SYSTEMS AND METHODS FOR INCLUDING SIMULATION NOISE IN PATH PLANNING

Systems and methods for training a neural network for generating a reasoning statement are provided. In one embodiment, a method includes receiving raw data regarding a raw simulation for an ego agent. The raw simulation includes at least one dynamic object. The method also includes generating a noisy simulation by applying a first input noise of a plurality of input noises to the raw simulation. The method further includes predicting one or more object trajectories for the at least one dynamic object. The method yet further includes generating an ego trajectory for the ego agent in the noisy simulation based on the one or more object trajectories. The method includes causing the ego agent to execute the ego trajectory.