Iterative projection and evaluation of scene states using end to end neural networks for autonomous and semi-autonomous systems and applications

Abstract

In various examples, a modular (e.g., control, drive, software, software/hardware) stack may evaluate candidate trajectory(ies) by predicting the subsequent scene or world state that would result from using the candidate trajectory(ies), and quantifying a measure of the quality, safety, and/or favorability of the predicted subsequent state. If the quality, safety, and/or favorability of the estimated subsequent state satisfies a designated threshold (e.g., the candidate trajectory is predicted to result in a favorable, safe, desirable, and/or beneficial future state), the candidate trajectory may be used to control navigation. If not, a determination may be made to disengage autonomous navigation, execute a safety procedure, and/or a parameterized optimization may be used to generate an optimized trajectory using an optimized planner to iteratively refine a trajectory predicted to lead to the most desirable state—or at least a state beyond a threshold desirability.

Dependencies are inferred from the claim text.