CN117782126A - High-precision map-guided autonomous driving path planning and decision-making method - Google Patents

Abstract

The high-precision map-guided autonomous driving path planning and decision-making method of this application is based on the large, rich, and accurate road static information provided by high-precision maps, based on the navigation application mode, and combines path planning and decision-making based on the perception system and simple element maps. Method, a path planning decision-making system framework, modules and methods suitable for smart cars are proposed, and improvements in three aspects are focused on: one is the lane track level path planning method guided by high-precision maps, and the other is the high-precision map-guided path planning method. The driving behavior decision-making method, the third is the local path planning and selection method guided by high-precision maps, integrates three different indicators that affect driving curve selection, including decision-making cost, road cost and comfort cost, through cost calculation, and replaces them through weighted summation Select the trajectory line with the lowest price. Through the above three improvements, the reliability, safety, intelligence and economy of autonomous driving have been greatly improved.

Description

High-precision map-guided autonomous driving path planning and decision-making method

Technical field

This application relates to a vehicle-assisted driving path planning and decision-making system, and in particular to a high-precision map-guided automatic driving path planning and decision-making method, which belongs to the field of automotive AI intelligent driving technology.

Background technique

In recent years, autonomous driving technology has developed vigorously. The traditional autonomous driving system is an extension of the robot architecture and is divided into three modules: perception, planning and decision-making, and control. Perception is the use of various sensors such as cameras and radars to detect and identify the surrounding environment, thereby "seeing" the surrounding environment, analogous to the human eye. Planning and decision-making are based on the fusion of multiple sensor information and a clear understanding of oneself and the surrounding environment. It simulates manual driving. The human brain comprehensively judges the importance of each element and ultimately makes various decisions and plans different driving paths. Therefore, the planning and decision-making module of the autonomous driving system is also the most complex part. Control is to implement the brain's planning into the actual control system including brakes, accelerators, and brakes for execution.

Currently, a large part of traffic accidents are caused by human driving errors. The bad habits of human driving also cause the increase of congestion. Compared with human driving, machines do not need to rest like humans, nor do they feel tired. They always keep detecting the surrounding environment to minimize the risk of collision. Therefore, the application of autonomous driving technology will greatly improve the travel efficiency of the city, thereby reducing the cost of related industries and improving people's living standards. In addition, excellent autonomous driving technology can reduce the occurrence of traffic accidents. Therefore, the early realization of autonomous driving technology also has great social significance.

However, practice has proven that existing technology for autonomous driving is unreliable. If there is a map in the system, the system can retrieve the entire intersection from a few kilometers away, and by saving historical data on the map, it is easy to find that the intersection is prone to traffic accidents, so the vehicle can give early warning, slow down, and sense Strengthen etc. to avoid accidents. Autonomous driving systems often make mistakes, and bloody lessons have made everyone realize that high-precision maps are indispensable in autonomous driving systems. However, the current application of high-precision maps for autonomous driving is still in its infancy.

Due to the extremely high safety requirements of automobiles, the most critical issue restricting the commercialization of autonomous driving technology is the reliability of the technology itself. Therefore, the addition of maps will improve the performance of the perception, positioning and planning decision-making modules that are closely related to the safety of autonomous driving systems, help greatly accelerate the popularization and application of this technology, and provide better travel services.

At present, high-precision maps have been produced in small batches, but the production cost of high-precision map data is ten times that of ordinary map data, and there are currently no sufficient application areas. The field of autonomous driving is the focus of high-precision map applications, and solving the application problems of high-precision maps in autonomous driving technology is a top priority. But at present, the combination of high-precision maps and autonomous driving technology is still in its early stages.

There are currently many methods applied to smart cars, but these methods are all based on simple feature maps. Due to the lack of engineering high-precision map support, they cannot adapt to complex real road conditions, so they cannot be truly further engineered. How to use high-precision map information to improve existing intelligent vehicle route planning and decision-making methods is the core of current research.

Existing smart car behavior decision-making methods rely solely on the detection of perception systems or the support of simple feature maps. For some complex scenarios that are not easily recognized by the perception system or cannot be recognized, decision-making errors are prone to occur, resulting in collision risks. . How to use the detailed data of high-precision maps to avoid decision-making errors and reduce risks is also the focus of current research.

The path planning decision-making module is the brain of smart cars and one of the most critical technologies in the field of smart cars. The performance indicators of this module also largely determine the height that the entire system can achieve. The existing smart car planning and decision-making system is divided into four levels. The first level is long-distance road-level planning based on conventional map data, which is the planning from the current location to the destination, that is, mission planning; the second level is high-precision planning. Map-guided lane track-level path planning is based on road-level planning to further obtain lane track-level planning information; the third level is behavioral decision-making guided by high-precision maps. This step is to combine vehicle driving experience and traffic information. Plan abstraction and summary to assist in final path selection. The fourth level is local path planning, which is to solve the driving path within one or two hundred meters in front of the current vehicle. This step takes into account the location and behavior prediction results of obstacles and the vehicle's movement restrictions, and finally selects a path. Executable, collision-free optimal local driving trajectory consistent with vehicle dynamics. Although there are many research contents on path planning methods, because the concept of high-precision maps has not been popularized before, the current intelligent vehicle path planning and decision-making system does not make good use of high-precision map data.

Taken together, there are still several problems and defects in the existing technology. The key technical difficulties of the autonomous driving path planning and decision-making system are as follows:

(1) Existing technology automatic driving systems often make mistakes and cannot meet the extremely high safety requirements of automobiles. The most critical issue restricting the commercialization of automatic driving technology is the reliability of the technology itself. Current automatic driving systems are closely related to safety. There are several problems in the perception, positioning and planning decision-making modules of the system, which are not conducive to the popularization and application of this technology. The current production cost of high-precision map data is high, and the combination of high-precision maps and autonomous driving technology is not yet mature. The existing methods applied to smart cars are all based on simple element maps. Due to the lack of support from engineered high-precision maps, they cannot adapt to complex real road conditions. Therefore, they cannot be further engineered and cannot use the information of high-precision maps to Improve existing intelligent vehicle path planning decision-making methods. Existing smart car behavior decision-making methods rely solely on the detection of perception systems or the support of simple feature maps. For some complex scenarios that are not easily recognized by the perception system or cannot be recognized, decision-making errors are prone to occur, resulting in collision risks. , unable to use the detailed data of high-precision maps to avoid decision-making errors and reduce risks. The existing technology of autonomous driving cannot meet the requirements in terms of reliability, safety, intelligence, and economy.

(2) The existing technology for autonomous driving lacks a lane-track-level path planning method guided by high-precision maps. The lane-track-level path planning method is a necessary step in the context of high-precision map data being accurate to the lane track level. Conventional lane track level planning methods in the prior art simply perform lane track level topology calculations because the corresponding maps are relatively simple. They lack high-precision maps with complete elements and cannot further calculate each lane track. Similar lanes in line-level planned road sections lack the drivable area range that accurately describes the structured road, thus failing to provide effective support for vehicle lane changes and the generation of local planned trajectory clusters. This is fatal to the autonomous driving system and cannot provide safety. Economic path planning decisions ultimately lead to autonomous driving losing its value for large-scale promotion and application.

(3) Autonomous driving lacks decision-making methods guided by high-precision maps. The existing decision-making methods are based on perceived information. However, some information (such as the color and type of lane lines and beyond-visual-range information) is not easy to be perceived and identified, but is of great significance to decision-making (double yellow lines and solid white lines cannot change lanes, and accidents are prone to occur at intersections outside the line of sight). Therefore, there is a certain probability of decision-making errors. Regarding the decision-making risks that may be caused by the uncertainty of the perception system, due to the lack of high-precision maps to provide accurate ground static information, it is impossible to avoid decision-making errors caused by the uncertainty of perception, resulting in huge potential risks to safety and reliability. As a result, autonomous driving decisions are not only unsafe, but also lack scientificity and authority.

(4) Autonomous driving lacks local path planning guided by high-precision maps. Most of the existing autonomous driving team technologies are derived from smart car competitions, so local path planning methods are all based on road network files (domestic and foreign teams are based on RNDF file defined by DARPA), but the road network file information is sparse and the data model is too simple, which cannot adapt to more complex road scenarios, which will bring additional difficulties to local planning. In view of the local road expression dilemma and blurred planning boundaries caused by road network files, the existing technology lacks detailed road expression models combined with high-precision maps and accurate lane geometry and topology information, and cannot solve the road boundary deletion of local path planning methods. Ambiguous question. At the same time, the path selection results cannot completely match the structured road driving rules, there is a lack of lane and lane line data combined with high-precision maps, and there is a lack of cost calculation for path selection. It is impossible to solve the problem of path selection to ensure that the vehicle drives in the center of the structured road as much as possible, making Autonomous driving poses serious safety risks and can easily affect the overall order of road traffic. Such autonomous driving models lose their practical value.

Contents of the invention

In view of the fact that the current autonomous driving technology cannot achieve sufficient safety, and it only relies on the perception system without the assistance of the map, which is not stable enough, this application combines the high-precision map with the multi-level advanced autonomous driving system. It can not only assist the vehicle's perception system, but also the positioning system can assist the vehicle's path planning and decision-making system, which overall improves the stability of autonomous driving by a big step. Specifically, through high-precision map-guided lane track-level path planning, high-precision map-guided behavioral decision-making methods, and high-precision map-guided local path planning and selection, the safety, scientificity, and authoritativeness of autonomous driving decision-making have been greatly improved. , Tests have proven that as a substitute for manual driving, autonomous driving in this application can not only significantly reduce the incidence of traffic accidents, but also greatly improve the operating efficiency of urban highway transportation. It has huge market value and social value, and effectively solves the constraints that restrict autonomous driving. The most direct security issue for the commercial implementation of technology.

In order to achieve the above technical effects, the technical solutions adopted in this application are as follows:

The high-precision map-guided autonomous driving path planning and decision-making method first constructs the elements and data model of the high-precision map, and analyzes the data modeling method of the high-precision map through typical map data model examples. Then, the path guided by the high-precision map is The planning and decision-making system is divided into three layers: the first layer is the lane track level path planning layer, the second layer is the behavior decision-making layer, and the third layer is the local path planning layer. Each layer of methods uses high-precision map data for automatic driving. Application direction path planning decision analysis;

1) High-precision map-guided lane track-level path planning method: First, improve the Dijkstra method to perform high-precision map lane track-level path planning, and then use the corresponding relationship between roads and lanes in the map to reversely calculate similar lanes Express the planned passable area to meet the vehicle's obstacle avoidance and lane changing needs, and then construct a virtual lane through the cosine function to tangentially smooth the planned lane to solve the lane track-level planned lane jump problem caused by map division and meet the path requirements Based on the demand for planning smoothness, and finally based on the advantage of consuming very few resources for map topological relationship calculation within short distances, a lane track level path planning method is constructed with triggering methods including insufficient planning distance and lane changing driving;

2) High-precision map-guided driving behavior decision-making method: first analyze the action decision items under structured road surface conditions, express them through the state switching mechanism of the finite state machine, and then analyze the switching conditions between each state item of the state machine. Important The state machine switching conditions include: using the map to map obstacles, predicting the behavior of obstacles through the map, classifying left and right lane changes as horizontal decisions, and classifying the remaining decision items as vertical decisions; for horizontal decisions, divide There are three steps to determine the starting point of lane change intention generation, lane change intention generation and lane change condition judgment. Among them, the absolute difference between the expected vehicle speed and the current vehicle speed is greater than a certain critical value as the starting point of lane change intention generation, and the accumulated driving efficiency loss is greater than A certain critical value is used to generate the lane change intention, and finally the lane change condition is judged by combining the lane change intention with the map data, and finally a decision is generated;

3) Local path planning and selection method guided by high-precision maps: Firstly, the steps of the local path planning method are proposed, and then the center line of the map lane track-level planning lane is extracted as the planning baseline, and the boundary of the drivable area is used as the maximum margin of the trajectory cluster. When the planning curve spans drivable areas with different widths, the narrowest drivable area boundary is used as the maximum margin of the trajectory cluster, and finally the decision-making cost, road cost and comfort that affect the driving curve selection are integrated through cost calculation. Three different cost indicators are used, and the trajectory line is selected by the method of weighted summation instead of minimum cost, in which the road cost is obtained by extracting the lane line type.

Preferably, the lane trajectory level planning method:

Based on the optimal lane planning, a comprehensive plan for multiple suboptimal alternative lanes is obtained. The plan includes three parts. The first part is to calculate the lane trajectory level topology based on the road-level planning LinkID sequence and the one-to-many inclusion relationship between the road LinkID and the lane LaneID to ensure connectivity. The second part is to completely obtain the lane trajectory level LaneID sequence to ensure lane connectivity, and then backtrack layer by layer to calculate the same type of lanes using the differences in attributes between lanes. The third part is to segment the road segments based on the high-precision map data itself, and make virtual lane connections for smoothing the step parts according to the tangent of the road.

The lane trajectory level path planning is refined in two aspects:

(1) Proposal of similar lanes. In addition to the lane ID sequence obtained from lane track-level planning, its similar lanes are also provided to ensure local path planning and lane changing needs, and reflect the real road conditions to vehicles for their reference. ;

(2) A tangent point search was performed on the search results and a tangent translation was performed using the cosine function to ensure a smooth connection between the center lines of the two lanes.

Preferably, high-precision map lane trajectory level topology calculation:

Through high-precision positioning, the current s_lane and the end point e_lane are matched. Each long-distance planning is calculated by road-level planning. Each lane-level planning only needs to be responsible for calculating the data 1 to 2km ahead;

Based on the lane track-level topology information, the lane changing action must be realized by opening up the connections of multiple LaneIDs belonging to the same LinkID;

The lane change itself is the same as the lane length. It is converted into a distance cost and participates in the path calculation. The lane cost is the direct lane distance. The lane change cost is estimated, a virtual road change route is generated, and the virtual road change route is calculated. distance and multiplied by a weight:

Cost＝dst*pFormula 1

Assuming that the virtual lane length is the product of driving speed and lane changing time, the default value used in this application is 6 seconds lane changing time at 120km/h, that is, 216m. The weight is adjustable, and the default value used in this application is 1.5.

Preferably, the lane transition is smooth: the cosine function is used to generate a virtual lane at the tangential transition of the lane center line, and the two tangentially translated lanes are connected by continuously translating the points at the tangent surface of the existing road section. Thereby reducing the strength of the jump;

The triggering mechanism of lane track level path planning includes:

(1) Lane track-level path planning has a single length of more than 2km, which is based on static information. The first trigger condition is that the vehicle is about to reach the critical value of the result data of the single lane track-level path planning, and the lane is automatically triggered. Rail-level path planning and further planning;

(2) When the vehicle leaves the recommended lane of the current lane track-level path planning by changing lanes, the lane track-level path planning is automatically triggered.

Preferably, the behavioral decision items based on structured road traffic regulations:

(1) Drive straight in the lane - conventional decision-making, keep driving in the current lane: regardless of whether there are obstacles in the current lane, the vehicle keeps driving in the lane until a new decision causes a state switch;

(2) Turn left at the intersection - conventional decision-making. In the case of intersections, you will encounter: after the intersection, the vehicle enters slow operation, adapts to more calculation delays caused by faster planning frequency and perception frequency, and some other turns Properties in the process will also be gradually added and applied;

(3) Turn right at the intersection - conventional decision-making, which will be encountered at the intersection: after the vehicle enters the predetermined lane, it no longer determines the traffic light information and directly enters the lane change state;

(4) Go straight at the intersection - conventional decision-making. In the case of an intersection, you will encounter: the vehicle will enter the intersection and run slowly, adapting to more computing delays caused by faster planning and sensing frequency;

(5) Parking at the intersection - a conventional decision. When you are about to reach the intersection, you will encounter: the vehicle enters the state of slow driving and following the car;

(6) Left lane change - conventional decision-making, encountered when maintaining the lane: the weight of the left road increases;

(7) Right lane change - conventional decision-making, encountered when maintaining the lane: the weight of the right road increases.

(8) U-turn driving—conventional decision: distinguish U-turn lanes and increase the weight of U-turn lanes;

The obstacle behavior prediction method combined with the map lane attribute information: further use its geometric information to estimate the lateral movement of the obstacle, detect the sudden lane change as soon as possible, model the lateral movement of the vehicle, decompose the lane direction and lane tangent speed according to the vehicle speed, and obtain the tangential vehicle speed Vx; at the same time, match the vehicle center point with the lane center point, and obtain the matching distance D1. The tangential distance between the vehicle center point and the lane boundary is half the lane width minus D1, and the lateral crossing lane time calculation formula is obtained as:

V represents the true speed _of the vehicle, V For the matching point on the center line, Width represents the width of the vehicle; to predict the time to cross the lane line, it is between 1.5 and 2 seconds for small cars and 2.5 seconds for large cars. It is determined that the vehicle has lateral movement and constraints are imposed on the decision-making.

Preferably, the lateral decision of the lane-changing vehicle: the lane-changing decision includes: first, the generation of the lane-changing intention, second, the timing of starting the lane-changing intention or the condition of abandoning the lane-changing intention, and third, the judgment of the lane-changing decision;

(1) Generate lane change intention: Based on the description of driving efficiency loss accumulation LCDT, first, the lane change intention is generated when the current vehicle speed is lower than the expected speed; second, this inefficiency needs to accumulate for a period of time; the integral of the time when the current speed is lower than the expected speed is compared with the pre-set critical value to determine whether the generation condition of the lane change intention is met;

(2) Lane change intention starting timing or abandonment conditions: Determine the time point to start the cumulative calculation of the cumulative driving efficiency loss. After the difference between the desired speed and the current speed is greater than the set critical value, the cumulative calculation of the driving efficiency loss will be started. When the difference between the desired speed and the current driving speed is less than the set critical value, the calculated cumulative value will be abandoned and the cumulative value will be reset to zero, waiting for the next start;

(3) Judgment of lane change decision: When the cumulative value of driving efficiency loss is greater than the set critical value, the feasibility judgment of lane change is initiated.

Preferably, real-time local path trajectory cluster planning:

The first step is to obtain the planning base baseline: obtain a centerline base line of the road that the trajectory follows, which serves as a baseline. All trajectory lines in the trajectory cluster use the trajectory centerline as the planning baseline, and then generate a baseline based on the method near the baseline. A series of trajectory lines, the baseline is not only the core line of the trajectory cluster but also provides the direction of motion at the end of the trajectory line;

The second step is baseline smoothing: each trajectory is generated to ensure that it is as smooth as possible and the curvature changes linearly. The baseline is composed of a series of hash points, and the trajectory generation needs to be further expressed by a function.

The third step is to obtain the tangential edge restriction conditions: effectively limit the tangential divergence of the trajectory cluster, and the boundaries are reasonable to ensure the effectiveness of the planning and the controllable amount of calculations;

The fourth step is the generation of trajectory clusters: through the lateral deviation function and superimposing it with the baseline, while adjusting the maximum lateral deviation distance, a series of trajectory data with a certain lateral spacing are generated. Each trajectory of these trajectory clusters satisfies the vehicle kinematic constraints and is related to the current vehicle position and posture.

Preferably, the boundary constraint conditions are extracted based on the high-precision map: the high-precision map is introduced, and after the lane trajectory-level path planning guided by the high-precision map, a lane trajectory-level path planning result and similar lanes in each lane that can be used for lane change and obstacle avoidance are obtained. The acquisition of the similar lanes is the maximum deviation of the lateral path planning;

The specific method is as follows: after extracting the planned lane centerline hash point string from the LaneID obtained through the lane track-level path planning each time, the similar lanes of each segment of the lane track-level planned LaneID are further obtained through the results of the lane track-level path planning. , these same kind of lanes support lanes for vehicles to change lanes. The maximum sideslip distance is determined by the data of the same kind of lanes. Assuming that the width of each normal lane is the same, the calculation of the maximum sideslip distance q:

q _left/right =NumofBrotherLanes _left/right *Width Formula 3

NumofBrotherLaneS _left/right represents the number of similar lanes, Width represents the lane width, and the maximum lateral distance q must be the minimum value of the lateral distance within a single planning distance, centered on the lane track level planning.

Preferably, the trajectory cluster is generated: after obtaining the center line of the lane as the base baseline of the trajectory cluster and the vehicle's side slip distance limit parameters q _left and q _right , the vehicle's drivable area is obtained, and a series of vectors are generated in this space. Driving trajectories. The generation of these trajectories ensures that the vehicle has sufficient maneuverability and makes driving safer and more comfortable. The steps for generating trajectory clusters include:

Step 1: Positioning and matching with the baseline: The vehicle is mapped from the Cartesian coordinate system to the curved coordinate system of the baseline through the matching process, using the heading information of the base baseline;

Step 2: Determine the length of a single local path planning: determine the distance within which the vehicle can correct the heading and offset errors between the current vehicle and the baseline;

Step 3: Construct the sideslip change function and generate trajectory clusters

The length of a single path planning consists of two parts. The first part is the adjustment stage of heading deviation and lateral distance deviation. A cubic spline curve is used to simulate the curve of the adjustment stage and is solved by the following formula:

Δs=ss _i Equation 7

Among them, q(s) is the cubic spline function in the adjustment stage, s _i represents the mapping point on the baseline obtained by the vehicle through matching, s _f represents the arc length in the adjustment stage, q _f is the side of a certain trajectory line in the trajectory cluster. The lateral offset value is calculated through similar lanes; q _i represents the lateral distance between the vehicle and the baseline calculated by current point matching, a, b, c represent the corresponding parameters, and s represents the function arc length variable.

Preferably, comfort cost path selection:

The cost calculation method is used to calculate a cost for all candidate trajectories in the trajectory cluster, normalize multiple factors that affect trajectory selection, and finally form a unified standard, and normalize each factor by giving each influence The cost factors are given a quantitative measurement value and assigned weights respectively, so as to finally obtain the cost calculation function;

The curvature of the entire trajectory is cumulatively calculated as the comfort cost of the trajectory. Based on the exponential influence of curvature, the square of the curvature value is used as the integration condition:

In the above formula, k _i represents the curvature of the i-th point, and s is the arc length along the trajectory line;

Finally, assuming that the three types of costs are cost ₁ , cost ₂ , and cost ₃ respectively, each type of cost has different influence factors on the final choice, and a weight must be added to obtain the final cost calculation formula:

cost＝q ₁ *cost ₁ +q ₂ *cost ₂ +q ₃ *cost ₃ Equation 9

Get the final cost diagram.

Compared with the existing technology, the innovations and advantages of this application are:

First, this application is based on the large, rich, and accurate road static information provided by real high-precision maps. Based on the navigation application model, this application combines the path planning and decision-making method based on the perception system and the simple element map, and proposes a high-precision road planning decision-making method based on The precision map is suitable for the path planning decision-making system framework, modules and methods of smart cars, and focuses on improvements in three aspects: First, the lane track level path planning method guided by high-precision maps: improved Dijkstra method Carry out high-precision map lane track-level path planning, use the corresponding relationship between roads and lanes in the map to reversely calculate the passable area of the same lane expression plan to meet the vehicle's obstacle avoidance and lane changing needs, and construct a virtual lane through the cosine function to thereby Perform tangential smoothing on the planned lane to solve the lane jump problem of lane track level planning caused by map division and meet the smoothness requirements of path planning. The trigger method of constructing a lane track level path planning method includes insufficient planning distance and lane change driving. item; the second is the driving behavior decision-making method guided by high-precision maps: expressed through the state switching mechanism of the finite state machine, and then analyzing the switching conditions between each state item of the state machine, classifying left and right lane changes as lateral decisions, and other decisions The first is classified as vertical decision-making; the third is the local path planning and selection method guided by high-precision maps: the map lane track-level planning lane center line is extracted as the planning baseline, and the boundary of the drivable area is used as the maximum margin of the trajectory cluster, with the narrowest The boundary of the drivable area is used as the maximum margin of the trajectory cluster. Three different indicators that affect the decision-making cost, road cost and comfort cost that affect the selection of driving curves are integrated through cost calculation, and the trajectory line is calculated using a weighted sum instead of the minimum cost method. Make your selection. Through the above three improvements, the reliability, safety, intelligence and economy of autonomous driving have been greatly improved, which has huge practical value.

Secondly, as the current autonomous driving technology cannot achieve sufficient safety, it only relies on the perception system without the assistance of maps and the stability is insufficient. This application combines high-precision maps with the multi-level advanced autonomous driving system. The function of the precise map can not only assist the vehicle's perception system, but also the positioning system can assist the vehicle's path planning and decision-making system, which overall improves the stability of autonomous driving by a big step. Specifically, through high-precision map-guided lane track-level path planning, high-precision map-guided behavioral decision-making methods, and high-precision map-guided local path planning and selection, the safety, scientificity, and authoritativeness of autonomous driving decision-making have been greatly improved. , Tests have proven that as a substitute for manual driving, autonomous driving in this application can not only significantly reduce the incidence of traffic accidents, but also greatly improve the operating efficiency of urban highway transportation. It has huge market value and social value, and effectively solves the constraints that restrict autonomous driving. The most direct security issue for the commercial implementation of technology.

Third, in response to the problem that conventional lane track-level path planning methods cannot express the complete drivable area required by smart cars, this application proposes the concept and calculation method of similar lanes, improving the conventional lane planning method based on perception systems and simple feature maps. Intelligent vehicle lane track-level path planning method. The actual drivable area is expressed through the combination of similar lanes. On the one hand, non-conventional lanes such as emergency lanes are included in the drivable area of smart cars on regular road sections, so that smart cars can choose to drive or park in the emergency lane in an emergency; On the other hand, in the scene of road forks, certain normal lanes on the same road segment that are different from the lane track level planning driving direction are filtered out, thereby avoiding interference, which is conducive to improving the overall order of road traffic and helping to Dramatically accelerate the popularization and application of this technology and provide better travel services.

Fourth, purely relying on smart car perception systems and simple element maps for drivable area detection brings about problems such as missing lane attributes, incorrect detection of marking attributes, and incorrect lane ID mapping of obstacles to which obstacles belong, causing errors in calculating the right of way of the vehicle. The risk of error in decision-making. This application reduces the risk of error in judgment of decision-making conditions by extracting the geometry, lane and marking attribute information of high-precision maps. At the same time, it proposes an obstacle mapping method guided by high-precision maps and developed by drawing on the national standard LDW early warning conditions. The right-of-way change expected calculation method can not only be well adapted to the track-level mapping problem of obstacle lanes on large curved roads, but can also calculate the changes in road driving conditions caused by the lane-changing behavior of obstacle vehicles in real time, making the entire The intelligence and flexibility of autonomous driving are improved, the negative and uncertain impact on road traffic is reduced, and autonomous driving becomes more mature and reliable.

Description of the drawings

Figure 1 is a schematic diagram of the lane trajectory-level path planning results guided by high-precision maps.

Figure 2 is a diagram of the lane-level planning results after topological connection relationships between multiple lanes and starting and ending point matching.

Figure 3 is a schematic diagram of the lane centerline jump.

Figure 4 is a flow chart of the tangential smoothing algorithm for the jump lane center line.

Figure 5 is a schematic diagram of obstacle lateral motion prediction.

Figure 6 is a cumulative schematic diagram of driving efficiency loss in this application.

Figure 7 is an abstract schematic diagram of the intersection decision-making problem.

Figure 8 is a schematic diagram of the local path planning process guided by high-precision maps.

FIG9 is a schematic diagram showing the relationship between the baseline and the generated trajectory clusters.

FIG. 10 is a schematic diagram of road boundary restriction condition parameters.

Figure 11 is a schematic diagram of the generated trajectory cluster and comfort cost.

Figure 12 is a high-precision map surveying and mapping preparation flow chart.

Figure 13 is a schematic diagram of simulation verification path selection when there are no obstacles.

FIG. 14 is a schematic diagram of simulation verification path selection when obstacles are added.

Detailed ways

The following, in conjunction with the accompanying drawings, further describes the technical solution of the high-precision map-guided autonomous driving path planning decision method provided in the present application, so that technical personnel in the field can better understand the present application and implement it.

With the breakthrough development of AI and big data related technologies in recent years, autonomous driving technology, as a hot technology with significant commercial potential, has attracted more and more attention from the society. The core of autonomous driving technology is to solve how to make the machine have the ability of a driver, including the ability to perceive and recognize the environment, the ability to position itself, and the ability to make plans and make decisions. High-precision maps are one of the necessary conditions for achieving advanced autonomous driving. At present, autonomous driving is developed using a simplified version of high-precision maps, similar to RNDF road network files, but because the road network files contain very little information, there is still room for further exploration of the application potential of high-precision maps. This application is based on the large amount of rich and accurate road static information provided by real high-precision maps, and based on the navigation application mode, proposes a path planning decision system guided by high-precision maps suitable for smart cars.

This application first builds the elements and data models of high-precision maps, and analyzes the data modeling methods of high-precision maps through typical map data model examples. Then, the path planning decision-making system guided by high-precision maps is divided into three layers: First The first layer is the lane track level path planning layer, the second layer is the behavior decision-making layer, and the third layer is the local path planning layer. Each layer of the method analyzes the high-precision map data for automatic driving application direction path planning decision-making, respectively:

(1) Lane trajectory level path planning method guided by high-precision maps: First, the Dijkstra method is improved to perform lane trajectory level path planning based on high-precision maps. Then, the correspondence between roads and lanes in the map is used to reversely calculate the drivable area of the same type of lane expression plan to meet the vehicle's obstacle avoidance and lane change requirements. Then, a virtual lane is constructed through a cosine function to perform tangential smoothing on the planned lane to solve the lane jump problem caused by map division in lane trajectory level planning and meet the requirements of path planning smoothness. Finally, based on the advantage of minimal resource consumption in short-distance map topological relationship calculation, the lane trajectory level path planning method triggering mode includes two items: insufficient planning distance and lane change.

(2) High-precision map-guided driving behavior decision-making method: first analyze the action decision items under structured road surface conditions, express them through the state switching mechanism of the finite state machine, and then analyze the switching conditions between each state item of the state machine, Important state machine switching conditions include: using the map to map obstacles, predicting the behavior of obstacles through the map, and finally, classifying left and right lane changes as horizontal decisions, and the remaining decision items as vertical decisions; for horizontal Decision-making is divided into three steps: starting point judgment of lane change intention generation and lane change intention generation and lane change condition judgment. Among them, the absolute difference between the expected vehicle speed and the current vehicle speed is greater than a certain critical value as the starting point of lane change intention generation. Through cumulative driving When the efficiency loss is greater than a certain critical value, a lane change intention is generated. Finally, the lane change intention is combined with the map data to judge the lane change conditions, and finally a decision is generated;

(3) Local path planning and selection method guided by high-precision maps: Firstly, the steps of the local path planning method are proposed, and then the center line of the map lane track-level planning lane is extracted as the planning baseline, and the boundary of the drivable area is used as the maximum margin of the trajectory cluster , when the planning curve spans drivable areas with different widths, the narrowest drivable area boundary is used as the maximum margin of the trajectory cluster, and finally the decision-making cost, road cost and comfort that affect the driving curve selection are integrated through cost calculation. There are three different indicators of degree cost, and the trajectory line is selected through the method of weighted summation and minimum cost, in which the road cost is obtained by extracting the lane line type.

1. Lane track-level path planning method guided by high-precision maps

The traditional navigation method is that the navigator calculates the road-level calculation results. When the vehicle travels to a specific area, the navigator broadcasts the corresponding voice message to remind the driver to change lanes and then change the road. However, in the autonomous driving system, traditional semantic information is not enough to be well adopted by vehicles. Multiple lanes on the same road lead to different directions. The results of road-level navigation alone are likely to cause the autonomous vehicle to go to the wrong lane and then go Wrong path. Therefore, for autonomous driving needs, data information down to the lane trajectory level and data-based lane trajectory level navigation calculation results are indispensable.

1. Lane trajectory level planning method

On the basis of optimal lane planning, multiple sub-optimal alternative lane comprehensive plans are derived. The plan includes three parts. The first part is based on road-level planning, based on the road-level planning LinkID sequence and the relationship between road LinkID and lane LaneID. The one-to-many inclusion relationship is used to calculate the lane track level topology to ensure connectivity; the second part is to fully obtain the lane track level LaneID sequence to ensure lane connectivity, and then backtrack layer by layer to take advantage of the differences in attributes between lanes. Calculate similar lanes; the third part is based on the high-precision map data itself. The segmentation of road segments is based on the tangential direction of the road. The result of lane track-level planning is likely to cause a lateral step of the lane, so the step part is Make virtual lane connections for smoothing. A schematic diagram of the complete lane track level planning results is shown in Figure 1. The closer to the intersection, the closer the lane planned at the lane track level is to the rightmost lane until it enters the rightmost lane. When the vehicle reaches the interchange area about to exit the ramp, there is still a lane on the left side of the lane planned at the lane track level. There are several free lanes, but they will not be counted as the same kind of lanes because left lane changes are no longer allowed here.

This application makes two refinements to the lane track-level path planning:

(1) Proposal of similar lanes. In addition to the lane ID sequence obtained from lane track-level planning, its similar lanes are also provided to ensure local path planning and lane changing needs, and reflect the real road conditions to vehicles for their reference. ;

(2) Based on the search results, a tangent point search was performed and the cosine function was used for tangential translation to ensure a smooth connection between the center lines of the two lanes.

1. High-precision map lane track level topology calculation

Through high-precision positioning, the current s_lane and the end point e_lane are matched. Each long-distance planning is calculated by road-level planning. Each lane-level planning only needs to be responsible for calculating the data 1 to 2km ahead, thus significantly Reduces computational complexity.

The lane track level topological information can be seen from Figure 2(a). When the number of lanes remains unchanged, there is no one-to-many relationship between the preceding and following lanes between different road sections, and lane changing is not supported in the topological relationship. However, in actual planning, the lane changing action must be realized by opening up the connections of multiple LaneIDs belonging to the same LinkID. as shown in picture 2.

The lane change itself is the same as the lane length, which is converted into a distance cost and involved in the path calculation. The lane cost is the direct lane distance. The cost of changing lanes is obtained by estimation, generating a virtual lane change route, and calculating the distance of the virtual lane change route and multiplying it by a weight:

Cost＝dst*p Formula 1

Assuming that the virtual lane length is the product of driving speed and lane changing time, the default value used in this application is 6 seconds lane changing time at 120km/h, that is, 216m. The weight is adjustable, and the default value used in this application is 1.5. As shown in Figure 2(b), it is the calculation result of lane planning.

2. Smooth lane transitions

As shown in Figure 3, the results of lane track-level topology calculation cannot avoid the occurrence of lane changes, and the generation of lane change requirements will cause tangential translation of the road centerline data.

This application proposes a solution to generate a virtual lane using a cosine function at the tangential jump of the lane centerline, and to reduce the strength of the jump by continuously translating the points at the tangent plane of the existing road section to connect the two tangentially translated lanes. The method flow chart is shown in FIG4 .

3. Lane track level path planning triggering mechanism

(1) The length of a single lane track-level path planning exceeds 2km. It is a plan based on static information and has no time-varying characteristics. The first trigger condition is that the vehicle is about to arrive at the result data of the single-time lane track-level path planning. Critical value, automatically triggers lane track level path planning and then further planning;

(2) Lane track-level path planning, in which recommended lanes are obtained, which plays an important role in the decision-making of the planning module and local path planning. The lane track-level path planning adopts virtual lane smoothing for the recommended lane tangential translation. Considering the above two reasons, when the vehicle leaves the recommended lane of the current lane track-level path planning by changing lanes, the lane track-level path planning is automatically triggered.

2. Driving behavior decision-making method guided by high-precision maps

(1) Behavioral decision-making items based on structured road traffic regulations

(1) Drive straight in the lane - conventional decision-making, keep driving in the current lane: the longitudinal content is not considered, that is, regardless of whether there are obstacles in the current lane (even the obstacle speed is 0), the vehicle keeps driving in the lane until a new decision causes a state switch;

(2) Turn left at the intersection - conventional decision-making. In the case of intersections, you will encounter: after the intersection, the vehicle enters slow operation, adapts to more calculation delays caused by faster planning frequency and perception frequency, and some other turns Properties in the process will also be gradually added and applied.

(3) Turn right at the intersection - conventional decision-making, which will be encountered at the intersection: after the vehicle enters the predetermined lane, it no longer determines the traffic light information (unless the right-turn lane also has indicator light attributes), and directly enters the lane change state;

(4) Go straight at the intersection - conventional decision-making. In the case of an intersection, you will encounter: the vehicle will enter the intersection and run slowly, adapting to more computing delays caused by faster planning and sensing frequency;

(5) Parking at an intersection - conventional decision-making. When approaching an intersection, you may encounter: vehicles entering a slow-moving or following state;

(6) Left lane change - conventional decision-making, encountered when maintaining the lane: the weight of the left road increases;

(7) Right lane change - conventional decision-making, encountered when maintaining the lane: the weight of the right road increases.

(8) U-turn driving - conventional decision-making: distinguish the U-turn lane and increase the weight of the U-turn lane.

(2) Obstacle behavior prediction method based on map lane attribute information

In addition to predicting the behavior of obstacles through the lane trajectory-level attributes of high-precision maps, its geometric information is further used to estimate the lateral movement of obstacles, so as to detect sudden lane changes as soon as possible. As shown in Figure 5, it is the estimation of the lateral motion of the obstacle vehicle.

As shown in Figure 5, the lateral motion of the vehicle is modeled, and the vehicle speed is decomposed into the lane direction and the lane tangential speed, and the tangential vehicle speed Vx is obtained. At the same time, the vehicle center point is matched with the lane center point, and the tangential vehicle speed Vx is obtained. Matching distance D1, the tangential distance between the vehicle center point and the lane boundary is half the lane width minus D1, and the calculation formula for the lateral lane crossing time is:

V represents the true speed _of the vehicle, V For the matching point on the center line, Width represents the width of the vehicle; to predict the time to cross the lane line, it is between 1.5 and 2 seconds for small cars and 2.5 seconds for large cars. It is determined that the vehicle has lateral movement and constraints are imposed on the decision-making.

(3) Lateral decision-making of vehicles changing lanes

The lane change decision affects the vehicle's tangential lane selection. The behavior of lateral lane change is an inevitable but high-risk decision when the vehicle is driving. The lane-changing decision includes: first, the generation of the lane-changing intention; second, the timing or abandonment conditions of the lane-changing intention; and third, the judgment of the lane-changing decision.

(1) Generate lane change intention: Based on the description of driving efficiency loss accumulation LCDT, first, the lane change intention is generated when the current vehicle speed is lower than the expected speed; second, this inefficiency needs to accumulate for a period of time; the integral of the time when the current speed is lower than the expected speed is compared with the pre-set critical value to determine whether the generation condition of the lane change intention is met. Figure 6 is a schematic diagram of driving efficiency loss accumulation.

(2) Lane change intention starting timing or abandonment conditions: Determine the time point to start the cumulative calculation of the cumulative driving efficiency loss. After the difference between the desired speed and the current speed is greater than the set critical value, the cumulative calculation of the driving efficiency loss will be started. When the difference between the desired speed and the current driving speed is less than the set critical value, the calculated cumulative value will be abandoned and the cumulative value will be reset to zero, waiting for the next start.

(3) Lane change decision judgment: When the cumulative value of driving efficiency loss is greater than the set critical value, the lane change feasibility judgment is initiated.

(4) Speed control vertical decision-making

The control that only affects speed while maintaining lane status, as shown in Figure 7, uses high-precision maps to construct lane-track-level topological connection relationships for intersections, and further constructs virtual lane geometry information based on logical relationships.

As shown in Figure 7, lane L6 is a virtual lane. The LaneID sequence planned at the lane track level is extracted. There is no essential difference from the non-intersection situation. The only difference is that the L6 lane is a virtual lane and has some special features in the intersection situation. Lane attributes, but the vehicle does not have the ability to change lanes.

3. Local path planning and selection method guided by high-precision maps

The discrete optimization method is combined with the prior information of the high-precision map to connect with the high-precision map, and an application method of the high-precision map is obtained. The specific calculation process is shown in Figure 8.

(1) Real-time local path trajectory cluster planning

Figure 9 is a schematic diagram of the baseline and the generated trajectory cluster. The trajectory cluster is generated as follows:

The first step is to obtain the planning base baseline: obtain a centerline base line of the road that the trajectory follows, which serves as a baseline. All trajectory lines in the trajectory cluster use the trajectory centerline as the planning baseline, and then generate a baseline based on the method near the baseline. A series of trajectory lines, the baseline is both the core line of the trajectory cluster and provides the movement direction of the end of the trajectory line. Figure 9 shows a schematic diagram of the relationship between baselines and trajectory clusters;

The second step is baseline smoothing: each trajectory generation is guaranteed to be as smooth as possible and the curvature change is linear. The baseline is composed of a series of hash points, and further function expression is required for trajectory generation;

The third step is to obtain the tangential edge restriction conditions: effectively limit the tangential divergence of the trajectory cluster, and the boundaries are reasonable to ensure the effectiveness of the planning and the controllable amount of calculations;

The fourth step is the generation of trajectory clusters: through the side deflection function and superimposed with the baseline, the maximum side deflection distance is adjusted at the same time, and a series of trajectory data with a certain lateral spacing is generated. Each trajectory of these trajectory clusters satisfies the vehicle motion. The learning constraints are related to the current vehicle position and attitude.

(2) Selection and extraction of lane center lines based on high-precision maps

First, planning is based on a base baseline that guides the direction. The baseline is not only to ensure that the vehicle is driving on the road and does not deviate from the lane, it must also be steerable and the curvature changes smoothly. Each lane line of the high-precision map has high-precision road centerline data. To reproduce the curvature of real roads, the lane centerline point sequence data of the high-precision map replaces the road network point sequence of RNDF as the baseline for trajectory cluster generation, and directly extracts the lane centerline of the lane ID where the current lane trajectory-level path planning is located. Just click to list the data.

3. Extracting boundary constraints based on high-precision maps

As shown in Figure 10, a schematic diagram of road boundary restriction parameters. The traditional method of using RNDF files to define the maximum side slip parameter q has the following problems: First, the value of the side slip parameter q is relatively lagging and cannot change in a very timely manner with changes in the lane. Second, during driving, even if some road surfaces are wide enough, these drivable areas often cannot become a truly drivable place for vehicles. When you are about to exit a ramp, it is not recommended to stay in the middle of the road section at the ramp. When driving in the lane or even the passing lane, you should change lanes in time to the rightmost lane and track-level planned road.

This application introduces high-precision maps and uses high-precision map-guided lane track-level path planning to obtain a lane track-level path planning result and the same type of lanes that can be used for lane changing and obstacle avoidance in each lane. The obtained is the maximum deviation of the path lateral planning.

The specific method is as follows: after extracting the planned lane centerline hash point string from the LaneID obtained through the lane track-level path planning each time, the similar lanes of each segment of the lane track-level planned LaneID are further obtained through the results of the lane track-level path planning. , some of these similar lanes are on the left side of the lane track level planned lane, and some are on the right side, but they are all lanes that support vehicle lane changes. The maximum lateral deviation distance is determined by the same type of lane data, assuming that the width of each normal lane is the same Below, the calculation of the maximum sideslip distance q:

q _left/right =NumofBrotherLanes _left/right *Width Formula 3

NumofBrotherLaneS _left/right represents the number of similar lanes, Width represents the lane width, and the maximum lateral distance q must be the minimum value of the lateral distance within a single planning distance, centered on the lane track level planning.

(4) Generating trajectory clusters

After obtaining the center line of the lane as the base baseline of the trajectory cluster and the vehicle's lateral distance limit parameters q _left and q _right , the vehicle's drivable area is obtained, and a series of drivable trajectories are generated in this space. The generation ensures that the vehicle has sufficient maneuverability and makes driving safer and more comfortable. The steps for generating trajectory clusters include:

Step 1: Match the positioning with the base line: through the matching process, the vehicle is mapped from the Cartesian coordinate system to the curved coordinate system of the baseline, using the heading information of the base baseline;

Step 2: Determine the length of a single local path planning: determine the distance within which the vehicle can correct the heading and offset errors between the current vehicle and the baseline;

Step 3: Construct the side deviation change function and generate trajectory clusters

The length of a single path planning consists of two parts. The first part is the adjustment stage of heading deviation and lateral distance deviation. A cubic spline curve is used to simulate the curve of the adjustment stage and is solved by the following formula:

Δs＝ss _i Formula 7

Among them, q(s) is the cubic spline function in the adjustment stage, _si represents the mapping point on the baseline obtained by matching the vehicle, _sf represents the arc length in the adjustment stage, _qf is the lateral offset value of a trajectory line in the trajectory cluster, which is calculated by the same lane; _qi represents the lateral distance between the vehicle and the baseline calculated by matching the current point, a, b, and c represent the corresponding parameters, and s represents the function arc length variable.

(5) Comfort cost path selection

The cost calculation method is used to calculate a cost for all candidate trajectories in the trajectory cluster, normalize multiple factors that affect trajectory selection, and finally form a unified standard, and normalize each factor by giving each influence The cost factors are given a quantitative measurement value and weighted respectively, so as to finally obtain the cost calculation function.

The driving comfort of the vehicle depends on whether the local path planning trajectory is smooth enough and the curvature radius of each point on the trajectory. The greater the curvature, the smallest the turning radius of the vehicle and the sharper the bend. In order to comprehensively consider the curvature of the entire trajectory , this application uses the cumulative calculation of the curvature sum of the entire trajectory line as the comfort cost of the trajectory line. The influencing factors based on the curvature are exponential, and the square of the curvature value is used as the integration condition. As shown in Figure 11, a straight line segment is selected, a trajectory cluster is generated and the cost is calculated.

In the above formula, k _i represents the curvature of the i-th point, and s is the arc length along the trajectory line;

Finally, assuming that the three types of costs are cost ₁ , cost ₂ , and cost ₃ , the factors affecting the final choice are different, and a weight is added to obtain the final cost calculation formula:

cost＝q ₁ *cost ₁ +q ₂ *cost ₂ +q ₃ *cost ₃ Equation 9

Get the final cost diagram.

4. Experimental testing and analysis

(1) Test background

In the early stage, MATLAB simulation was used to verify the path planning method. In the later stage, the method was transplanted to the self-driving car as a verification platform to verify the selected solution. The verification site was selected as the large circular runway in the traffic test field of Beijing Highway Research Institute. The large outer ring of the site is a formal road with two-way four lanes. Common roads such as intersections and one-way streets are also built in the middle, and a variety of different traffic signs are set up for verification and testing.

To prepare the high-precision map of the experimental site, a professional high-precision map collection vehicle was used to survey and map the field, and the data was layered and processed in blocks according to the high-precision map format, as shown in Figure 12.

(2) Test plan projects

(1) MATLAB simulation verification method: The center line of a lane is randomly selected from the high-precision map data, and the local path planning method is verified based on the center line. After the trajectory cluster generation method can be implemented, randomly introduce some obstacles and see whether the trajectory selection method after cost calculation is feasible. As shown in Figure 13 and Figure 14.

(2) Smart car field verification platform: A test vehicle modified from Dongfeng AX7 is used. The test vehicle uses high-precision combined positioning equipment and RTK differential services. The lane centerline trajectory data of the center lane is collected through manual driving, and then edited using a dedicated map. The tool performs editing operations such as translating, interrupting, adding attributes, dividing lanes, etc. on the lane centerline data to generate a high-precision map data. The verification was carried out at the Tongzhou Highway Test Site.

(3) Test analysis

In the Matlab indoor simulation test, the road baseline is obtained by extracting a lane centerline from the high-precision map. As shown in Figure 13, the center line generated as a trajectory cluster has a large jagged run after zooming in, which is not as smooth as imagined. In addition, in order to reduce the amount of lane centerline data, the data extracted from the high-precision map centerline data was decimated in the experiment, which further increased the unsmoothness of the baseline. However, because the baseline is not smooth enough, the trajectories of the generated trajectory clusters have a tendency similar to overfitting, which often cannot well reflect the vehicle operating trajectories in real scenarios. In addition, as shown in Figure 14, when no obstacles are introduced, the trajectory of the final selected lane line is basically the trajectory parallel to the trajectory center line. This situation is due to the fact that when calculating the cost, the comfort cost is related to the along the trajectory. Comprehensive reflection of the cost of traveling along the center line of the trajectory line. As shown in Figure 14, after introducing obstacles of different shapes and sizes, the trajectory is selected correctly.

(4) Test summary

In view of the path planning decision-making solution guided by high-precision maps, this application conducted simulation and field testing, and verified the feasibility of high-precision map-assisted planning in terms of process. In the actual process, thanks to the assistance of high-precision maps, the adoptability and flexibility of the overall system have been further improved. Especially in specific scenarios, it is completely achievable to use the assistance of high-precision maps to meet the stability and reliability required by the autonomous driving system under limited conditions.

Claims (10)

1. The automatic driving path planning decision-making method guided by the high-precision map is characterized in that firstly, elements and a data model of the high-precision map are constructed, a data modeling method of the high-precision map is analyzed through a typical map data model example, and then, the path planning decision-making system guided by the high-precision map is divided into three layers: the first layer is a lane trajectory level path planning layer, the second layer is a behavior decision layer, the third layer is a local path planning layer, and each layer of method carries out automatic driving application direction path planning decision analysis on high-precision map data;

1) The high-precision map guided lane trajectory level path planning method comprises the following steps: firstly, improving a Dijiestra method to conduct high-precision map lane trajectory level path planning, then, reversely calculating a passable area of similar lane expression planning by utilizing the corresponding relation between a road and a lane in a map so as to meet the obstacle avoidance and lane change requirements of a vehicle, constructing a virtual lane by a cosine function so as to conduct tangential smoothing on a planned lane to solve the lane trajectory level planning lane jump problem caused by map division, meeting the requirement of path planning smoothness, and finally, calculating the advantage of little resource consumption based on the map topological relation in a short distance, wherein the triggering mode of the lane trajectory level path planning method is constructed to comprise two items of insufficient planning distance and lane change driving;

2) The driving behavior decision method guided by the high-precision map comprises the following steps: firstly, analyzing action decision items under the condition of structured road conditions, expressing the action decision items through a state switching mechanism of a finite state machine, and then analyzing switching conditions among all the state items of the state machine, wherein the important state machine switching conditions comprise: mapping the obstacle by using the map, prejudging the behavior of the obstacle by using the map, and classifying left and right lane changes into a horizontal decision, wherein the rest decision items are classified into a vertical decision; dividing a lane change intention generation starting point judgment, a lane change intention generation and lane change condition judgment aiming at a transverse decision, wherein the lane change intention is generated by accumulating that the driving efficiency loss is greater than a certain critical value by taking the absolute difference between the expected vehicle speed and the current vehicle speed as the starting point of the lane change intention generation, and finally the lane change intention is combined with map data to carry out lane change condition judgment to finally generate the decision;

3) The high-precision map guided local path planning and selecting method comprises the following steps: firstly, providing a local path planning method, then extracting a map lane trajectory level planning lane central line as a planning base line, taking the boundary of a drivable region as the maximum margin of a track cluster, taking the boundary of the narrowest drivable region as the maximum margin of the track cluster under the condition that the planning curve spans the drivable region with different widths, finally integrating three different indexes of decision cost, road cost and comfort cost which influence driving curve selection in a cost calculation mode, and selecting a trajectory line in a method of taking the minimum cost by weighting summation, wherein the road cost is obtained by extracting a lane line type.

2. The high-precision map-guided automatic driving route planning decision method according to claim 1, characterized in that the lane trajectory level planning method:

obtaining a plurality of suboptimal alternative lane comprehensive schemes on the basis of optimal lane planning, wherein the scheme comprises three parts, and the first part is to perform lane trajectory level topology calculation on the basis of road level planning based on a road level planning LinkID sequence and a one-to-many inclusion relationship between a road LinkID and a lane LaneID so as to ensure connectivity; the second part is that lane line level LaneID sequences are completely obtained, and on the basis of ensuring lane communication, backtracking is performed layer by layer, and the similar lanes are calculated by utilizing the attribute difference between the lanes; the third part is to make virtual lane connection for step parts for smoothing the tangential direction of the road by cutting the road section based on the high-precision map data per se;

two-point refinement is carried out on the lane trajectory level path planning:

(1) The method comprises the steps of providing the same kind of lanes, providing the same kind of lanes besides a lane ID sequence obtained by the lane trajectory level planning, ensuring the local path planning and the lane changing requirements, and reflecting the real road condition to the vehicle for reference;

(2) Tangential point search is carried out on the search result, tangential translation is carried out by utilizing a cosine function, and smooth connection of the center lines of the two lanes is ensured.

3. The high-definition map-guided automatic driving path planning decision method according to claim 1, characterized in that the high-definition map lane trajectory level topology calculation:

the current s_lane and the e_lane of the terminal point are matched through high-precision positioning, each long-distance planning is completed by road level planning, and each lane trajectory level planning only needs to be responsible for calculating the data of 1-2 km in front;

based on the lane trajectory level topology information, the lane changing action must be realized by opening the connection of a plurality of LaneIDs belonging to the same LinkID;

the lane change is the same as the lane length, the distance cost is converted, the lane change participates in the path calculation, the cost of the lane is the direct lane distance, the cost of the lane change is obtained through estimation, a virtual lane change line is generated, the distance of the virtual lane change line is calculated, and the distance is multiplied by a weight value:

cost=dst×p type 1

Assuming that the virtual lane length is the product of the running speed and the lane change time, the default value adopted in the method is that the lane change time of 6s, namely 216m, is adjustable under the condition that the default value adopted in the method is 120km/h, and the default value adopted in the method is 1.5.

4. The high-precision map-guided automatic driving path planning decision method according to claim 1, characterized in that the lane change jump is smoothed: the scheme of generating a virtual lane at the tangential jump position of the lane center line by utilizing a cosine function is that two tangential translation lanes are connected by continuously translating at a point at the tangent plane of the existing road section, so that the jump force is reduced;

The lane trajectory level path planning triggering mechanism includes:

(1) The single length of the lane trajectory level path planning exceeds 2km, which is based on static information, and the first triggering condition is a critical value of result data of the lane trajectory level path planning of the vehicle which is about to arrive at the single lane trajectory level path planning, and the lane trajectory level path planning is automatically triggered to be further planned forward;

(2) And when the vehicle leaves the recommended lane of the current lane trajectory level path planning in a lane changing mode, automatically triggering the lane trajectory level path planning.

5. The high-precision map-guided automatic driving route planning decision method according to claim 1, wherein the behavior decision term based on the structured road traffic rules:

(1) Lane straight-through-normal decision, keep current lane running: regardless of whether the current lane is obstructed, the vehicle keeps driving on the lane until a new decision results in a state switch;

(2) Left turn of the intersection-conventional decision, the situation of the intersection encounters: after the intersection, the vehicle enters into slow running, so that more calculation delay caused by faster planning frequency and perceived frequency is adapted, and other attributes in the turning process are gradually added for application;

(3) Right turn at intersection-conventional decision, the situation at intersection encounters: after the vehicles at the intersection enter a preset lane, no traffic light information is judged, and the vehicles directly enter a lane changing state;

(4) Crossing straight-going-conventional decision, the situation of crossing will meet: the vehicle slowly runs in the crossing, and is suitable for faster planning and more calculation time delay brought by the perception frequency;

(5) Intersection parking-conventional decision, the intersection is about to be reached, the following is met: the vehicle enters a state of slow running and following;

(6) Left lane change-conventional decision, encounter with lane keeping: the weight of the left road is increased;

(7) Right lane change-conventional decision, encounter with lane keeping: the weight of the right road is increased;

(8) Turning around-conventional decision: distinguishing a turning lane and weighting the turning lane;

the obstacle behavior prejudging method combining map lane attribute information comprises the following steps: further estimating the lateral movement of the obstacle by utilizing the geometric information of the obstacle, finding the condition of sudden lane change as timely as possible, modeling the lateral movement of the vehicle, decomposing the speed of the lane direction and the tangential speed of the lane according to the speed of the vehicle, and obtaining the tangential speed Vx of the vehicle; meanwhile, the vehicle center point is matched with the lane center point, the matching distance D1 is obtained, the tangential distance between the vehicle center point and the lane boundary is obtained by subtracting D1 from the width of the half lane, and the calculation formula of the time for the lateral crossing lane is obtained as follows:

V represents the actual speed of the vehicle, V _x The component of the vehicle speed V in the tangential direction of the road is represented, delta represents an included angle between the vehicle speed direction and the lane direction, lanePt.head represents the lane direction, vehicle.head represents the vehicle running direction, lanePt represents a matching point of the vehicle center on the lane center line, and Width represents the vehicle Width; predicting the time of crossing the lane line, taking 1.5-2 seconds for a small vehicle, and 2.5 seconds for a large vehicle, recognizing that the vehicle moves sideways, and making a constraint on decision.

6. The high-precision map-guided automatic driving path planning decision method according to claim 1, characterized in that the lane-changing vehicle lateral decision: lane change decisions include: firstly, generating a lane changing intention, secondly, starting the time or giving up the condition of the lane changing intention, and thirdly, judging a lane changing decision;

(1) Generating a lane change intention: based on running efficiency loss accumulated LCDT description, firstly, lane change intention is generated that the current speed is smaller than the expected speed; second, this inefficiency requires accumulation for a period of time; comparing the integral of the current speed less than the expected speed with a preset critical value, and judging whether the generation condition of the lane change intention is reached;

(2) Lane change intention starting timing or abort conditions: determining a time point for starting to calculate the running efficiency loss accumulated value in an accumulated manner, starting to start running efficiency loss accumulated calculation after the difference between the expected speed and the current speed is larger than a set critical value, giving up the calculated accumulated value when the difference between the expected speed and the current running speed is smaller than the set critical value, and returning the accumulated value to zero to wait for the next starting;

(3) Judging lane change decision: and when the running efficiency loss accumulated value is larger than a set critical value, starting the lane change feasibility judgment.

7. The high-precision map-guided automatic driving path planning decision method of claim 1, wherein the real-time local path trajectory cluster planning:

first, acquiring a planned base baseline: obtaining a central line base line of a road followed by a track, playing a role of a base line, taking the central line of the track as a planning base line for all the track lines in the track cluster, then generating a series of track lines near the base line according to a method, wherein the base line is a core line of the track cluster and provides a movement direction of the tail end of the track line;

second, baseline smoothing: each track generation ensures that the curvature change is as smooth as possible and linear as possible, a base line is formed by a series of hash points, and further function expression is needed for track generation;

thirdly, acquiring tangential edge limiting conditions: the tangential divergence of the track clusters is limited effectively, the boundary is reasonable, the planning effectiveness is ensured, and the calculated amount is controllable;

fourth, generating a track cluster: and superposing the track clusters with the base line through a lateral deviation function, simultaneously adjusting the maximum lateral deviation distance, and simultaneously generating a series of track data with a certain lateral distance, wherein each track of each track cluster meets the kinematic constraint of the vehicle and is related to the current position and the gesture of the vehicle.

8. The high-definition map guided automatic driving route planning decision method according to claim 1, wherein the boundary constraint condition is extracted based on the high-definition map: introducing a high-precision map, and obtaining a lane trajectory level path planning result and a similar lane which can be used for the obstacle avoidance of the lane in each section after the lane trajectory level path planning guided by the high-precision map, wherein the obtaining of the similar lane is the maximum deviation of the path lateral planning;

the specific method comprises the following steps: after each LaneID obtained through lane trajectory level path planning extracts a planned lane center line hash point string, similar lanes of each section of lane trajectory level planning LaneID are further obtained through the lane trajectory level path planning result, the similar lanes support lanes for vehicle lane change driving, the maximum cornering distance is determined by similar lane data, and under the condition that the widths of normal lanes are consistent, the maximum cornering distance q is calculated:

q _left/right ＝NumofBrotherLanes _left/right * Width 3

NumofBrotherLaneS _left/right The Width represents the Width of the lane, and the maximum lateral deviation distance q is the minimum value of the lateral distance under the condition of taking the lane trajectory level planning as the center in the single planning distance.

9. The high-precision map-guided automatic driving path planning decision method of claim 1, wherein a track cluster is generated: obtaining the center line of the lane as the base baseline of the track cluster and the lateral offset distance limit of the vehicleParameter q _left And q _right Then, a drivable area of the vehicle is obtained, a series of drivable tracks are generated in the space, the generation of the tracks ensures that the vehicle has enough mobility, the driving behavior is safer and more comfortable, and the track cluster generation step comprises the following steps:

step one, positioning and matching with a base line: mapping the vehicle from a Cartesian coordinate system to a curve coordinate system of a base line through a matching process, and utilizing heading information of the base line;

step two, determining a single local path planning length: determining how long the vehicle is within to correct the heading and offset errors existing between the current vehicle and the base line;

step three, constructing a lateral deviation change function and generating a track cluster

The length of the single path planning comprises two parts, wherein the first part is an adjustment stage of course deviation and lateral distance deviation, a curve of the adjustment stage is simulated by adopting a cubic spline curve, and the problem is solved by the following formula:

Δs＝s-s _i 7. The method of the invention

Wherein q(s) is a cubic spline function of the adjustment stage, s _i Representing a mapping point on a baseline of the vehicle obtained by matching s _f Arc length, q, representing the adjustment phase _f The lateral offset value of a certain track line in the track cluster is calculated through the similar lanes; q _i Representing the lateral distance of the vehicle from the base line calculated by the current point matchingAnd a, b and c represent corresponding parameters, and s represents a function arc length variable.

10. The high-precision map-guided automatic driving route planning decision method of claim 1, wherein comfort cost route selection:

calculating a cost for all alternative tracks in the track cluster by adopting a cost calculation mode, normalizing a plurality of factors influencing track selection to finally form a unified standard, and normalizing each factor by giving a quantitative measure value to each factor influencing the cost and giving weights to each factor so as to finally obtain a cost calculation function;

the curvature on the whole track line is calculated in an accumulated mode and the comfort cost serving as the track line is calculated, the influence factor based on the curvature is exponential, and the square of the curvature value is used as an integral condition:

in the above, k _i Representing the curvature of the ith point, s being the arc length along the trajectory;

Finally, assume three types of costs, costs respectively ₁ ，cost ₂ ，cost ₃ The influence factors of various cost values on the final selection are different, and a weight is added to obtain a final cost calculation formula:

cost＝q ₁ *cost ₁ +q ₂ *cost ₂ +q ₃ *cost ₃ 9. The invention is applicable to

And obtaining a final cost schematic diagram.