Trajectory planning for an autonomous vehicle

Abstract

Method for planning a trajectory for a vehicle, which moves in particular along a static reference path, with the following steps: - Providing (S1) linearly sorted boundary conditions for the movement of the vehicle; - Capturing and analyzing (S2) a vehicle environment based on vehicle environment-related sensor data regarding collision risks of the vehicle's trajectories; - Determining (S3) collision-free trajectories or excluding trajectories that are not collision-free; - Evaluating (S4) the collision-free or non-excluded trajectories, based on a lexicographical comparison of the linearly sorted boundary conditions.

Description

AREA OF INVENTION

[0001] The present invention relates to a method for determining a trajectory for a vehicle. TECHNICAL BACKGROUND

[0002] Patent application 102022210586.9 of the applicant shows trajectory planning based on lexicographically sorted boundary conditions. SUMMARY OF THE INVENTION

[0003] Based on this, the invention aims to assess the safety of an autonomous vehicle.

[0004] Accordingly, the following is planned: - A method for planning a trajectory for a vehicle, which moves in particular along a static reference path, comprising the following steps: providing linearly sorted boundary conditions for the movement of the vehicle; capturing and analyzing a vehicle environment with respect to collision risks of the vehicle's trajectories; determining collision-free trajectories or excluding trajectories that are not collision-free; evaluating the collision-free or non-excluded trajectories by means of a lexicographical comparison of the linearly sorted boundary conditions, if a collision-free or non-excluded trajectory exists.

[0005] A sensor, also known as a detector, transducer, or probe, is a technical component that can qualitatively or quantitatively detect certain physical or chemical properties or conditions, such as temperature, humidity, pressure, speed, brightness, acceleration, pH value, ionic strength, electrochemical potential, and / or the material composition of its environment. These quantities are detected using physical or chemical effects and converted into a processable electrical signal as sensor data.

[0006] Linear sorting means that all elements of the set are comparable, and for every pair of elements, there exists a unique order according to which they can be sorted. This means that no element is simultaneously less than, greater than, and equal to any other element in the order.

[0007] Vehicle sensors are either mountable or already mounted on a vehicle. Vehicle sensors include optical sensors, such as cameras, lidar, radar, time-of-flight (TOF), and other sensor technologies, such as acoustic sensors.

[0008] A trajectory, also called path or route, is the course of a spatial or surface curve along which a body or a point, for example, the center of gravity of a body, moves. In this patent application, the trajectory includes additional location-dependent information, such as time or velocity.

[0009] Collision-free trajectories are trajectories that are collision-free with certainty or with a predetermined probability. Accordingly, a collision-free trajectory may still carry a residual risk of collision due to the movement of another object.

[0010] A reference path is the trajectory or path along the route, i.e., the trajectory without the additional location-dependent information. In this patent application, the reference path runs through static reference points, such as bus stops or other fixed points on a route, which lie on the trajectory. The trajectory between the reference points can be predefined, e.g., for rail vehicles, or variable, e.g., for road vehicles.

[0011] Boundary conditions in this patent application are not necessarily of a physical nature. Traffic regulations or specifications regarding driving dynamics or comfort are examples of boundary conditions that affect the movement of the vehicle. Accordingly, boundary conditions do not restrict the physical movement possibilities of a vehicle. Compliance with boundary conditions is a target specification that is not technically necessary for the operation of the vehicle.

[0012] Traffic rules in Germany are partly regulated by the Road Traffic Regulations (StVO). On private property, traffic rules are set by the operator of the property.

[0013] A comfort requirement could be, for example, that acceleration or deceleration remains within a predetermined range.

[0014] An efficiency requirement could be, for example, that a distance is covered within a predetermined time.

[0015] A mission objective is, for example, reaching a destination, such as a bus stop, within a certain time.

[0016] Temporal logics, or time logics, are extensions of Boolean logic that can formally describe the spatiotemporal behavior of systems.

[0017] For the definition of STL (Signal Temporal Logic), reference is made to the publication by O. Maler and D. Nickovic, “Monitoring temporal properties of continuous signals,” in Formal Techniques, Modelling and Analysis of Timed and Fault-Tolerant Systems. Springer, 2004, pp. 152–166. The publication was available on the filing date at “https: / / www.researchgate.net / publication / 220958490_Monitoring_Temporal_Properties_of_Continuous_Signals / link / 00b4951adab57d79c4000000 / download”.

[0018] Lexicographic sorting is a method for deriving a linear order for compound objects, such as words composed of letters, from a linear order for objects, for example, alphabetically arranged letters. The eponymous example is the arrangement of words in a lexicon: they are first sorted according to their initial letters, then words with the same initial letters are sorted according to their second letter, and so on. In the context of this patent application, "lexicographical comparison" means that trajectories are lexicographically compared with one another based on compliance (or the degree of violation, if compliance is not possible) with the linearly ordered boundary conditions. If the boundary conditions are linearly sorted according to their importance, the lexicographical comparison of trajectories yields a linear sorting of the trajectories according to their conformity with the boundary conditions ordered by their importance.

[0019] Optimizing a cost function refers to the process of finding the best parameters or variables that minimize or maximize the cost function. A cost function is a mathematical representation that describes the relationship between the input variables (or parameters) and the costs associated with those inputs.

[0020] Quantifiable criteria are criteria that are evaluated using a measure of quality on a continuous scale. In contrast, it is conceivable to initially assess compliance with a traffic rule or the achievement of a mission objective using a binary approach (compliant / non-compliant or objective achieved / not achieved) and only continuously evaluate the duration and / or a measure of a violation in the event of a violation.

[0021] Computer program products typically comprise a sequence of instructions that, when the program is loaded, cause the hardware to perform a specific procedure that leads to a particular result.

[0022] The basic idea of ​​the invention is to provide linearly sorted boundary conditions for the movement of a vehicle, ordered according to their importance. To assess the vehicle's movement with regard to collision risks of trajectories, sensor data concerning the vehicle's surroundings are acquired and analyzed.

[0023] Based on the collision risks of trajectories, collision-free trajectories are identified, as are collision-prone trajectories. Collision-prone trajectories are excluded from trajectory planning. If collision-free trajectories are identified, only these are analyzed further.

[0024] It is understood that the problem of the invention can be solved both by determining collision-free trajectories and by determining and excluding non-collision-free trajectories.

[0025] If a collision-free trajectory exists, collision-free trajectories are evaluated based on the linearly ordered boundary conditions and compared lexicographically.

[0026] The plan is to provide boundary conditions for the vehicle's movement and, in particular, to define a measure that quantifies whether or not these boundary conditions are met by a given trajectory. The boundary conditions are numerous and prioritized according to their importance, i.e., they are linearly sorted.

[0027] Advantageous designs and further developments result from the further sub-claims as well as from the description with reference to the figures in the drawing.

[0028] According to a preferred embodiment of the invention, the lexicographically sorted boundary conditions include traffic rules and / or mission objectives.

[0029] Accordingly, the boundary conditions contain expressions that formulate traffic rules and / or mission objectives in a way that can be interpreted by a computer.

[0030] Alternatively or additionally, it is useful if the boundary conditions contain expressions that formulate comfort requirements and / or efficiency requirements in a way that can be interpreted by a computer.

[0031] According to a preferred embodiment of the invention, a measure for evaluating a trajectory depends on the degree or duration of non-fulfillment of a boundary condition. If the boundary conditions are sorted according to their importance, a measure can, for example, quantify the duration of non-fulfillment of the boundary condition with the highest priority.

[0032] According to a preferred embodiment of the invention, a state contains information about a position, a velocity, an acceleration and / or a time.

[0033] According to a preferred embodiment of the invention, collision-free trajectories are determined based on achievable quantities.

[0034] In trajectory planning, achievable quantities refer to the range of states or positions that a system can reach, taking into account its dynamic constraints and environmental conditions. They can be determined using an reachability analysis.

[0035] Trajectory planning involves finding a suitable path or trajectory for a system that moves from an initial state to at least one target state, taking various constraints into account. The achievable set is the set of states that the system can reach while adhering to these constraints. This embodiment of the invention provides for considering collision avoidance as a constraint on the achievable set. This means that the achievable set excludes states that would lead to a collision with objects in the environment.

[0036] For autonomous vehicles, an achievable quantity can encompass the possible positions and speeds of the vehicle, taking into account collision avoidance and optionally other constraints.

[0037] Knowing the achievable quantities is advantageous for planning safe, realistic, and physically feasible trajectories. This allows for the generation of realistic movement plans that take into account the system's physical capabilities as well as the given environmental conditions.

[0038] According to a preferred embodiment of the invention, collision-free trajectories are further evaluated using a cost function. It is particularly advantageous if the cost function continuously evaluates quantifiable criteria. In this regard, it is conceivable, for example, to evaluate criteria relating to ride comfort using a cost function.

[0039] According to a preferred embodiment of the invention, the boundary conditions are formulated using temporal logic, in particular STL. In this regard, it is conceivable, for example, to formalize traffic rules and mission objectives using STL and to use the STL robustness measure as a cost function.

[0040] The following table shows a formulation of traffic rules using STL. For example, φ_G1 ensures the necessary safety distance to vehicles ahead and that this distance is restored after k_c time steps if a vehicle ahead performs an illegal merging maneuver; φ_G2 penalizes unnecessary braking without reason; φ_G3 ensures compliance with speed limits; φ_G4 ensures the necessary traffic flow; and φ_M1 ensures that the target area is reached within the time interval l_g. Rule Definition φG1 G(in_same_lane ∧ in_front_of ∧ ¬O [0,kc] (φ cul_in ∧ O [1,1] (¬φ cul_in )) ⇒ keeps_safe_distance_prec) φ cul_in := ¬single_lane ∧ ((is_left ∧ ¬orientation_is_positive)V (¬is_left ∧ orientation_is_positive)) ∧ in_same_lane φG2 G (is_braking ⇒ ¬φ reason _t ∧ ¬φ reason_2 ) φ reason_1 :=brakes_abruptly ∧ (¬in_same_lane V ¬in_front_of) φ reason_2 :=(in_same_lane ∧ in_front_of ∧ keeps_safe_distance_prec∧ brakes_abruptly_relative) φG3 G (is_after_limit_start ∧ is_before_limit_end) ⇒ is_below_speed_limit) φG4 G (¬φ slow_leading_vehicle ⇒ φ preserves_flow ) φ slow_leading_vehicle :=in_Same_lane ∧ in_front_of ∧ is_slow φ preserves_flow :=(is_after_limit_start ∧ is_before_limit_end) ⇒ is_above_required_speed φM1 F Ig (is_after_goal_start ∧ is_before_goaLend)

[0041] It is also advantageous if the boundary conditions are formulated using linear, and especially linearized, predicates. This allows for an efficient evaluation of the STL robustness measure.

[0042] In Signal Temporal Logic (STL), a predicate is a statement or condition that is monitored over a temporal context to determine whether it is true or not. A predicate describes a property or behavior that is monitored over time.

[0043] In STL rules, a predicate is often represented by a combination of temporal operators and expressions. Linear predicates contain expressions such as... p(y,k):=1v(α(k)Tyκ−β(κ)) α:K→ℝny,β:K→ℝ, and v∈ℝ v is a rule-specific scaling factor.

[0044] Nonlinear predicates can be linearized.

[0045] According to a preferred embodiment of the invention, an optimal trajectory is determined based on comparisons of trajectories, wherein a comparison is terminated as soon as only one of the trajectories fails to meet a boundary condition, wherein the boundary conditions are examined in their lexicographical order for conformity with the trajectories.

[0046] This allows for savings in runtime and computational effort if an optimal trajectory is determined based on the lexicographical order of the boundary conditions.

[0047] It is conceivable that a lexicographically optimal trajectory can be determined by searching the reachable set for the set of trajectories that satisfy the boundary condition with the highest priority (at the first position of the linear order) or violate it only minimally if compliance is not possible. Subsequently, the remaining set is searched for the trajectories that satisfy the boundary condition with the second-highest priority (at the second position of the linear order) or violate it only minimally if compliance is not possible. This procedure can be repeated for all boundary conditions, including the boundary condition with the lowest priority (at the last position of the linear order).

[0048] This ensures that at least the highest priority boundary conditions are met or only minimally violated if the computation time is limited or the calculation is unexpectedly interrupted, since the boundary conditions are traversed from "important" to "unimportant".

[0049] According to a preferred embodiment of the invention, the lexicographically optimal trajectories are determined using numerical optimization methods as described above. For this purpose, a numerical optimization problem is successively solved for each boundary condition, and its solution set is added as an additional boundary condition for the numerical optimization problem of the subsequent boundary condition.

[0050] Alternatively or additionally, the lexicographically optimal trajectories are determined using sampling-based methods according to the procedure described above. For this purpose, a specific set of trajectories is sampled from the reachable set. Subsequently, a search is conducted within this set for those trajectories that satisfy or minimally violate the highest-priority boundary condition. Then, a search is conducted within the resulting set for those trajectories that satisfy or minimally violate the second-highest-priority boundary conditions. This procedure is repeated for all boundary conditions.

[0051] According to a preferred embodiment of the invention, a collision planner is activated if a collision-free trajectory cannot be determined. Accordingly, the method was used to determine that a collision is unavoidable. In this case, it is advantageous to limit property damage and personal injury resulting from a collision.

[0052] According to a preferred embodiment of the invention, it is determined whether no collision-free trajectory exists by checking whether the reachable quantity is an empty quantity.

[0053] Collision planners are software programs designed to limit damage in the event of a collision and are also known as "crash management software" or "collision mitigation software." This type of software is used in vehicles to take measures in the event of a collision or accident to minimize the impact or limit damage. Such systems can, for example, activate automatic braking, deploy airbags, tighten seat belts, and activate other safety mechanisms to protect occupants and reduce damage to the vehicle.

[0054] According to a preferred embodiment of the invention, the vehicle is controlled autonomously by means of the method according to one of the preceding claims.

[0055] According to a preferred embodiment of the invention, a trajectory curve between static reference points of the reference path is specified and the trajectory to be determined is designed as a velocity profile.

[0056] This aspect of the invention is particularly applicable where vehicles move along highly defined paths. Examples include rail vehicles or vehicles that have only one track available for the entire length of a route.

[0057] It is also conceivable to determine trajectories using grid graphs and a vehicle model.

[0058] The vehicle's motion can be modeled and discretized using states. Accordingly, states and state transitions that can be represented by a vehicle model are modeled in a grid graph whose nodes represent states of a trajectory at a given time and whose edges represent state transitions. To determine an optimal trajectory, an optimal path in the graph between a start and a final state is identified.

[0059] Such trajectories are evaluated based on the described procedure for fulfilling or not fulfilling the boundary conditions and based on the prioritization of the boundary conditions.

[0060] According to a preferred embodiment of the invention, paths, i.e., collision-free or non-excluded trajectories in the grid graph, are determined using a modified A* algorithm. Accordingly, unlike in a conventional A* algorithm, edge weights are not assigned to the edges, representing whether or not all boundary conditions of a state transition are met. Instead, the edges are lexicographically sorted according to the highest lexicographical order of the boundary conditions that are not met by a state transition. This corresponds to a recursive sorting of the edges of the grid graph.

[0061] A computer program product according to a method of an embodiment of the invention performs the steps of a method according to the preceding description when the computer program product runs on a computer, in particular an in-vehicle computer. When the program in question is used on a computer, the computer program product produces an effect, namely the control of an autonomous vehicle along an optimal trajectory. CONTENT OF THE DRAWINGS

[0062] The present invention will be explained in more detail below with reference to the exemplary embodiments shown in the schematic figures of the drawings. These show: Fig. 1 A schematic block diagram of an embodiment of the invention.

[0063] The accompanying drawings are intended to provide a further understanding of the embodiments of the invention. They illustrate embodiments and, in conjunction with the description, serve to explain the principles and concepts of the invention. Other embodiments and many of the advantages mentioned will become apparent with reference to the drawings. The elements of the drawings are not necessarily shown to scale.

[0064] In the figures of the drawings, identical, functionally equivalent and equally effective elements, features and components - unless otherwise stated - are each provided with the same reference symbols. DESCRIPTION OF EXAMPLES OF EXECUTION

[0065] Fig. Figure 1 shows a schematic block diagram of a procedure for planning a trajectory for a vehicle moving along a static reference bicycle, with steps S1 to S4.

[0066] In step S1, linearly sorted boundary conditions for the vehicle's motion are provided. In step S2, the vehicle's environment is captured and analyzed with regard to the collision risks of the vehicle's trajectories. In step S3, collision-free trajectories are identified, or trajectories that are not collision-free are excluded. In step S4, collision-free trajectories are evaluated based on a lexicographical comparison of the linearly sorted boundary conditions, if a collision-free trajectory exists. Reference sign S1-S4 process steps QUOTES INCLUDED IN THE DESCRIPTION

[0000] This list of documents cited by the applicant was automatically generated and is included solely for the reader's convenience. The list is not part of the German patent or utility model application. The DPMA accepts no liability for any errors or omissions. Cited non-patent literature

[0000] by O. Maler and D. Nickovic, “Monitoring temporal properties of continuous signals,” in Formal Techniques, Modeling and Analysis of Timed and Fault-Tolerant Systems. Springer, 2004, pp. 152-166

[0017] https: / / www.researchgate.net / publication / 220958490_Monitoring_Temporal_Properties_of_Continuous_Signals / link / 00b4951adab57d79c4000000 / download

[0017]

Claims

[1] Method for planning a trajectory for a vehicle, which moves in particular along a static reference path, with the following steps: - Providing (S1) linearly sorted boundary conditions for the movement of the vehicle; - Detecting and analyzing (S2) a vehicle environment based on vehicle environment-related sensor data with regard to collision risks of vehicle trajectories; - Determining (S3) collision-free trajectories or excluding trajectories that are not collision-free; - Evaluation (S4) of the collision-free or non-excluded trajectories, based on a lexicographical comparison of the linearly sorted boundary conditions. [2] The method of claim 1, wherein the lexicographically sorted constraints comprise traffic rules and / or mission objectives. [3] Method according to one of the preceding claims, wherein collision-free trajectories are determined on the basis of achievable quantities. [4] Method according to one of the preceding claims, wherein collision-free trajectories are further continuously evaluated by means of a cost function with regard to quantifiable criteria, in particular comfort. [5] Method according to one of the preceding claims, wherein the boundary conditions are formulated by means of a temporal logic, in particular by means of STL, and in particular an STL robustness measure is modeled by means of a cost function. [6] Method according to claim 5, wherein the boundary conditions are formulated by means of linear, in particular linearized, predicates. [7] Method according to one of the preceding claims, wherein an optimal trajectory is determined based on comparisons of trajectories, wherein a comparison is aborted as soon as only one of the trajectories does not satisfy a boundary condition, wherein the boundary conditions are examined in their lexicographical order for conformity with the trajectories. [8] Method according to one of the preceding claims, wherein a collision planner is started if no collision-free trajectory can be determined. [9] Computer program product with program code means for carrying out the method according to one of the preceding claims.