US 11,741,274 B1
Perception error model for fast simulation and estimation of perception system reliability and/or for control system tuning
Andrew Scott Crego, Foster City, CA (US); Sai Anurag Modalavalasa, Foster City, CA (US); Subhasis Das, Menlo Park, CA (US); Siavosh Rezvan Behbahani, San Francisco, CA (US); and Aditya Pramod Khadilkar, Foster City, CA (US)
Assigned to Zoox, Inc., Foster City, CA (US)
Filed by Zoox, Inc., Foster City, CA (US)
Filed on Nov. 20, 2020, as Appl. No. 17/100,787.
Int. Cl. G06F 30/20 (2020.01); G06F 30/15 (2020.01); G06N 7/01 (2023.01)
CPC G06F 30/20 (2020.01) [G06F 30/15 (2020.01); G06N 7/01 (2023.01)] 20 Claims
OG exemplary drawing
 
1. A method comprising:
generating a simulation comprising a simulated environment and a simulated object;
controlling a simulated autonomous vehicle to traverse the simulated environment based at least in part on a configuration of a component of the simulated autonomous vehicle;
determining an error distribution by a perception component error model based at least in part on a state of the simulated object during the simulation, wherein the perception component error model is trained based at least in part on an object detection determined by a perception component using sensor data and wherein the error distribution comprises a plurality of contours around the simulated object;
determining, based at least in part on the error distribution, a maximum probability of an intersection of the simulated autonomous vehicle with the simulated object; and
determining, based at least in part on the maximum probability, a performance metric associated with the configuration.