Systems and methods for disengagement prediction and triage assistant

1. A method comprising, by a computing system:

receiving perception data associated with a perceived scenario encountered by a vehicle while operating in an autonomous driving mode;

identifying the perceived scenario based at least on the perception data;

generating a performance metric associated with a vehicle navigation plan to navigate the vehicle in accordance with the perceived scenario, the performance metric indicating a predicted ability of the vehicle to navigate the perceived scenario along a planned trajectory, the perceived scenario representing different arrangements of elements including i) a known configuration of identified objects that is not supported by a planning system and ii) an arrangement of unidentifiable objects;

in response to a determination that the performance metric associated with the vehicle navigation plan fails to satisfy one or more criteria for navigating the vehicle in accordance with the perceived scenario, triggering a disengagement operation related to disengaging the vehicle from the autonomous driving mode; and

generating a disengagement record associated with the disengagement operation, the disengagement record comprising information associated with the perceived scenario encountered by the vehicle related to the disengagement operation.

2. The method of claim 1 , wherein the disengagement operation comprises performing at least one of:

sending an alert message to an operator that the vehicle is going to disengage from the autonomous driving mode;

slowing down a speed of the vehicle; and

automatically disengaging the vehicle from the autonomous driving mode.

3. The method of claim 1 , wherein, prior to the determination that the performance metric fails to satisfy the one or more criteria, the method further comprises:

receiving additional perception data associated with the perceived scenario encountered by the vehicle;

based on the additional perception data, determining that the perceived scenario cannot be identified; and

in response to determining that the perceived scenario cannot be identified, triggering the vehicle to disengage from the autonomous driving mode.

4. The method of claim 3 , further comprising;

updating a database corresponding to an operational design domain of the vehicle to include the perceived scenario that is unidentifiable;

receiving feedback associated with the perceived scenario; and

using the feedback to identify the perceived scenario.

5. The method of claim 1 , wherein the perceived scenario is identified based at least on determining whether the perception data associated with the perceived scenario corresponds to data associated with one or more previously identified scenarios that are defined by an operational design domain of the vehicle.

6. The method of claim 5 , wherein, prior to identifying the perceived scenario based at least on the perception data, the method further comprises:

identifying a known scenario based on further perception data received by the vehicle, wherein the known scenario is excluded from a plurality of pre-determined scenarios corresponding to the operational design domain;

automatically disengaging the vehicle from the autonomous driving mode in accordance with the known scenario being excluded from the plurality of pre-determined scenarios; and

automatically generating a new disengagement record based on the known scenario of the vehicle.

7. The method of claim 1 , wherein the disengagement record is automatically generated to include a pre-filled template, and wherein the pre-filled template includes the perceived scenario.

8. The method of claim 7 , further comprising:

receiving feedback information associated with the pre-filled template;

determining, based on the feedback information, that the perceived scenario included in the pre-filled template is incorrect; and

replacing, in the pre-filled template, the perceived scenario with another scenario.

9. The method of claim 1 , wherein the one or more criteria comprise one or more of:

a distance between the vehicle and another agent or object being greater than or equal to a pre-determined threshold distance;

a moving speed being within a pre-determined speed range;

an acceleration or deceleration being within a pre-determined acceleration range;

a turning radius being within a pre-determined radius range; or

a relative speed with respect to a nearby object being within a pre-determined relative speed range.

10. The method of claim 1 , further comprising:

determining a probability score for each of a plurality of pre-determined scenarios with respect to each of a plurality of scenario categories, wherein the probability score indicates a probability level of that scenario belonging to an associated scenario category, and wherein the plurality of pre-determined scenarios are classified into the plurality of scenario categories based on associated probability scores.

11. The method of claim 1 , further comprising:

capturing vehicle operation data after the vehicle is disengaged from the autonomous driving mode; and

including the vehicle operation data in the disengagement record associated with the perceived scenario, wherein the vehicle operation data includes at least one of a vehicle acceleration, a vehicle velocity, or a vehicle trajectory.

12. The method of claim 1 , further comprising:

determining an importance score for each scenario of a plurality of scenarios concurrently encountered by the vehicle, wherein the importance score is based on a severeness level associated with each scenario; and

selecting a scenario from the plurality of scenarios concurrently encountered by the vehicle in response to a determination that the scenario is associated with a highest importance score among the plurality of scenarios concurrently encountered by the vehicle.

13. The method of claim 1 , further comprising:

generating an image raster for an image associated with the perceived scenario; and

categorizing a disengagement event associated with the disengagement operation based on an analysis on pixels of the image raster, wherein the disengagement record is generated based on the disengagement event.

14. The method of claim 1 , wherein the disengagement record comprises a text content describing the perceived scenario, further comprising:

analyzing the text content of the disengagement record using a text recognition algorithm; and

assigning, based on an analysis result of the text recognition algorithm, the disengagement record to a particular team for triaging the perceived scenario related to the disengagement operation included in the disengagement record.

15. The method of claim 14 , wherein the disengagement record comprises an image associated with the perceived scenario, further comprising:

analyzing the image of the disengagement record using an image classification algorithm; and

assigning, based on an analysis result of the image classification algorithm, the disengagement record to the particular team for triaging the disengagement record.

16. The method of claim 1 , further comprising:

in response to the perceived scenario being an un-identifiable scenario, automatically disengaging the vehicle from the autonomous driving mode;

generating a new disengagement record including the perception data associated with the un-identifiable scenario; and

associating the perception data included in the new disengagement record to an identifiable scenario based on an operator feedback.

17. The method of claim 16 , further comprising:

in response to a subsequent scenario being the identifiable scenario, navigating the vehicle in the autonomous driving mode in accordance with a navigation plan generated by the vehicle.

18. The method of claim 1 , further comprising:

extracting scenario information and vehicle performance information from a plurality of disengagement records associated with a plurality of scenarios; and

training the planning system based on the scenario information and vehicle performance information, wherein the planning system, once trained, generates vehicle navigation plans that allow the vehicle to have a performance metric meeting one or more pre-determined criteria under the plurality of scenarios.

19. One or more non-transitory computer-readable storage media embodying software that is operable, when executed by one or more processors of a computing system, to:

receive perception data associated with a perceived scenario encountered by a vehicle while operating in an autonomous driving mode;

identify the perceived scenario based at least on the perception data;

generate a performance metric associated with a vehicle navigation plan to navigate the vehicle in accordance with the perceived scenario, the performance metric indicating a predicted ability of the vehicle to navigate the perceived scenario along a planned trajectory, the perceived scenario representing different arrangements of elements including i) a known configuration of identified objects that is not supported by a planning system and ii) an arrangement of unidentifiable objects;

in response to a determination that the performance metric associated with the vehicle navigation plan fails to satisfy one or more criteria for navigating the vehicle in accordance with the perceived scenario, trigger a disengagement operation related to disengaging the vehicle from the autonomous driving mode; and

generate a disengagement record associated with the disengagement operation, the disengagement record comprising information associated with the perceived scenario encountered by the vehicle related to the disengagement operation.

20. A system comprising:

one or more non-transitory computer-readable storage media embodying instructions; and

one or more processors coupled to the one or more non-transitory computer-readable storage media and operable to execute the instructions to:

receive perception data associated with a perceived scenario encountered by a vehicle while operating in an autonomous driving mode;

identify the perceived scenario based at least on the perception data;

generate a performance metric associated with a vehicle navigation plan to navigate the vehicle in accordance with the perceived scenario, the performance metric indicating a predicted ability of the vehicle to navigate the perceived scenario along a planned trajectory, the perceived scenario representing different arrangements of elements including i) a known configuration of identified objects that is not supported by a planning system and ii) an arrangement of unidentifiable objects;

in response to a determination that the performance metric associated with the vehicle navigation plan fails to satisfy one or more criteria for navigating the vehicle in accordance with the perceived scenario, trigger a disengagement operation related to disengaging the vehicle from the autonomous driving mode; and

generate a disengagement record associated with the disengagement operation, the disengagement record comprising information associated with the perceived scenario encountered by the vehicle related to the disengagement operation.