer can control the vehicle to track the curve and complete the entire obstacle avoidance operation.

[0122] Therefore, the present invention adopts the above-mentioned intelligent driving obstacle avoidance trajectory planning method based on the Internet of Vehicles, which can complete the obstacle avoidance operation of intelligent driving in the Internet of Vehicles and is suitable for intelligent driving under more complex environmental conditions.

[0123] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that they can still modify or replace the technical solution of the present invention with equivalents, and these modifications or equivalent replacements cannot cause the modified technical solution to deviate from the spirit and scope of the technical solution of the present invention.

Claims

1. A method for intelligent driving obstacle avoidance trajectory planning based on Internet of Vehicles, characterized in that: The following steps are involved: S1. Obtain information about surrounding vehicles through the Internet of Vehicles and on-board sensors; S2: The acquired surrounding vehicle information is feature-fused through the LSTM network, and the fused features are passed to the Transformer to predict the trajectory of the surrounding vehicles, and then the predicted trajectory of the surrounding vehicles is sent to the planning layer; S3. According to the predicted trajectory of the surrounding vehicles, the trajectory of the vehicle is planned by using dynamic programming and quadratic programming through path-speed planning decoupling, thus completing the trajectory obstacle avoidance operation of the intelligent driving vehicle.

2. The intelligent driving obstacle avoidance trajectory planning method based on the Internet of Vehicles according to claim 1 is characterized in that: The vehicle information obtained includes the location, speed, and acceleration of surrounding threatening vehicles.

3. The intelligent driving obstacle avoidance trajectory planning method based on the Internet of Vehicles according to claim 1 is characterized in that: Intelligent driving vehicle trajectory obstacle avoidance operation, including obtaining trajectory information including speed through lateral path planning and longitudinal speed planning; Among them, lateral path planning is used to avoid static obstacles, including evenly spreading points in the driving range, connecting points with quintic polynomials, and optimizing the path through dynamic programming and quadratic programming; Longitudinal speed planning is used to avoid surrounding vehicles. It includes constructing an ST map and projecting the surrounding vehicle predictions. Based on the predicted trajectory of the surrounding vehicles and the planned trajectory of the vehicle, it is determined whether to perform obstacle avoidance operations on the ST map.

4. The intelligent driving obstacle avoidance trajectory planning method based on the Internet of Vehicles according to claim 3 is characterized in that: Longitudinal speed planning includes: first, evenly scattering points on the ST graph, and dynamically programming through obstacle distance cost, speed cost and destination cost; then, optimizing and solving the dynamic programming results of speed by using quadratic programming; Among them, the obstacle cost is used to measure the distance between the trajectory and the obstacle, expressed as: In the formula, Cost obs is the obstacle cost, s i is the longitudinal displacement of the sampling point, s obs is the minimum longitudinal displacement of the obstacle, s safe is the obstacle safety threshold, g is a piecewise function; The speed cost is used for the smoothness and comfort of the lateral trajectory and is expressed as: In the formula, Cost speed is the longitudinal velocity penalty, v ref For road speed limit, Cost a Cost is the acceleration cost. jerk is the acceleration cost, W speed , W a , W jerk are the weight coefficients of velocity, acceleration and jerk cost terms respectively, is the longitudinal velocity of the sampling point; The destination cost is used to measure the distance to the end of the path and is expressed as: Cost des =W des (s n -s i ) 2 ; In the formula, Cost des is the destination cost, W des is the weight coefficient of the destination cost, s n is the longitudinal displacement value of the path end point.

5. The intelligent driving obstacle avoidance trajectory planning method based on the Internet of Vehicles according to claim 4 is characterized in that: The quadratic programming is used to optimize the dynamic programming results of speed, including optimizing the target variable using the position cost, reference speed cost and comfort cost functions; Among them, the optimization target variables include longitudinal displacement, longitudinal velocity, and longitudinal acceleration; The position cost is the deviation between the current position of the vehicle and the optimal position in the planned trajectory, and the expression is: Cost station =W station ∑(s i -s ref ) 2 ; In the formula, Cost station is the location cost, W station The weight coefficient of the position cost, s ref is the longitudinal displacement of the initial path; The reference speed cost is the deviation between the current vehicle speed and the reference speed calculated by speed planning, and the expression is: In the formula, Cost velocity is the reference speed cost, W velocity is the weight coefficient of the reference speed cost; The comfort cost is used to control the vehicle acceleration and the rate of change of vehicle acceleration, expressed as: In the formula, Cost conf At the expense of comfort, W a , W jerk are respectively the acceleration weight coefficient and the acceleration change rate weight coefficient, is the longitudinal acceleration of the sampling point, and ΔT is the sampling interval in the time T direction.

6. The intelligent driving obstacle avoidance trajectory planning method based on the Internet of Vehicles according to claim 1 is characterized in that: The constraints of quadratic programming include road boundary constraints, kinematic constraints, and starting point constraints; Road boundary constraints are used to ensure that the vehicle path is within the road boundary conditions during planning and cannot exceed the road speed limit. The expression is: In the formula, s mini is the lower boundary of the road, s maxi is the boundary on the road, v limit Set speed limits for roads; Kinematic constraints are used to ensure that the planned speed, acceleration, and acceleration rate of the vehicle meet the vehicle dynamics constraints. The expression is: In the formula, v min 、v max is the maximum speed limit and minimum speed limit of the vehicle, a min 、a max is the maximum acceleration limit and minimum acceleration limit of the vehicle, jerk min , jerk max The limit value of the acceleration change rate allowed for the vehicle; The starting point constraint is used to ensure that the starting point of the initial trajectory does not change. The expression is: Where s begin , is the starting point state of the initial trajectory.

Images