CN114261399B - Decision planning method for intelligent driving automobile under ice and snow road surface - Google Patents
Decision planning method for intelligent driving automobile under ice and snow road surface Download PDFInfo
- Publication number
- CN114261399B CN114261399B CN202111403108.1A CN202111403108A CN114261399B CN 114261399 B CN114261399 B CN 114261399B CN 202111403108 A CN202111403108 A CN 202111403108A CN 114261399 B CN114261399 B CN 114261399B
- Authority
- CN
- China
- Prior art keywords
- vehicle
- state
- ice
- planning
- snow
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 41
- 230000008859 change Effects 0.000 claims abstract description 44
- 238000002360 preparation method Methods 0.000 claims description 17
- 230000006870 function Effects 0.000 claims description 15
- 239000000203 mixture Substances 0.000 claims description 15
- 238000005070 sampling Methods 0.000 claims description 7
- 230000007704 transition Effects 0.000 claims description 7
- 238000009825 accumulation Methods 0.000 claims description 6
- 230000009471 action Effects 0.000 claims description 4
- 230000001960 triggered effect Effects 0.000 claims description 3
- 238000000926 separation method Methods 0.000 claims description 2
- 230000006872 improvement Effects 0.000 claims 1
- 230000008901 benefit Effects 0.000 description 6
- 230000001133 acceleration Effects 0.000 description 5
- 238000013461 design Methods 0.000 description 4
- 238000010586 diagram Methods 0.000 description 4
- 230000008569 process Effects 0.000 description 4
- 238000004088 simulation Methods 0.000 description 4
- 238000006073 displacement reaction Methods 0.000 description 3
- 206010039203 Road traffic accident Diseases 0.000 description 2
- 230000002411 adverse Effects 0.000 description 2
- 230000007547 defect Effects 0.000 description 2
- 238000001514 detection method Methods 0.000 description 2
- 239000011159 matrix material Substances 0.000 description 2
- 230000007246 mechanism Effects 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 230000001186 cumulative effect Effects 0.000 description 1
- 238000007405 data analysis Methods 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 230000007774 longterm Effects 0.000 description 1
- 230000003340 mental effect Effects 0.000 description 1
- 230000008447 perception Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000012216 screening Methods 0.000 description 1
- 238000010845 search algorithm Methods 0.000 description 1
- 230000011218 segmentation Effects 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 238000012546 transfer Methods 0.000 description 1
Landscapes
- Traffic Control Systems (AREA)
Abstract
The invention relates to an intelligent driving automobile decision planning method under an ice and snow road surface, which mainly comprises the following steps: step one: a decision method of a layered state machine is designed in an accumulated mode by considering the lane change safety distance and the dissatisfaction degree of a driver in an ice and snow environment; step two: improved mixing A based on ice and snow environment * Planning a planned expected path by a rule; step three: and optimizing the generated path by adopting a cubic spline interpolation method, and planning a smooth track.
Description
Technical Field
The invention relates to a humanized advanced driving support system (ADAS) design, in particular to an intelligent driving automobile decision planning method based on environment.
Background
The ice and snow environment exists in the northeast, northwest and other vast areas of China for a long time, and vehicles in cold areas exist in the ice and snow environment for a long time under special working conditions. In northeast, northwest and other areas, the winter is long, the air temperature is low, the snow fall is large, and the driving condition of an automobile driver is very poor before snow is removed. Some residual snow may also remain after the road is cleared of snow. Under the action of ice and snow environment, the traffic running environment is bad, the road adhesion coefficient is only 1/8-1/4 of that of a normal road, the driving comfort of a driver is poor, and the discomfort such as high mental stress and fatigue is easy to generate. Under the ice and snow environment, the traffic running efficiency and the running safety are lower than those of the non-ice and snow road surface, xing Enhui and the like obtain that the average running speed of the vehicle under the ice and snow environment is lower than that of the non-ice and snow road surface by analyzing the traffic data of the highway in Harbin city. These factors lead to a number of potential safety hazards, lin et al, through data analysis of traffic accidents in the United states 1967-2005, found that the accident rate in ice and snow environments was significantly higher than that in adverse weather such as rain, fog, etc., and these adverse driving factors also made the study of intelligent driving in ice and snow environments more significant.
The decision planning module in intelligent driving is one of indexes for measuring and evaluating the intelligent core of the intelligent driving automobile, and has the main task of analyzing the current environment after receiving various perception information of the sensor and then giving instructions to the bottom control module. The rule-based decision method is a classical approach, and is widely used in the last 2004 and later calendar DAPPA unmanned challenged games. A Talos decision system of the university of Massa staff adopts a serial state machine structure, the automatic driving system is divided into positioning and navigation, obstacle detection, lane line detection, road sign recognition, and module planning modules such as a drivable area map construction module, a motion planning module, a motion control module and the like adopt a rapid exploration random tree algorithm to generate a feasible path, and the system has the advantages of being capable of effectively aiming at a highly dynamic environment with poor prior information, and has the defects of adopting a serial structure and incomplete scene coverage. In order to overcome the defects, the Stanford Junior unmanned vehicle firstly proposes to adopt a parallel structure state machine to process each state in a modular manner, and the whole system can respond to input rapidly and flexibly. However, under complex conditions, the algorithm structure becomes huge due to the increase of traversal states, so that the division between states and the solution of state conflicts become difficult. While the parallel configuration is suitable for more complex operating conditions. Compared with a serial structure, the parallel structure has the advantages of wider scene traversal range, easier realization of complex function combination, better modularization and expandability and the like. The disadvantage is that the scene lacks depth and subtle changes in the environment can lead to decision errors. The learner combines the advantages of the serial structure and the parallel structure, chen Yongshang proposes a layered decision method for decoupling adherence to traffic regulations and avoidance of traffic participants, firstly selects actions which can be adopted according to the regulations to form a candidate behavior set, then synthesizes factors such as collision risks, traffic regulations and the like, adopts a hidden Markov model to infer the intention of a driver of the vehicle, the driving efficiency, a macroscopic recommended path and other evaluation methods to comprehensively decide an optimal behavior, and then performs path planning according to the behavior, a current area and a target area and gives a recommended vehicle speed.
The intelligent driving automobile planning method is mainly divided into a sampling-based planning and a search-based planning. The sampling-based method has the problem of high solution complexity and low sampling efficiency in some boundary scenes. Zhang Yu et al applied a hybrid a-algorithm and conjugate gradient descent path smoother for generating smooth paths by dynamically expanding four node patterns, introducing a kinematic model relative to a to make the resulting trajectory of each edge of the graph, i.e., the generation, smoother, with the disadvantage of lacking obstacle avoidance modules and not taking into account peri-vehicular obstacles. JorgeGodoy et al designed a local path planning module based on a path optimal decision method, and the module interacted with a global path planning module and had the functions of generating smooth fitting tracks, avoiding static and dynamic obstacles and the like, and had the disadvantage of too high time complexity.
Disclosure of Invention
In the conventional intelligent driving automobile decision planning research, a method aiming at a special working condition of an ice and snow mixed road surface rarely appears.
In order to achieve the above purpose, the present invention adopts the following technical scheme:
an intelligent driving automobile decision planning method under an ice and snow road surface mainly comprises the following steps:
step one: a decision framework of a layered state machine is designed in an accumulated mode by considering the lane change safety distance and the dissatisfaction degree of a driver in an ice and snow environment;
step two: improved mixing A based on ice and snow environment * The planning algorithm plans out the expected path(s),
step three: optimizing the planned expected path by adopting a cubic spline interpolation method so as to obtain a smooth track;
2. designing a decision architecture for a hierarchical state machine includes the following steps:
1) A state machine decision model is designed, the model is divided into two layers, and the upper layer is divided into two states, namely normal running and crossing. The lower layer is divided into three states, namely lane keeping, lane changing preparation, lane changing and stopping/waiting. The states of the two-layer state machine are stored in a key value pair, indicating the current state of the vehicle, the state set of the state machine is given by equation (1),
(normal travel, intersection, lane keeping, preparation for lane change, stop/wait } (1)
The state transition relation of the decision model is specifically as follows:
upper layer: the normal running state is converted into an intersection state when entering an intersection, and the intersection state is converted into the normal running state after exiting the intersection;
the lower layer: the stopping state is converted into a lane keeping state after starting, when dissatisfaction reaches a threshold value, the lane keeping state is converted into a lane change preparation state, the lane change preparation state is judged according to whether lane change is currently required, whether a safety distance is met or not and whether collision occurs or not, and when the condition is met, the lane change preparation state is converted into the lane change preparation state (including left lane change and right lane change); after the vehicle reaches a target lane or the track planning fails, the vehicle can not reach the target lane, the lane change state is converted into a lane keeping state, and the vehicle enters a stop/wait state when the vehicle reaches a stop condition brake in the lane keeping state;
2) Under the ice and snow environment, due to the fact that the ice and snow road surface adhesion condition is poor, factors such as low-speed front vehicles exist, the running continuity and speed of the vehicles cannot meet the expectations of drivers, dissatisfaction of the drivers can be gradually accumulated to generate overtaking intention, and a driver dissatisfaction accumulation formula (2) is obtained.
Wherein D (k) is the dissatisfaction at this time; d (k-1) is the dissatisfaction at the last time point; v pre Is the desired vehicle speed; v des For the current running speed, θ is the running continuity weighting coefficient, count avg For average number of stops per kilometer, T is the sampling time.
3) Whether a lane change or overtaking is possible or not is considered in view of safety of the lane change or overtaking, and depends on the front-back position between each traffic participant on adjacent lanes, namely, the lane change safety distance. Considering the safety braking distance under the ice and snow environment, the braking action distance and the braking driving distance can be divided, and the front workshop separation distance is as shown in formula (3).
Wherein s is a safe braking distance, s 0 A is the difference between the front and rear distances after braking bmax For maximum braking deceleration, s 0 T is the difference between the distance of the front car and the distance of the rear car 1 ,t 2 V is the speed of the vehicle for braking time.
3. Improved mixing A based on ice and snow environment * The planning algorithm comprises the following steps:
1) Under the mixed working condition of ice and snow road surface and normal road surface, A is as follows * Algorithmic 2D search expansion to 4D space [ x, y, θ, D]Where θ is the heading of the host vehicle, D indicates whether the vehicle is heading forward or heading backward; inputting position information and Zhou Che track prediction information of the ice and snow road surface into the planning model, and if the searched node is covered by ice and snow, the cost of the node is reduced; if mix A * The track between the searched node and the current node coincides with the Zhou Che predicted track, so that the accumulated cost on the track is reduced.
2) As shown in fig. 3, the distance between the own vehicle and the other vehicle is larger than the safety distance between the own vehicle and the front vehicle and meets the speed expectations of the own vehicle and the other vehicle, the lane change condition is triggered, and the own vehicle passes through the mixture a * The algorithm searches for two trajectories a, b. Since the trajectory node b passes through the ice and snow road surface, the overall cost becomes small, and the planning module outputs the trajectory a to the control module. As shown in fig. 4, the vehicle triggers the lane change condition to search out two tracks, a and b, and the tracks a and b overlap with the track predicted by the other vehicle on the right side of the vehicle in a certain period of time, that is, collide, so that both tracks a and b cannot be selected, the state machine module returns to the state of failed planning, and the state machine transitions to the lane keeping state.
4. Optimizing the generated path by adopting a cubic spline interpolation method comprises the following steps:
1) The path is divided into several parts, each of which is interpolated to obtain a different function, and the interpolated functions are used to represent the overall curve.
2) Constructing a third equation for each part, i.e. S i (x)＝a i +b i (x-x i )+c i (x-x i ) 2 +d i (x-x i ) 3 。
Wherein S is i (x) For piecewise interpolation functions (i=0, 1, …, n), a i ,b i ,c i ,d i To interpolate the unknowns over the interval, x i Is an interpolation point.
3) Let equation S i (x) Four interpolation conditions are satisfied. In condition one, the function must traverse all known nodes. In condition two, the 0 th derivative must be continuous at the node. Condition three, 1 st order succession at all nodes except the first and last nodes. Condition four except for the order 2 continuity at the first node and last node.
Compared with the prior art, the invention has the following advantages due to the adoption of the technical scheme:
1. the invention designs a layered state machine architecture, wherein the upper layer state and the lower layer state are decoupled, the upper layer is a scene, the lower layer is a sub-state in the scene, the upper layer state machine is used as a base class, the lower layer state machine is used as a derivative class, the framework has the advantage of convenient expansion, when a new scene needs to be dealt with or scene tasks need to be processed more carefully, the new state can be added in the original framework conveniently, the influence on the original various states is small, and the number of state transition conditions needing to be set is avoided by adopting the layered state machine.
2. In the layered state machine framework, constraints such as an dissatisfaction accumulated lane change mechanism and a driving safety distance in an ice and snow environment are introduced, so that the safety and the comfort of the decision planning model in the ice and snow environment driving are improved.
3. The invention improves the mixture A based on ice and snow environment * The planning algorithm is used for planning an expected path and optimizing the track by adopting a cubic spline interpolation method, and the algorithm introduces the information of the ice and snow road surface and Zhou Che track prediction information, considers the behavior of the cycle and the special working condition under the ice and snow mixed road surface, thereby planning a road surfaceThe bars are safe and relatively smooth.
Drawings
FIG. 1 is a schematic diagram of a hierarchical state machine decision model
FIG. 2 is a schematic diagram of an ice and snow hybrid road surface
FIG. 3 is a track planning scenario 1
FIG. 4 is a track planning scenario 2
FIG. 5 is a schematic diagram of a trajectory plan
FIG. 6 is a simulation environment of the setup
FIG. 7 is a diagram of simulator set parameters
FIG. 8 is a longitudinal displacement of the host vehicle and the lead vehicle during the following condition
FIG. 9 is a longitudinal displacement difference between the own vehicle and the front vehicle in the following condition
FIG. 10 is a graph of the velocity of the vehicle during the following condition
FIG. 11 is a graph of acceleration of the vehicle during a following condition
FIG. 12 is a lane change status value for an overtaking situation
FIG. 13 is a graph of the speed of a vehicle during an overrun condition
FIG. 14 is a graph of acceleration of a vehicle during an overtaking condition
Detailed Description
The following further describes the details of the present invention and its embodiments.
The invention relates to a decision planning method for an intelligent driving automobile under an ice and snow road surface, which mainly comprises the following steps:
step one: a decision method of a layered state machine is designed in an accumulated mode by considering the lane change safety distance and the dissatisfaction degree of a driver in an ice and snow environment; step two: improved mixing A based on ice and snow environment * Generating a path by a planning algorithm; step three: and optimizing the generated path by adopting a cubic spline interpolation method, and planning a smooth track.
1 rule-based State machine decision model establishment
The decision model of the invention adopts a layered state machine, wherein the upper layer is divided into two states, namely normal running and crossing, and the lower layer is divided into four states, namely lane keeping, lane changing preparation, lane changing and stopping/waiting, and the state machine model is shown in figure 1. The transition between states is determined by the safe distance between the own vehicle and the front vehicle in the ice and snow environment, the expected speed of the own vehicle on the ice and snow road surface and whether the reasonable track can be planned by the planning module.
{ normal Driving, crossing, lane keeps, preparation for lane change, stop/wait } (1)
The state transition relation of the decision model is specifically as follows:
upper layer: the normal running state is converted into an intersection state when entering an intersection, and the intersection state is converted into the normal running state after exiting the intersection;
the lower layer: the stopping state is converted into a lane keeping state after starting, when dissatisfaction reaches a threshold value, the lane keeping state is converted into a lane change preparation state, the lane change preparation state is judged according to whether lane change is currently required, whether a safety distance is met or not and whether collision occurs or not, and when the condition is met, the lane change preparation state is converted into the lane change preparation state (including left lane change and right lane change); after the vehicle reaches the target lane or the track planning fails, the vehicle can not reach the target lane, the lane change state is converted into the lane keeping state, and the vehicle enters the stop/wait state when the vehicle reaches the stop condition braking in the lane keeping state.
The hierarchical state machine architecture of the present invention decouples upper and lower layer states. The upper layer is a scene, and the lower layer is a sub-state in the scene; the upper layer state machine is used as a base class, and the lower layer state machine is used as a derivative class. The framework has the advantage of convenient expansion, when a new scene needs to be dealt with or a scene task needs to be processed more carefully, new states can be added conveniently in the original framework, the influence on various original states is small, and the number of state transition conditions needing to be set is avoided by adopting the hierarchical state machine. In the related layered state machine architecture, constraints such as an dissatisfaction accumulated lane change mechanism and a driving safety distance under the ice and snow environment are introduced, and the safety and the comfort of the decision planning model under the ice and snow environment driving are improved.
1.1 dissatisfaction accumulation
In consideration of the fact that under the ice and snow environment, due to the fact that the ice and snow road surface adhesion condition is poor, factors such as low-speed front vehicles exist, the running continuity and speed of the vehicles cannot meet the expectations of drivers, and when the vehicles run on the ice and snow road surface, the drivers do not have any intention of changing lanes due to sudden acceleration or deceleration. The speed of the vehicle is not too high due to poor adhesion conditions on ice and snow roads, so the vehicle will change lanes on ice and snow roads. Most of the problems are that the speed of the passengers of the rear vehicle cannot be met due to long-term low-speed running of the front vehicle. The cumulative dissatisfaction level of the vehicle speed is set in the task file of the upper layer or by other means, and the expected value of the vehicle speed in the driving process is set. But during actual driving the vehicle will follow the lead vehicle at a very slow speed. Ensuring the driving safety, but not reaching the expected driving speed. The degree of accumulation of vehicle speed dissatisfaction increases over time. The dissatisfaction of the driver is gradually accumulated to generate the intention to overtake, and the accumulated formula (2) of the dissatisfaction of the driver is obtained.
Wherein D (k) is dissatisfaction at the current time; d (k-1) is the dissatisfaction of the last moment; v pre Is the desired vehicle speed; v des For the current running speed, θ is the running continuity weighting coefficient, count avg For average number of stops per kilometer, T is the sampling period.
1.2 safety distance between own vehicle and front vehicle
The minimum safe distance of the vehicle following the front vehicle means that emergency stop occurs due to the sudden occurrence of the front vehicle when the vehicle is traveling on a road. The vehicle can be braked within an effective safe distance without causing traffic accidents. The distance traveled by the target vehicle from the vehicle in front of the lane should exceed a minimum safe distance to prevent danger. The minimum safe distance may be obtained from the stopping distance. According to the analysis of the braking performance of the automobile, the distance has a direct relation with the driving safety of the automobile. From the whole braking process, it comprises two parts: the travel distance of the brake application time and the travel distance of the continuous brake.
Wherein a is bmax For maximum braking deceleration, s 0 T is the difference between the distance of the front car and the distance of the rear car 1 ,t 2 V is the speed of the vehicle for braking time.
According to the safety distance between the vehicle and the front vehicle and the establishment of a vehicle speed expected formula, a transfer method of a state machine model can be obtained.
2 track planning model establishment
The idea of planning model building is to first use improved mixture A * The algorithm determines the future position of the vehicle and fits a curve to the controller by using a trajectory planning model of cubic spline interpolation.
2.1 mixing A * Planning method
Mix A * The algorithm is divided into two phases, the first phase being essentially for the traditional A * The algorithm is improved, unlike the hybrid A * Heuristic searches are performed under a continuous coordinate system and it can be ensured that the generated trajectory satisfies the vehicle non-integrity constraints, but the path is not necessarily globally optimal during the algorithm operation, but is nevertheless "near" the globally optimal solution. Traditional A * With mix A * Both algorithms are based on the grid world, A * Is to give each grid a corresponding penalty to the center points and the algorithm only accesses these center points, whereas mix a * Points satisfying the 3D continuous state of the vehicle are selected from the grids, and losses are assigned to the points. Different from graph search algorithm A * Mix A * Consider a kinematic model, will A * Algorithmic 2D search expansion to 4D space [ x, y, θ, D]Where θ is the heading of the host vehicle and D indicates whether the vehicle is heading forward or heading backward. In the figure, the searching method consists of six searching directions, and the radius of the circular arc is the minimum turning radius of the vehicle as shown in figure 5.
2.2 trajectory planning for specific conditions
The invention aims at ice and snow road surface and normalThe road surface mixing condition is as shown in fig. 2 for establishing a track planning model. Traditional mix A * The method does not consider the track of ice and snow road surface and Zhou Che, and the invention is applied to classical mixing A * The algorithm is improved, and the position information and Zhou Che track prediction information of the ice and snow road surface are input to the planning model. If the searched node is covered with ice and snow, the cost of the node is reduced. If mix A * The track between the searched node and the current node coincides with the Zhou Che predicted track, so that the accumulated cost on the track is reduced.
As shown in fig. 3, the distance between the own vehicle and the other vehicle is larger than the safety distance between the own vehicle and the front vehicle and meets the speed expectations of the own vehicle and the other vehicle, the lane change condition is triggered, and the own vehicle passes through the mixture a * The algorithm searches two tracks, namely track a and track b. Since the trajectory node b passes through the ice and snow road surface, the overall cost becomes small, and the planning module outputs the trajectory a to the control module. As shown in fig. 4, the self-vehicle triggers the channel change condition to search out two tracks, a and b, and the tracks a and b overlap with the track predicted by the other vehicle on the right side of the self-vehicle in a certain period of time, so that collision occurs, both tracks a and b cannot be selected, the state machine module returns a state of failed planning, and the state machine keeps following.
2.3 cubic spline sampling method
Due to the mixture A * The planned path is not smooth, partial path [ x, y]And outputting by adopting a track planning model based on a cubic spline interpolation method. And performing track planning on the intelligent driving automobile. And planning a lane change track according to the initial state and the target state of the vehicle, so that the vehicle reaches the target position at the designated time.
Optimizing the generated path by adopting a cubic spline interpolation method comprises the following steps:
1) The path is divided into several parts, each of which is interpolated to obtain a different function, and the interpolated functions are used to represent the overall curve.
2) A third equation is constructed for each part as in equation (4).
S i (x)＝a i +b i (x-x i )+c i (x-x i ) 2 +d i (x-x i ) 3 (4)
Wherein S is i (x) For piecewise interpolation functions (i=0, 1, …, n), a i ,b i ,c i ,d i To interpolate the unknowns over the interval, x i Is an interpolation point.
3) Let equation S i (x) Four interpolation conditions are satisfied. In condition one, the function must traverse all known nodes. In condition two, the 0 th derivative must be continuous at the node. Condition three, 1 st order succession at all nodes except the first and last nodes. Condition four except for the order 2 continuity at the first node and last node.
It is necessary to find 4n equations to solve for 4n unknowns, first of all because all points must satisfy the interpolation condition S (x i )＝y i (i=0, 1, …, n), except for two end points, each of all n-1 internal points satisfies S i (x i+1 )＝y i+1 The front and back two segmentation cubic equations have 2 (n-1) equations, and the two endpoints respectively meet the first and last cubic equations, so that the total number of the equations is 2 n; secondly the first derivative of n-1 interior points should be continuous, i.e. the end point of the i-th interval and the start point of the i+1-th interval are the same point, their first derivatives should also be equal, there are n-1 equations, and the second derivative of the interior points should also be continuous, i.e. there are now a total of 4n-2 equations, the other two equations being available by passing the boundary conditions.
The boundary conditions adopt natural boundary conditions, and the second derivative of the designated endpoint is 0, namely S "(x) 0 )＝S”(x n ) =0. The interpolation point is brought into an interpolation function to obtain a i ＝y i By h i ＝x i+1 -x i Represents the step size, by S i (x i+1 )＝y i+1 Can derived (5)
From S' i (x i+1 )＝S′ i+1 (x i+1 ) Equation (6) can be derived.
S′ i+1 (x i+1 )＝b i+1 +2c i (x i+1 -x i+1 )+3d i (x i+1 -x i+1 ) 2 ＝b i+1 (6)
By S' i (x i+1 )＝S″ i+1 (x i+1 ) Can push out 2c i +6h i d i ＝2c i+1 . Let m be i ＝S” i (x i )＝2c i Then equation (7) is derived.
Will a i ,c i ,d i Brought into formula (6), formula (8) can be obtained.
At natural boundary conditions, m 0 ＝0，m n =0. From equation (8), a linear system of equations with m as the unknowns can be derived as in equation (9).
The coefficient matrix on the left side is a strict diagonal dominant matrix, so that the linear equation set has a unique solution, and the equation set solution is brought into formula (4) to obtain:
the invention improves the mixture A based on ice and snow environment * The planning algorithm is used for planning an expected path and optimizing the track by adopting a cubic spline interpolation method, and the algorithm introduces the information of the ice and snow road surface and Zhou Che track prediction information, and considers the behavior of the cycle and the special characteristics under the ice and snow mixed road surfaceAnd the working condition is special, so that a safe and relatively smooth track is planned.
3 acquisition and screening of observed data
3.1 Driving simulation experiment design
The invention takes a simulated driver built on the basis of CarMaker vehicle dynamics simulation software as an experimental platform, designs a simulated driving experiment, adopts various working conditions to verify a decision planning algorithm, and collects driving behavior, vehicle state and road environment information. The main software and hardware components of the simulated driving platform are as follows:
the simulation driver is real, and the compass G27 kit comprises an accelerator pedal, a brake pedal, a clutch pedal, a gear handle, a steering wheel and other driving position structures which are the same as those of a real vehicle. Luo Ji G27 provides a force feedback technique for providing road feel information to the driver, and the magnitude of the feedback force scaling factor can also be set. The scene can be customized through the IPGRoad module in the CarMaker main interface, and the invention establishes a straight-going-rotary island-straight-going working condition through the IPGRoad. The vehicle parameters are set in CIT of the car maker software, and as shown in fig. 7, the vehicle-related parameters and the running strategy simulation progress can be controlled through the interface, the vehicle is selected as audior 8, the parameters such as tires are selected as default values, and the experimental setting is driven by the driver only, so that the load is set to 60kg here. The experimental lane is a three-lane urban road, ice and snow road surfaces are randomly inserted into the lanes, and the friction coefficient of the road surfaces is 0.28. The road conditions are shown in fig. 6, and the light gray road surface is an ice and snow covered road surface.
3.1.1 following Condition
The front vehicle is driven at 30km/h and set to cruise at a constant speed, the front vehicle is positioned in front of the host vehicle by 15m, the self vehicle is started from rest to follow the front vehicle, the expected speed is set to 30km/h, the distance between the self vehicle and the front vehicle is kept, the self vehicle accelerates from rest to the driving speed of 30km/h, and the speed of the self vehicle is controlled to be stable.
The decision of the own vehicle is a state machine decision, and the own vehicle keeps constant-speed cruising in the following process. Fig. 8 is a longitudinal displacement of the own vehicle and the front vehicle. The front vehicle accelerates from rest to traveling at a speed of 30km/h, and fig. 10 is a speed profile of the own vehicle.
From the formula (3), it can be calculated that the following distance is more than 20m to meet the safety requirement. As can be seen from fig. 9, the travel distance between the front vehicle and the own vehicle is reduced and then increased, and finally the heel is full of the safety distance. The acceleration is as shown in figure 11, the minimum is-3.9, and the maximum is 1.0, so that the comfort requirement and the safety requirement are met.
3.1.2 Overtaking conditions
Setting two front vehicles to cruise at a constant speed of 20km/h, wherein the distance between the two front vehicles is 50m, the distance between the own vehicle and the front vehicle is 50m, determining whether to select lane change by the own vehicle decision planning module, generating a lane change intention when the dissatisfaction accumulation reaches a set value, and carrying out lane change decision and path planning by the own vehicle. During driving, dissatisfaction of the front vehicle gradually accumulates to reach a threshold value, the state machine is switched from lane keeping to lane changing state as shown in fig. 12, the speed of the self vehicle is as shown in fig. 13, and the acceleration of the self vehicle is as shown in fig. 14. Minimum is-4.3, maximum is 2.0, and the comfort requirement and the safety requirement are met.
Claims (5)
1. An intelligent driving automobile decision planning method under an ice and snow road surface mainly comprises the following steps:
step one: a decision framework of a layered state machine is designed in an accumulated mode by considering the lane change safety distance and the dissatisfaction degree of a driver in an ice and snow environment;
step two: improved mixing A based on ice and snow environment * The planning algorithm plans out the expected path(s),
step three: optimizing the planned expected path by adopting a cubic spline interpolation method so as to obtain a smooth track;
the state machine decision model is divided into two layers, wherein the upper layer is divided into two states, namely normal running and crossing, and the lower layer is divided into four states, namely lane keeping, lane changing preparation, lane changing and stopping/waiting; the states of the two-layer state machine are stored in a key value pair, indicating the current state of the vehicle, the state set of the state machine is given by equation (1),
(normal travel, intersection, lane keeping, preparation for lane change, stop/wait } (1)
The state transition relation of the decision model is specifically as follows:
upper layer: the normal running state is converted into an intersection state when entering an intersection, and the intersection state is converted into the normal running state after exiting the intersection;
the lower layer: the stopping state is converted into a lane keeping state after starting, when dissatisfaction reaches a threshold value, the state is converted into a lane change preparation state, the lane change preparation state is judged according to whether lane change is needed, whether a safety distance is met or not and whether collision occurs at present, and when the condition is met, the state is converted into the lane change preparation state; after the vehicle reaches a target lane or the track planning fails, the vehicle can not reach the target lane, the lane change state is converted into a lane keeping state, and the vehicle enters a stop state when the vehicle reaches a stop condition for braking in the lane keeping state;
improved mixing A based on ice and snow environment in step two * The planning algorithm is that under the mixed working condition of ice and snow road surface and normal road surface, A is carried out * Algorithmic 2D search expansion to 4D space [ x, y, θ, D]Where θ is the heading of the host vehicle, D indicates whether the vehicle is heading forward or heading backward; inputting position information and Zhou Che track prediction information of the ice and snow road surface into the planning model, and if the searched node is covered by ice and snow, the cost of the node is reduced; if mix A * The track between the searched node and the current node coincides with the Zhou Che predicted track, so that the accumulated cost on the track is reduced.
2. The decision planning method for intelligent driving of a car under an icy or snowy road surface according to claim 1, wherein the driver dissatisfaction accumulation method is specifically as follows:
driver dissatisfaction accumulation formula (2);
wherein D (k) is the dissatisfaction at this time; d (k-1) is the dissatisfaction at the last time point; v pre Is the desired vehicle speed; v des For the current running speed, θ is the running continuity weighting coefficient, count avg For average number of stops per kilometer, T is the sampling time.
3. The decision planning method for intelligent driving of vehicles under ice and snow road surface according to claim 1, wherein the lane changing safety distance can be divided into a braking action distance and a braking driving distance, a front workshop separation distance as shown in formula (3),
wherein s is a safe braking distance, s 0 A is the difference between the front and rear distances after braking bmax For maximum braking deceleration, s 0 T is the difference between the distance of the front car and the distance of the rear car 1 ,t 2 V is the speed of the vehicle for braking time.
4. The decision-making planning method for intelligent driving automobiles under ice and snow roads according to claim 1, wherein step two is based on ice and snow environment improvement mixing a * The specific method for planning the expected path by the planning algorithm is as follows:
the distance between the own vehicle and the other vehicle is larger than the safety distance between the own vehicle and the front vehicle, the speed expectations of the own vehicle and the other vehicle are met, the lane changing condition is triggered, and the own vehicle passes through the mixture A * The algorithm searches out two tracks a, b; because the track b node passes through the ice and snow road surface, the total cost is reduced, and the planning module outputs an a track to the control module; the self-vehicle triggers the lane change condition to search out two tracks, a and b, and the tracks a and b are overlapped with the tracks predicted by the other vehicle on the right side of the self-vehicle in a certain time period, namely collision occurs, so that the two tracks a and b cannot be selected, a planning failure state can be returned to the state machine module, and the state machine is transferred to a lane keeping state.
5. The decision planning method for intelligent driving of a car under an icy or snowy road surface according to claim 1, wherein the following steps: optimizing the planned expected path by adopting a cubic spline interpolation method, thereby obtaining a smooth track, which comprises the following steps:
1) Dividing the path into a plurality of parts, interpolating each small part to obtain different functions, and representing an integral curve by using the interpolated functions;
2) Constructing a third equation for each part, i.e. S i (x)＝a i +b i (x-x i )+c i (x-x i ) 2 +d i (x-x i ) 3 ；
Wherein S is i (x) For piecewise interpolation functions (i=0, 1, …, n), a i ,b i ,c i ,d i To interpolate the unknowns over the interval, x i Is an interpolation point;
3) Let equation S i (x) Four interpolation conditions are satisfied; in condition one, the function must traverse all known nodes; condition two, the 0 th derivative must be continuous at the node; condition three, 1 st order continuity at all nodes except the first node and the last node; condition four except for the order 2 continuity at the first node and last node.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111403108.1A CN114261399B (en) | 2021-11-24 | 2021-11-24 | Decision planning method for intelligent driving automobile under ice and snow road surface |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111403108.1A CN114261399B (en) | 2021-11-24 | 2021-11-24 | Decision planning method for intelligent driving automobile under ice and snow road surface |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114261399A CN114261399A (en) | 2022-04-01 |
| CN114261399B true CN114261399B (en) | 2023-11-07 |
Family
ID=80825479
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111403108.1A Active CN114261399B (en) | 2021-11-24 | 2021-11-24 | Decision planning method for intelligent driving automobile under ice and snow road surface |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114261399B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116620287B (en) * | 2023-06-30 | 2024-03-19 | 南京项尚车联网技术有限公司 | Intelligent driving method and system |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101297337A (en) * | 2005-09-29 | 2008-10-29 | 微软公司 | Methods for predicting destinations from partial trajectories employing open-and closed-world modeling methods |
| CN108454623A (en) * | 2018-01-22 | 2018-08-28 | 大连理工大学 | A kind of unmanned electric vehicle Trajectory Tracking Control method of four motorized wheels |
| CN110304074A (en) * | 2019-06-28 | 2019-10-08 | 电子科技大学 | A kind of hybrid type driving method based on stratification state machine |
| CN111890951A (en) * | 2020-08-07 | 2020-11-06 | 吉林大学 | Intelligent electric automobile trajectory tracking and motion control method |
| WO2021102955A1 (en) * | 2019-11-29 | 2021-06-03 | 华为技术有限公司 | Path planning method for vehicle and path planning apparatus for vehicle |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US10503172B2 (en) * | 2017-10-18 | 2019-12-10 | Luminar Technologies, Inc. | Controlling an autonomous vehicle based on independent driving decisions |
| US20200406928A1 (en) * | 2019-06-28 | 2020-12-31 | Robert Bosch Gmbh | Method for controlling a vehicle |
-
2021
- 2021-11-24 CN CN202111403108.1A patent/CN114261399B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101297337A (en) * | 2005-09-29 | 2008-10-29 | 微软公司 | Methods for predicting destinations from partial trajectories employing open-and closed-world modeling methods |
| CN108454623A (en) * | 2018-01-22 | 2018-08-28 | 大连理工大学 | A kind of unmanned electric vehicle Trajectory Tracking Control method of four motorized wheels |
| CN110304074A (en) * | 2019-06-28 | 2019-10-08 | 电子科技大学 | A kind of hybrid type driving method based on stratification state machine |
| WO2021102955A1 (en) * | 2019-11-29 | 2021-06-03 | 华为技术有限公司 | Path planning method for vehicle and path planning apparatus for vehicle |
| CN111890951A (en) * | 2020-08-07 | 2020-11-06 | 吉林大学 | Intelligent electric automobile trajectory tracking and motion control method |
Non-Patent Citations (1)
| Title |
|---|
| 无人驾驶车辆行为决策系统研究;熊璐;康宇宸;张培志;朱辰宇;余卓平;;汽车技术(第08期);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114261399A (en) | 2022-04-01 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| WO2021136130A1 (en) | Trajectory planning method and apparatus | |
| JP6451111B2 (en) | Driving support device and driving support method | |
| JP3714258B2 (en) | Recommended operation amount generator for vehicles | |
| US11370435B2 (en) | Connected and automated vehicles, driving systems, and control logic for info-rich eco-autonomous driving | |
| CN111867911A (en) | Vehicle control method and apparatus | |
| CN110304074B (en) | Hybrid driving method based on layered state machine | |
| US20190286151A1 (en) | Automated driving systems and control logic for cloud-based scenario planning of autonomous vehicles | |
| Bertolazzi et al. | Supporting drivers in keeping safe speed and safe distance: the SASPENCE subproject within the European framework programme 6 integrating project PReVENT | |
| WO2017159539A1 (en) | Vehicle control apparatus, vehicle control method and vehicle control program | |
| KR101601889B1 (en) | Method and module for controlling a vehicle's speed based on rules and/or costs | |
| KR20190030199A (en) | Supervision of vehicles | |
| CN107813820A (en) | A kind of unmanned vehicle lane-change paths planning method for imitating outstanding driver | |
| CN113012448B (en) | Intelligent driving system for vehicle distribution formation for solving intersection traffic jam | |
| CN113297721B (en) | Simulation method and device for signal intersection vehicle selective exit road | |
| Sakhdari et al. | Ecological adaptive cruise control of a plug-in hybrid electric vehicle for urban driving | |
| US20230019462A1 (en) | MPC-Based Trajectory Tracking of a First Vehicle Using Trajectory Information on a Second Vehicle | |
| CN114261399B (en) | Decision planning method for intelligent driving automobile under ice and snow road surface | |
| Zhao et al. | “InfoRich” Eco-Driving Control Strategy for Connected and Automated Vehicles | |
| Aoki et al. | Multicruise: eco-lane selection strategy with eco-cruise control for connected and automated vehicles | |
| CN115050183A (en) | Method for generating simulated traffic flow | |
| CN110426215B (en) | Model establishing method for vehicle ride comfort test and intelligent driving system | |
| CN117315970A (en) | Lane change of autonomous vehicles involving traffic congestion at an intersection | |
| Gao et al. | Predictive cruise control under cloud control system for urban bus considering queue dissipation time | |
| GB2560487A (en) | Vehicle route guidance | |
| Zhao et al. | Cellular automaton model for three-lane urban road considering Internet of Vehicles lane |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |