Systems and methods for streaming processing for autonomous vehicles
Generally, the present disclosure is directed to systems and methods for streaming processing within one or more systems of an autonomy computing system. When an update for a particular object or region of interest is received by a given system, the system can control transmission of data associated with the update as well as a determination of other aspects by the given system. For example, the system can determine based on a received update for a particular aspect and a priority classification and/or interaction classification determined for that aspect whether data associated with the update should be transmitted to a subsequent system before waiting for other updates to arrive.
1. A motion planning system for an autonomous vehicle, the motion planning system comprising:
one or more processors; and
one or more non-transitory computer-readable media that store instructions that, when executed by the one or more processors, cause the one or more processors to perform operations, the operations comprising:
receiving autonomy data descriptive of an update for a first aspect of a scene of an environment surrounding the autonomous vehicle;
determining a priority classification for the first aspect of the scene of the environment surrounding the autonomous vehicle;
determining that the priority classification meets one or more predetermined criteria;
in response to determining that the priority classification meets the one or more predetermined criteria, determining at least one of a discrete-type decision relative to navigation of the autonomous vehicle or an optimized motion plan for the autonomous vehicle; and
updating the priority classification for the first aspect of the scene of the environment in a different system other than the motion planning system.
2. The motion planning system of claim 1 , wherein the first aspect of the scene of the environment surrounding the autonomous vehicle comprises a region of interest within the scene.
3. The motion planning system of claim 1 , wherein the first aspect of the scene of the environment surrounding the autonomous vehicle comprises an object within the scene.
4. The motion planning system of claim 1 , wherein determining a priority classification for the first aspect of the scene of the environment surrounding the autonomous vehicle comprises determining a level of influence of the first aspect of the scene of the environment on a recently obtained motion plan for the autonomous vehicle.
5. The motion planning system of claim 1 , wherein:
receiving autonomy data descriptive of an update for an aspect of a scene of an environment surrounding the autonomous vehicle comprises receiving autonomy data from at least one of a perception system or a prediction system; and
wherein the different system other than the motion planning system comprises at least one of the perception system or the prediction system.
6. The motion planning system of claim 1 , wherein:
the operations further comprise determining that a second aspect of the scene has an interaction dependence relative to the first aspect of the scene;
the one or more predetermined criteria includes receiving autonomy data descriptive of an update for the second aspect of the scene.
7. The motion planning system of claim 3 , wherein determining at least one of a discrete-type decision relative to navigation of the autonomous vehicle or an optimized motion plan for the autonomous vehicle is based at least in part on the autonomy data descriptive of the update for the first aspect of the scene and the autonomy data descriptive of the update for the second aspect of the scene.
8. The motion planning system of claim 6 , wherein determining that a second aspect of the scene has an interaction dependence relative to the first aspect of the scene comprises accessing a dependence graph that identifies interacting aspects.
9. The motion planning system of claim 1 , wherein the operations further comprise determining a set of one or more discrete-type decisions from a plurality of discrete-type decisions, the set of one or more discrete-type decisions corresponding to those discrete-type decisions that are based at least in part on the first aspect of the scene of the environment surrounding the autonomous vehicle.
10. A computer-implemented method, comprising:
receiving, by a motion planning system comprising one or more computing devices, autonomy data descriptive of an update for a first aspect of a scene of an environment surrounding an autonomous vehicle;
determining, by the motion planning system, a priority classification for the first aspect of the scene of the environment surrounding the autonomous vehicle;
determining, by the motion planning system, an interaction classification for the first aspect of the scene relative to other aspects of the scene of the environment surrounding the autonomous vehicle;
determining that at least one of the priority classification or the interaction classification for the first aspect meets one or more predetermined criteria;
in response to determining that at least one of the priority classification or the interaction classification for the first aspect meets the one or more predetermined criteria, determining, by the motion planning system, at least one of a discrete-type decision relative to navigation of the autonomous vehicle or an optimized motion plan for the autonomous vehicle; and
transmitting, by the motion planning system, at least one of a priority classification update or an interaction classification update for the first aspect of the scene of the environment to a different system other than the motion planning system.
11. The computer-implemented method of claim 10 , wherein the first aspect of the scene of the environment surrounding the autonomous vehicle comprises at least one of a region of interest within the scene or an object perceived within the scene.
12. The computer-implemented method of claim 10 , wherein determining, by the motion planning system, a priority classification for the first aspect of the scene of the environment surrounding the autonomous vehicle comprises determining, by the motion planning system, a level of influence of the first aspect of the scene of the environment on a recently obtained motion plan for the autonomous vehicle.
13. The computer-implemented method of claim 10 , wherein:
receiving, by the motion planning system comprising the one or more computing devices, the autonomy data descriptive of the update for the first aspect of the scene of the environment surrounding the autonomous vehicle comprises receiving, by the motion planning system, the autonomy data from at least one of a perception system or a prediction system; and
wherein the different system other than the motion planning system comprises at least one of the perception system or the prediction system.
14. The computer-implemented method of claim 10 , further comprising:
determining, by the motion planning system, that a second aspect of the scene has an interaction dependence relative to the first aspect of the scene; and
receiving, by the motion planning system, autonomy data descriptive of an update for the second aspect of the scene.
15. The computer-implemented method of claim 14 , wherein determining, by the motion planning system, at least one of the discrete-type decision relative to navigation of the autonomous vehicle or the optimized motion plan for the autonomous vehicle is based at least in part on the autonomy data descriptive of the update for the first aspect of the scene and the autonomy data descriptive of the update for the second aspect of the scene.
16. The computer-implemented method of claim 14 , wherein determining that the second aspect of the scene has an interaction dependence relative to the first aspect of the scene comprises accessing an object dependence graph that identifies interacting aspects.
17. The computer-implemented method of claim 10 , further comprising determining, by the motion planning system, a set of one or more discrete-type decisions from a plurality of discrete-type decisions, the set of one or more discrete-type decisions corresponding to those discrete-type decisions that are based at least in part on the first aspect of the scene of the environment surrounding the autonomous vehicle.
18. An autonomous vehicle, comprising:
one or more processors; and
one or more non-transitory computer-readable media that store instructions that, when executed by the one or more processors, cause the one or more processors to perform operations, the operations comprising:
receiving autonomy data descriptive of an update for a first aspect of a scene of an environment surrounding the autonomous vehicle;
determining a priority classification for the first aspect of the scene of the environment surrounding the autonomous vehicle;
determining an interaction classification for the first aspect of the scene relative to other aspects of the scene of the environment surrounding the autonomous vehicle;
determining that at least one of the priority classification or the interaction classification for the first aspect meets one or more predetermined criteria;
in response to determining that at least one of the priority classification and the interaction classification for the first aspect meets one or more predetermined criteria, determining at least one of a discrete-type decision relative to navigation of the autonomous vehicle or an optimized motion plan for the autonomous vehicle; and
updating at least one of the priority classification or the interaction classification for the first aspect of the scene of the environment in a different system other than the motion planning system.
19. A motion planning system for an autonomous vehicle, the motion planning system comprising:
one or more processors; and
one or more non-transitory computer-readable media that store instructions that, when executed by the one or more processors, cause the one or more processors to perform operations, the operations comprising:
receiving autonomy data descriptive of an update for a first aspect of a scene of an environment surrounding the autonomous vehicle;
determining that a second aspect of the scene has an interaction dependence relative to the first aspect of the scene by accessing a dependence graph that identifies interacting aspects;
determining a priority classification for the first aspect of the scene of the environment surrounding the autonomous vehicle;
determining that the priority classification meets one or more predetermined criteria, the one or more predetermined criteria including receiving autonomy data descriptive of an update for the second aspect of the scene; and
in response to determining that the priority classification meets the one or more predetermined criteria, determining at least one of a discrete-type decision relative to navigation of the autonomous vehicle or an optimized motion plan for the autonomous vehicle.
20. A computer-implemented method, comprising:
receiving, by a motion planning system comprising one or more computing devices, autonomy data descriptive of an update for a first aspect of a scene of an environment surrounding an autonomous vehicle;
determining, by the motion planning system, that a second aspect of the scene has an interaction dependence relative to the first aspect of the scene based on accessing an object dependence graph that identifies interacting aspects;
determining, by the motion planning system, a priority classification for at least one of the first aspect or the second aspect of the scene of the environment surrounding the autonomous vehicle;
determining, by the motion planning system, that the priority classification meets one or more predetermined criteria, the one or more predetermined criteria including receiving autonomy data descriptive of an update for the at least one of the first aspect or the second aspect of the scene; and
in response to determining that the priority classification meets the one or more predetermined criteria, determining at least one of a discrete-type decision relative to navigation of the autonomous vehicle or an optimized motion plan for the autonomous vehicle.