Providing actionable uncertainties in autonomous vehicles
Systems and methods are provided for detecting objects of interest. A computing system can input sensor data to one or more first machine-learned models associated with detecting objects external to an autonomous vehicle. The computing system can obtain as an output of the first machine-learned models, data indicative of one or more detected objects. The computing system can determine data indicative of at least one uncertainty associated with the one or more detected objects and input the data indicative of the one or more detected objects and the data indicative of the at least one uncertainty to one or more second machine-learned models. The computing system can obtain as an output of the second machine-learned models, data indicative of at least one prediction associated with the one or more detected objects. The at least one prediction can be based at least in part on the detected objects and the uncertainty.
1. A computer-implemented method, comprising:
determining, by a perception system of an autonomous vehicle, data indicative of an uncertainty associated with a detected object in an environment of the autonomous vehicle at a current cycle, wherein the data indicative of the uncertainty is determined based on data indicative of the detected object and history data associated with the detected object at a previous cycle;
providing the data indicative of the detected object and the data indicative of the uncertainty associated with the detected object to one or more machine-learned models within a motion planning system of the autonomous vehicle; and
generating, by the motion planning system, a motion plan for the autonomous vehicle based at least in part on the uncertainty.
2. The computer-implemented method of claim 1 , wherein the data indicative of the uncertainty indicates at least one probability distribution with respect to the detected object.
3. The computer-implemented method of claim 1 , wherein the data indicative of the uncertainty indicates an uncertainty with respect to a detection of the detected object.
4. The computer-implemented method of claim 1 , wherein the data indicative of the uncertainty indicates an uncertainty with respect to a classification of the detected object.
5. The computer-implemented method of claim 1 , wherein:
the data indicative of the detected object includes data indicative of a classification of the detected object; and
the data indicative of the uncertainty includes a probability distribution associated with the classification.
6. The computer-implemented method of claim 1 , wherein the data indicative of the uncertainty indicates an uncertainty with respect to state data indicative of a state of the detected object.
7. The computer-implemented method of claim 1 , wherein the history data comprises information associated with one or more detections of the detected object at the previous cycle.
8. The computer-implemented method of claim 1 , wherein the history data comprises a length of time during which the detected object has been tracked.
9. The computer-implemented method of claim 1 , wherein the history data comprises a number of cycles during which the detected object has been tracked.
10. The computer-implemented method of claim 1 , wherein the history data comprises at least one of a classification or a state of the detected object from the previous cycle.
11. An autonomous vehicle control system for controlling an autonomous vehicle, the autonomous vehicle control system comprising:
one or more processors; and
one or more tangible, non-transitory, computer-readable media that store instructions that are executable to cause the autonomous vehicle control system to perform operations, the operations comprising:
determining, by a perception system of the autonomous vehicle control system, data indicative of an uncertainty associated with a detected object in an environment of the autonomous vehicle at a current cycle, wherein the data indicative of the uncertainty is determined based on data indicative of the detected object and history data associated with the detected object at a previous cycle;
providing the data indicative of the detected object and the data indicative of the uncertainty associated with the detected object to one or more machine-learned models within a motion planning system of the autonomous vehicle control system; and
generating, by the motion planning system, a motion plan for the autonomous vehicle based at least in part on the uncertainty.
12. The autonomous vehicle control system of claim 11 , wherein the data indicative of the uncertainty indicates at least one probability distribution with respect to the detected object.
13. The autonomous vehicle control system of claim 11 , wherein the data indicative of the uncertainty indicates an uncertainty with respect to at least one of a detection of the detected object or a classification of the detected object.
14. The autonomous vehicle control system of claim 11 , wherein the data indicative of the uncertainty indicates an uncertainty with respect to state data indicative of a state of the detected object.
15. The autonomous vehicle control system of claim 11 , wherein the history data comprises information associated with one or more detections of the detected object at the previous cycle.
16. The autonomous vehicle control system of claim 11 , wherein the history data comprises a length of time during which the detected object has been tracked.
17. The autonomous vehicle control system of claim 11 , wherein the history data comprises a number of cycles during which the detected object has been tracked.
18. The autonomous vehicle control system of claim 11 , wherein the history data comprises at least one of a classification or a state of the detected object from the previous cycle.
19. An autonomous vehicle, comprising:
one or more processors; and
one or more tangible, non-transitory, computer-readable media that store instructions that are executable to cause the one or more processors to perform operations, the operations comprising:
determining, by a perception system of the autonomous vehicle, data indicative of an uncertainty associated with a detected object in an environment of the autonomous vehicle at a current cycle, wherein the data indicative of the uncertainty is determined based on data indicative of the detected object and history data associated with the detected object at a previous cycle;
providing the data indicative of the detected object and the data indicative of the uncertainty associated with the detected object to one or more machine-learned models within a motion planning system of the autonomous vehicle; and
generating, by the motion planning system, a motion plan for the autonomous vehicle based at least in part on the uncertainty.
20. The autonomous vehicle of claim 19 , wherein the history data comprises at least one of a length of time or a number of cycles during which the detected object has been tracked.