JPWO2018162521A5 - - Google Patents

Description

TECHNICAL FIELD This disclosure relates generally to autonomous vehicles and, more particularly, to action planning systems and methods for autonomous vehicles.

BACKGROUND Unless otherwise indicated herein, the subject matter described in this section is not prior art to the claims of this application, and their inclusion in this section is not an admission that they are. do not have.

Conventional action planners are based on a finite state machine (FSM) approach in which an autonomous vehicle decides to take specific actions based on heuristic transition conditions. However, in this FSM approach, the autonomous vehicle can only react passively to changes in the surrounding environment of the autonomous vehicle, and how a given action will affect future traffic conditions. No future outlook is provided on how to progress to Figures 1 and 2 show the "lane following" behavior with one target vehicle or between two target vehicles on adjacent lanes evaluated by a prior art action planner using the FSM approach. Various possible actions are shown, such as a "lane change" action to the left.

SUMMARY A summary of certain embodiments disclosed herein follows. It should be understood that these aspects are presented merely to provide the reader with a brief overview of these particular embodiments and are not intended to limit the scope of the disclosure. should. Rather, the disclosure can encompass various aspects that may not be described below.

Embodiments of the present disclosure relate to a non-transitory computer-readable storage medium having stored thereon a computer program for evaluating responses and interactions with one or more vehicles. , the computer program searches through a tree structure of combinations of possible sequences of actions for the host vehicle and captures reactions and interactions with one or more vehicles. contains a routine consisting of a set of instructions to be executed by

Another aspect of embodiments of the present disclosure is a method in which, by one or more processors, estimating a future surrounding environment for an autonomous vehicle and generating a possible trajectory for the autonomous vehicle; Based on current local traffic conditions, it predicts the motion and reaction of each dynamic obstacle in the future surroundings of the autonomous vehicle, generating predictions iteratively over multiple time steps. The method further decouples the action decision temporal resolution from the iterative prediction resolution by the one or more processors.

Another aspect of embodiments of the present disclosure is one or more processors and one or more non-transitory computer-readable storage media on which computer programs for use by the one or more processors are stored. , wherein the computer program causes one or more processors to estimate a future surrounding environment of the autonomous vehicle, generate a possible trajectory for the autonomous vehicle, and determine a current local Based on traffic conditions, the motion and reaction of each dynamic obstacle in the future surroundings of the autonomous vehicle is predicted, and predictions are iteratively generated over multiple time steps.

BRIEF DESCRIPTION OF THE FIGURES These and other features, aspects and advantages of this disclosure will be described in the following specific illustrative examples with reference to the accompanying drawings, in which like numerals represent like techniques throughout the several views. It will be better understood upon reading the detailed description of the preferred embodiments.

1 is a simplified diagram illustrating one possible lane following behavior using an action planner for autonomous vehicles implementing FSMs according to the prior art; FIG. 1 is a simplified diagram illustrating one possible lane change behavior using an action planner for autonomous vehicles implementing FSMs according to the prior art; FIG. 1 illustrates an automated driving system according to one embodiment of the present disclosure; FIG. FIG. 2 is a simplified diagram showing a multi-lane road and map bird's eye view with multiple adjacent vehicles in the vicinity of the autonomous vehicle according to the above embodiments of the present disclosure; 1 is a diagram of an autonomous vehicle driving autonomously in the presence of multiple adjacent vehicles in the vicinity, according to the described embodiment of the present disclosure; FIG. 1 is a diagram of an autonomous vehicle driving autonomously in the presence of multiple adjacent vehicles in the vicinity, according to the described embodiment of the present disclosure; FIG. 1 is a diagram of an autonomous vehicle driving autonomously in the presence of multiple adjacent vehicles in the vicinity, according to the described embodiment of the present disclosure; FIG. 1 is a diagram of an autonomous vehicle driving autonomously in the presence of multiple adjacent vehicles in the vicinity, according to the described embodiment of the present disclosure; FIG.

DETAILED DESCRIPTION The following description is presented to enable any person skilled in the art to make and use the described embodiments, and is provided in the context of particular applications and needs thereof. Various modifications to the described embodiments will be apparent to those skilled in the art, and the general principles defined herein may be modified in other ways without departing from the spirit and scope of the described embodiments. can be applied to the embodiments and applications of Accordingly, the described embodiments should not be limited to the illustrated embodiments, but should be accorded a very broad scope consistent with the principles and features disclosed herein.

3A and 3B illustrate an automated driving system 100 according to one aspect of the present disclosure. As illustrated in FIG. 3A, operating system 100 may be incorporated into a vehicle 114 or mechanical device, or may be incorporated into any suitable portable or mobile device/container. . Certain aspects of the present disclosure are particularly useful in combination with certain types of vehicles, however, including automobiles, trucks, motorcycles, buses, boats, sport utility vehicles (SUVs), bicycles, airplanes, helicopters. , lawn mowers, recreational vehicles, amusement park vehicles, streetcars, golf carts, trains, trolleys, ultralight aircraft, and the like. However, vehicles are not limited to these. A vehicle may have one or more automated driving systems. The mechanical device may be any type of device including a mobile phone, laptop computer, tablet, or wearable device such as a watch, glasses, goggles, or any suitable portable device. However, the device is not limited to these. As illustrated in FIG. 3A, automated driving system 100 includes processor 108 , computer readable media 110 , and communication module 112 . A route planning module 102 , an action planning module 104 and a trajectory planning module 106 may be provided in the automated driving system 100 .

Although route planning module 102, action planning module 104, trajectory planning module 106, processor 108, computer readable medium 110, and communication module 112 reside within the same block 100, one of ordinary skill in the art would: Route planning module 102, action planning module 104, trajectory planning module 106, processor 108, computer readable medium 110, and communication module 112 may or may not be housed in the same block 100. You will understand that you may In various aspects described herein, processor 108 , computer-readable medium 110 , and communication module 112 may be incorporated into a computer located external to automated driving system 100 . In other aspects, some of the processes described herein run on a processor located within vehicle 114 and other processes run on a remote processor located within machine 116. .

Computer readable medium 110 stores information that may be accessed by route planning module 102, action planning module 104, trajectory planning module 106, and processor 108, including route planning module 102; Included are computer-executable instructions that may be executed by or otherwise used by the action planning module 104, the trajectory planning module 106, and the processor 108. Processor 108 may be any conventional processor. Alternatively, processor 108 may be a dedicated device such as an ASIC.

The communication module 112 communicates directly or indirectly by wire/wireless with other computers or networks, such as one or more electronic control units, one or more processors. Other sensing devices such as sensors, user interfaces such as touch displays, mice, keyboards, voice inputs, cameras, and other suitable computer-implemented modules may be incorporated into system 100, or , can be communicatively connected to the system 100 .

The process of route processing performed in route planning module 102, action processing performed in action planning module 104, and trajectory processing performed in trajectory planning module 106, during a configurable time period T, A candidate route, action and trajectory is generated that the ego vehicle can follow as it passes through the . In some aspects, route planning module 102, action planning module 104, and trajectory planning module 106 are computer-executable instructions programmed in processor 108 to generate routes, actions, and trajectory candidates. Processing, behavior processing and trajectory processing can be performed in processor 108 .

As illustrated in FIG. 3B, the action planning module 104 performs a definitive horizontal search of a tree of combinations of possible action sequences for the autonomous vehicle, this horizontal search and an iterative full - scale search. In combination with ambient environment prediction, it can operate to capture reactions and interactions with other traffic participants. For example, action planning module 104 performs an optimal graph search to efficiently search a tree structure of possible actions of the ego vehicle for a limited time period or planning range. The aforementioned time period includes a starting point and an ending point at which this period can be evaluated. Also, the aforementioned time period is necessary to constrain the size of the search space to achieve efficient replanning.

Graph search techniques are used to be able to reach the end point quickly so that the optimal path can be found without going through all possible paths or edges. The graph search technique may be a horizontal search technique. However, other suitable graph search techniques can be used. Pre-computed static costs are used to estimate the utility cost for a given ego vehicle position at the end of the planning period. In one embodiment, the trajectory planning module 106 includes a graph search technique that the ego vehicle 114 can follow as it traverses its surroundings for a configurable period of time T. In other embodiments, graph search techniques are stored in action plan module 104 . The generated trajectory is then stored on computer readable medium 110 . Further details on action planning using iterative action search surround prediction are provided below.

4A-4D are diagrams of an autonomous vehicle 114 autonomously driving on a two-lane roadway 200 in the vicinity of a plurality of adjacent vehicles 160, 162, according to one illustrated embodiment of the present disclosure. As illustrated in FIG. 4A, vehicle 160 is currently following the same potential path 202 on the left lane of two-lane road 200 . Autonomous vehicle 114 is currently located in the right lane of dual-lane road 200 . A boundary line 204 separates the left and right lanes. The autonomous vehicle 114 or "self-vehicle" described herein enumerates all possible behaviors that can initiate lane following behavior and switch at the next time step. In one aspect, ego vehicle 114 continues lane following behavior and follows the same potential path 208 on the right lane of two-lane road 200, as illustrated in FIG. 4A. In another aspect, the automated driving system 100 of the ego-vehicle 114 can switch from lane-following behavior to lane-changing behavior if the lane-changing behavior has a lower initial cost, and in FIG. 4B is shown as the planned path 210 for the ego vehicle 114 crosses the boundary line 204 and approaches the probable path 202 of the adjacent vehicle 160 . The action planning module 104, which is programmed with horizontal searching, looks for the possibility of switching from lane following behavior to lane change behavior. In one embodiment, an iterative global environment prediction is used whereby the horizontal search is performed without explicitly comprehensively searching for possible behaviors of other vehicles 160, 162. Reactions of other vehicles 160, 162 and interactions with other vehicles 160, 162 can be evaluated. Horizontal search tractability is maintained by iterative global environment prediction. Each time a new behavior is evaluated in the horizontal graph search, ego-vehicle 114 assumes that it follows the selected behavior, e.g., lane follow behavior, lane change behavior, or lane change abort behavior. A prediction process is performed at the next time step to estimate the future surroundings of . This prediction process involves the generation of possible trajectories for the ego-vehicle 114 based on the current local traffic conditions of the ego-vehicle 114, followed by the dynamic trajectory of each obstacle in the ego-vehicle's 114 surrounding environment. motion and response prediction. Additionally, the prediction process includes generating predictions of how other vehicles are likely to react to current traffic conditions repeatedly over short time steps. A horizontal graph search can model interactions between the ego-vehicle 114 and other traffic participants without requiring any active planning for the other vehicles 160,162. This iterative, passive prediction is much less costly and scales much better than attempting to implement joint active planning for the ego-vehicle 114 and other traffic participants. be able to.

In one aspect, the ego-vehicle 114 can continue the lane change behavior or can switch from lane change behavior to stop lane change behavior. 4C, ego-vehicle 114 continues its lane change behavior by entering probable route 202 of adjacent vehicle 160, thereby merging planned route 210 with probable route 202. As shown in FIG. The ego-vehicle 114 decides not to change lanes or, as illustrated in FIG. 4D, changes lanes due to the approach of an adjacent vehicle 160 coming from behind onto the probable path 202. is not likely to be possible, the automatic driving system 100 of the host vehicle 114 switches from the lane change action to the lane change action stop. A planned path 212 for ego vehicle 114 is shown crossing boundary line 204 to approach the left lane of two-lane road 200 .

Referring again to FIG. 4B, at time step t1, ego vehicle 114 produces a lane change trajectory, however, other vehicles 160 in the left lane have not yet responded. At time step t2, the vehicle continues its lane change behavior as shown in FIG. 4C or switches to stop lane change behavior as shown in FIG. partially enters the left lane, the vehicle 160 in the left lane will react accordingly. In some embodiments, the iterative prediction temporal resolution is effectively decoupled from the action search temporal resolution, as the action search is restricted to evaluate only switching behaviors in a small subset of time steps. . This greatly improves computational efficiency and allows longer planning horizons to be evaluated, resulting in more intelligent action decisions. A limited horizontal search of the ego - vehicle's possible behavior tree for intelligent autonomous driving decision-making significantly improves decision-making performance. In one embodiment, the ego-vehicle 114 explicitly evaluates the expected utility of actions in future time steps. In another embodiment, the ego vehicle 114 understands and incrementally predicts the behavior of other agents based on the expected future state of the ego vehicle 114 itself. The ego-vehicle 114 explicitly evaluates and compares the different possible timings of each behavioral transition in order to maximize safety, comfort and progress toward its goals. Through iterative prediction, how other dynamic obstacles will react to the ego-vehicle 114 without having to implement a joint prediction plan for each dynamic obstacle in the environment, and Behavioral exploration is implemented to understand how the ego vehicle 114 interacts. A limited branching point decouples the action decision temporal resolution from the iterative prediction resolution, which allows for more efficient exploration of longer planning horizons.

It should be understood that the above-described embodiments have been presented by way of example and that these embodiments are susceptible to various modifications and alternative forms. Further, it is understood that the claims are not limited to the particular forms disclosed, but rather cover all modifications, equivalents and alternatives falling within the spirit and scope of this disclosure. want to be

Embodiments within the scope of the present disclosure may also include non-transitory computer-readable storage media or machine-readable media for carrying or having computer-executable instructions or data structures stored thereon. Such non-transitory computer-readable storage media or machine-readable media may be any available media that can be accessed by a general purpose or special purpose computer. By way of example, and not limitation, such non-transitory computer-readable or machine-readable storage media may include RAM, ROM, EEPROM, CD-ROM or other optical disk storage devices, magnetic disk storage devices or other magnetic storage devices; It may include any other medium that can be used for carrying or storing desired program code means in the form of computer-executable instructions or data structures. Combinations of the above should also be included within the scope of non-transitory computer-readable storage media or machine-readable media.

Embodiments may also be practiced in distributed computing environments where tasks are performed by local and remote processing devices that are linked through a communications network (either by hardwired links, wireless links, or a combination thereof). good.

Computer-executable instructions include, for example, instructions and data that cause a general purpose computer, special purpose computer, or special purpose processing device to perform a particular function or group of functions. Computer-executable instructions also include program modules that are executed by computers in stand-alone or network environments. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform particular tasks or implement particular abstract data types. Computer-executable instructions, associated data structures, and program modules represent examples of the program code means for executing steps of the methods disclosed herein. Any particular sequence of such executable instructions or associated data structures represents examples of corresponding acts for implementing the functionality described in such steps.

While this patent has been described with reference to various embodiments, it should be understood that these embodiments are exemplary and the scope of the disclosure is not limited thereto. Many variations, modifications, additions and improvements are possible. More generally, embodiments in accordance with this patent have been described either in context or in specific embodiments. Functionality may be separate, or combined into blocks, or referred to with different terms, in different embodiments of the present disclosure. These and other variations, modifications, additions and improvements may fall within the scope of this disclosure as defined in the following claims.

Claims (2)

predicting, by one or more processors, the future ambient environment of the autonomous vehicle;
generating possible trajectories for the autonomous vehicle by performing, by one or more of the processors, a tree -structured graph search of combinations of possible sequences of actions for the autonomous vehicle;
predicting, by one or more of the processors, dynamic obstacle movement and reaction in the future global environment of the autonomous vehicle based on current local traffic conditions;
Iteratively over multiple time steps, by one or more of the processors, predicting the future ambient environment of the autonomous vehicle, generating a possible trajectory for the autonomous vehicle, and each of the dynamic obstacles. perform motion and reaction predictions of
method involving In systems for autonomous vehicles,
one or more processors;
one or more non-transitory computer-readable storage media storing a computer program for use by one or more of the processors;
and
The computer program, on one or more of the processors,
predict the future surroundings of autonomous vehicles,
generating possible trajectories for the autonomous vehicle by performing a tree -structured graph search of combinations of possible action sequences for the autonomous vehicle;
predict dynamic obstacle movement and reaction in the future global environment of the autonomous vehicle based on current local traffic conditions;
Iteratively over a plurality of time steps, causing prediction of the full future surrounding environment of the autonomous vehicle, generation of possible trajectories for the autonomous vehicle, and prediction of motion and reaction of each of the dynamic obstacles. ,
Systems for autonomous vehicles.