Optimization and selection of compute paths for an autonomous vehicle

1. A method comprising:

identifying at least one compute path for processing input sensor data by an autonomous vehicle, wherein the at least one compute path includes a set of nodes that are interconnected by one or more edges;

implementing a first test iteration of at least one simulation scenario for testing the at least one compute path, wherein the first test iteration of the at least one simulation scenario applies a first artificial delay to at least one edge from the one or more edges by configuring an artificial software workload associated with the at least one edge to model a processing delay;

determining a result of the first test iteration based on at least one of a safety parameter and a comfort parameter, wherein the safety parameter and the comfort parameter are based on a result of the at least one simulation scenario;

determining, based on the result of the first test iteration, a maximum tolerable artificial delay for the at least one edge;

configuring the autonomous vehicle for real-world operation based on the determined maximum tolerable artificial delay by setting a safe stop safety threshold used by the autonomous vehicle to initiate a safe stop maneuver during navigation; and

initiating the safe stop maneuver based on the safe stop safety threshold.

2. The method of claim 1 , further comprising:

determining that the safety parameter or the comfort parameter is less than a threshold value; and

in response, implementing a second test iteration of the at least one simulation scenario for testing that at least one compute path, wherein the second test iteration of the at least one simulation scenario applies a second artificial delay to the at least one edge from the one or more edges, wherein the second artificial delay is greater than the first artificial delay.

3. The method of claim 1 , further comprising:

disabling one or more alternate compute paths for processing the input sensor data by the autonomous vehicle.

4. The method of claim 1 , further comprising:

identifying a plurality of compute paths for processing the input sensor data by the autonomous vehicle, wherein the at least one compute path is part of the plurality of compute paths; and

selectively enabling one or more compute paths from the plurality of compute paths during the first test iteration of the at least one simulation scenario.

5. The method of claim 4 , further comprising:

identifying a preferred compute path from the plurality of compute paths for processing the input sensor data in a real-world environment corresponding to the at least one simulation scenario.

6. The method of claim 1 , wherein the set of nodes includes at least one of a perception stack, a prediction stack, a planning stack, and a control stack.

7. An apparatus comprising:

at least one memory comprising instructions; and

at least one processor configured to execute the instructions and cause the at least one processor to:

identify at least one compute path for processing input sensor data by an autonomous vehicle, wherein the at least one compute path includes a set of nodes that are interconnected by one or more edges;

implement a first test iteration of at least one simulation scenario for testing the at least one compute path, wherein the first test iteration of the at least one simulation scenario applies a first artificial delay to at least one edge from the one or more edges by configuring an artificial software workload associated with the at least one edge to model a processing delay;

determine a result of the first test iteration based on at least one of a safety parameter and a comfort parameter, wherein the safety parameter and the comfort parameter are based on a result of the at least one simulation scenario;

determine, based on the result of the first test iteration, a maximum tolerable artificial delay for the at least one edge;

configure the autonomous vehicle for real-world operation based on the determined maximum tolerable artificial delay by setting a safe stop safety threshold used by the autonomous vehicle to initiate a safe stop maneuver during navigation; and

initiate the safe stop maneuver based on the safe stop safety threshold.

8. The apparatus of claim 7 , wherein the at least one processor is further configured to:

determine that the safety parameter or the comfort parameter is less than a threshold value; and

in response, implement a second test iteration of the at least one simulation scenario for testing that at least one compute path, wherein the second test iteration of the at least one simulation scenario applies a second artificial delay to the at least one edge from the one or more edges, wherein the second artificial delay is greater than the first artificial delay.

9. The apparatus of claim 7 , wherein the at least one processor is further configured to:

disable one or more alternate compute paths for processing the input sensor data by the autonomous vehicle.

10. The apparatus of claim 7 , wherein the at least one processor is further configured to:

identify a plurality of compute paths for processing the input sensor data by the autonomous vehicle, wherein the at least one compute path is part of the plurality of compute paths; and

selectively enable one or more compute paths from the plurality of compute paths during the first test iteration of the at least one simulation scenario.

11. The apparatus of claim 10 , wherein the at least one processor is further configured to:

identify a preferred compute path from the plurality of compute paths for processing the input sensor data in a real-world environment corresponding to the at least one simulation scenario.

12. The apparatus of claim 7 , wherein the set of nodes includes at least one of a perception stack, a prediction stack, a planning stack, and a control stack.

13. A non-transitory computer-readable storage medium having stored thereon instructions which, when executed by one or more processors, cause the one or more processors to:

identify at least one compute path for processing input sensor data by an autonomous vehicle, wherein the at least one compute path includes a set of nodes that are interconnected by one or more edges;

implement a first test iteration of at least one simulation scenario for testing the at least one compute path, wherein the first test iteration of the at least one simulation scenario applies a first artificial delay to at least one edge from the one or more edges by configuring an artificial software workload associated with the at least one edge to model a processing delay;

determine a result of the first test iteration based on at least one of a safety parameter and a comfort parameter, wherein the safety parameter and the comfort parameter are based on a result of the at least one simulation scenario;

determine, based on the result of the first test iteration, a maximum tolerable artificial delay for the at least one edge;

configure the autonomous vehicle for real-world operation based on the determined maximum tolerable artificial delay by setting a safe stop safety threshold used by the autonomous vehicle to initiate a safe stop maneuver during navigation; and

initiate the safe stop maneuver based on the safe stop safety threshold.

14. The non-transitory computer-readable storage medium of claim 13 , comprising further instructions which, when executed by the one or more processors, cause the one or more processors to:

determine that the safety parameter or the comfort parameter is less than a threshold value; and

in response, implement a second test iteration of the at least one simulation scenario for testing that at least one compute path, wherein the second test iteration of the at least one simulation scenario applies a second artificial delay to the at least one edge from the one or more edges, wherein the second artificial delay is greater than the first artificial delay.

15. The non-transitory computer-readable storage medium of claim 13 , comprising further instructions which, when executed by the one or more processors, cause the one or more processors to:

disable one or more alternate compute paths for processing the input sensor data by the autonomous vehicle.

16. The non-transitory computer-readable storage medium of claim 13 , comprising further instructions which, when executed by the one or more processors, cause the one or more processors to:

identify a plurality of compute paths for processing the input sensor data by the autonomous vehicle, wherein the at least one compute path is part of the plurality of compute paths; and

selectively enable one or more compute paths from the plurality of compute paths during the first test iteration of the at least one simulation scenario.

17. The non-transitory computer-readable storage medium of claim 13 , wherein the set of nodes includes at least one of a perception stack, a prediction stack, a planning stack, and a control stack.