Method for optimizing a vehicle's path planning

Abstract

Method for optimizing path planning of a vehicle (1), comprising a search algorithm that calculates a path to a predetermined destination of the vehicle (1) based on static information used as attractive or repulsive forces, wherein the search algorithm additionally integrates dynamic information as attractive forces into the planning.

Description

[0001] The invention relates to a method for optimizing path planning of a vehicle.

[0002] Autonomous vehicles are well known from the field of robotics, for example, in logistics or transportation. The main task of such vehicles is to recognize their environment and adapt their movements to reach their destination while avoiding obstacles as much as possible. In more recent approaches, unknown environments are also incorporated into the planning of possible paths to reach the destination.

[0003] However, path planning is becoming increasingly important in the automotive industry to advance the development of (partially) automated driving and even autonomous driving. A path planning problem is the search for a path from a starting point to possible destinations, which can also take into account possible collisions with obstacles. To optimize the path, costs, for example, which can include the distance or the time required, are minimized.

[0004] For path planning in the automotive industry, it is not only important to consider unknown environments, but also to consider moving objects. Methods such as the potential field method are used here, in which static obstacles, lane markings, or object positions are considered repulsive forces and the path around these obstacles is avoided.

[0005] For current path planning applications, the environment model provides, for example, lane markings, a collective lane path obtained from the movement of the adjacent vehicles, information about peripheral buildings and the distance traveled by the vehicle in front. It is difficult to generate a drivable path from the information about peripheral buildings, adjacent vehicles and the vehicle in front. For this reason, control is currently only based on a selected reference, e.g. either only in relation to the lanes or only to a selected object history. In addition, it is difficult to determine a meaningful path from the multitude of information, especially if information for part of the route is missing, e.g. if information about lane markings is missing in a section of the route.

[0006] DE 10 2012 100 164 A1 describes a lane control system for generating a collision-free vehicle path. The space surrounding a vehicle is divided into grid cells. Each cell is classified as occupied or unoccupied. Using a lane information fusion module, an artificial potential field is created around the vehicle. A repulsive force generated by the potential field ensures that the vehicle follows the curve without the threat of collision with overlapping occupied cells.

[0007] DE 10 2005 002 719 A1 discloses a method for course prediction in driver assistance systems for motor vehicles, in which a course hypothesis is created based on information from different information sources. The method comprises extracting a set of raw data for the course hypothesis from each information source, representing the various sets of raw data in a unified description system, and merging the raw data to form the course hypothesis.

[0008] In a method proposed in DE 10 2014 200 638 A1 for estimating a lane course for an ego vehicle moving on a roadway with multiple lanes, the following steps are repeatedly carried out: detecting one or more other vehicles moving on the roadway in the same direction as the ego vehicle; detecting vehicle positions for each of the detected other vehicles; forming a trajectory for each of the detected other vehicles based on the respectively detected vehicle positions; assigning the trajectories to possible lanes based on a course of the formed trajectories in the environment of the ego vehicle; selecting one of the possible lanes to which at least one trajectory is assigned; and estimating the lane course based on the at least one trajectory assigned to the selected lane.

[0009] A method described in DE 10 2014 201 382 A1 for operating a driver assistance system of a vehicle comprises the steps of: determining a current position of a pedestrian in an environment of the vehicle; determining a first current movement state of the pedestrian; determining a second current movement state of the vehicle; calculating a location probability distribution of the pedestrian, wherein the location probability distribution is a function of time and space and is based on a pedestrian movement model in conjunction with the determined current position of the pedestrian and the determined current movement state of the pedestrian; calculating a trajectory, based on the calculated location probability distribution of the pedestrian and the second current movement state of the vehicle, with a minimum collision probability for the vehicle and the pedestrian;and operating the vehicle's driver assistance system based on the calculated trajectory;

[0010] It is an object of this invention to provide a method that enables the optimization of path planning. This object is achieved according to the invention by the features of the independent patent claims. Advantageous embodiments are the subject of the dependent claims.

[0011] According to the invention, a method for optimizing a path planning of a vehicle is proposed, comprising a search algorithm that calculates a path to a predetermined destination of the vehicle based on static information used as attractive or repulsive forces, wherein the search algorithm additionally integrates non-static data or dynamic information as attractive forces into the planning.

[0012] In one embodiment, data includes non-static data or dynamic information of a history of a path of a leading vehicle, a trajectory of one or more detected objects, such as the trajectory of a leading vehicle, an orientation of one or more detected objects, such as the orientation of other road users.

[0013] In one embodiment, a cost function based on a weighting of the influence of the repulsive and attractive forces is integrated into the search algorithm.

[0014] Previously known search algorithms only use static information as attractive and repulsive forces, whereby static information can be defined as static obstacles such as edge buildings, lane markings or object positions as repulsive forces and a target position or support points as attractive forces. By using non-static data or dynamic information such as the history of a vehicle in front or a trajectory of a vehicle in front, the robustness and availability of the reference line for path maintenance or planning can be increased. This means that even when information, e.g. about the lane, continuous and improved path planning takes place, i.e. gaps can be bridged and closed and local minima can be avoided. Improved planning can be achieved by weighting the attractive and repulsive forces.

[0015] In one embodiment, the search algorithm additionally uses predefined maneuver templates for predefined maneuvers.

[0016] By using predefined maneuver templates, the computational effort can be reduced without affecting the quality of the results.

[0017] In one embodiment, path smoothing is performed over the calculated path. Smoothing the path makes it possible to obtain a reference line that can be continuously differentiated at least twice, thus achieving a smoother curve or line. This increases comfort for the driver.

[0018] Furthermore, a computer program is proposed which is designed to carry out the method according to one of the preceding claims.

[0019] Further features and advantages of the invention will become apparent from the following description of exemplary embodiments of the invention, with reference to the figures of the drawing, which illustrate details of the invention, and from the claims. The individual features can be implemented individually or in combination in a variant of the invention.

[0020] Preferred embodiments of the invention are explained in more detail below with reference to the accompanying drawings. Fig. 1 shows a path planning according to an embodiment of the present invention. Fig. 2 shows the path planning with path smoothing from Fig. 1.

[0021] In the following descriptions of the figures, the same elements or functions are provided with the same reference symbols.

[0022] Fig. Figure 1 shows an exemplary path planning for a vehicle according to an embodiment of the invention. The planning of the path for the vehicle 1 is carried out within a predetermined area, which is indicated by thick solid lines in Fig. 1. The space is selected by the driving strategy in such a way that a sensible guidance of vehicle 1 is possible for the path. This means that paths very close to vehicle 1 and at an acute angle leading away from vehicle 1 are not considered, as these are classified as impossible paths.

[0023] The optimization area is restricted based on environmental information such as lane markings. Maneuver templates are used to accelerate the determination of the optimal path. The type of maneuver template is determined based on situations previously defined offline by experts and stored in the control unit for retrieval when needed. Online, the driving strategy selects the appropriate maneuver template based on situation recognition, for example, from sensors or camera systems. This significantly reduces the search area and thus the computing power required to calculate the optimal path.

[0024] To further plan the path of vehicle 1, not only static obstacles such as edge buildings, lane markings and the positions of other objects such as vehicles are included in the search algorithm's calculation. Non-static data or dynamic information is also used, e.g. the path history of a vehicle 2 in front, whereby the vehicle in front does not necessarily have to be the vehicle in front, but can also be a vehicle 2 that has since changed lanes, etc. A path history of another object, e.g. another road user, can also be used. It is advantageous if a history, e.g. a trajectory, is covered or recorded for at least part of the route leading to the destination for the vehicle 1. Other parameters such as other road users and their orientation can also be used as data to calculate the optimal path.

[0025] To further improve the algorithm, a cost function can be integrated, as shown in formula (1), which includes a heuristic, i.e., an estimation. Here, attractive and repulsive forces are assigned a weighting factor according to predefined criteria, e.g., criteria stored for specific situations. This allows the influence of individual situations or events, e.g., obstacles, on the path to be assessed or weighted and then incorporated as a factor into the path planning. More heavily weighted factors or situations are evaluated either as larger obstacles, i.e., with higher repulsion or repulsion, or with higher attraction, so that an optimal path for vehicle 1 can be found.

[0026] By incorporating non-static or dynamic information, which also flows into the cost function, areas with missing information can be covered with additional available information. Thus, path planning can be controlled not only based on lanes, but also based on information about the path traveled by a leading vehicle 2.

[0027] The cost function is preferably set up as follows: fb=∑i=1nwi∗fb,i (1), where f b the cost function, n the number of existing repulsive or attractive forces, w i the weighting factor and f b,i represent the influence of the respective repulsive or attractive force on the path of the vehicle.

[0028] The search algorithm is advantageously an RRT algorithm, but can also be any other suitable algorithm as long as the inclusion of repulsive and attractive forces leads to an improvement in path planning.

[0029] Fig. 2 shows the path planning with path smoothing from Fig. 1. To smooth the calculated path, known methods, such as filters or interpolation, can be used. Smoothing serves to compensate for any discontinuities in the path and thus achieve an even more drivable path. This increases comfort for the driver.

[0030] By means of the method of the present invention, which can be implemented as a computer program which in turn can be executed on a computing unit, a higher robustness and availability of the reference line for path planning of the vehicle 1 is achieved.

Claims

[1] Method for optimizing a path planning of a vehicle (1), comprising a search algorithm that calculates a path to a predetermined destination of the vehicle (1) based on static information used as attractive or repulsive forces, wherein the search algorithm additionally integrates dynamic information as attractive forces into the planning. [2] The method of claim 1, wherein the dynamic information comprises a history of a path of a leading vehicle (2), a trajectory of one or more detected objects and / or an orientation of one or more detected objects. [3] Method according to claim 1 or 2, wherein a cost function based on a weighting of the influence of the repulsive and attractive forces is integrated into the search algorithm. [4] Method according to one of the preceding claims, wherein the search algorithm additionally uses predetermined maneuver templates for predetermined maneuvers. [5] Method according to one of the preceding claims, wherein path smoothing (21) is carried out over the calculated path. [6] Computer program comprising instructions which, when the computer program is executed by a computing unit, cause the computing unit to carry out the method according to one of the preceding claims. [7] A computing unit which is programmed by means of the computer program according to claim 6 to carry out the method according to one of claims 1 to 5.

Images