Approximate Inverse Reinforcement Learning from Vision-based Imitation Learning
This work addresses vision-based autonomous driving by providing a more generalizable and robust costmap generation method, though it appears incremental as it builds on existing imitation learning and MPC techniques.
The authors tackled vision-based navigation by developing an approximate inverse reinforcement learning method that generates cost functions from imitation learning, which when combined with model predictive control outperforms other state-of-the-art costmap generators in novel environments with improved robustness.
In this work, we present a method for obtaining an implicit objective function for vision-based navigation. The proposed methodology relies on Imitation Learning, Model Predictive Control (MPC), and an interpretation technique used in Deep Neural Networks. We use Imitation Learning as a means to do Inverse Reinforcement Learning in order to create an approximate cost function generator for a visual navigation challenge. The resulting cost function, the costmap, is used in conjunction with MPC for real-time control and outperforms other state-of-the-art costmap generators in novel environments. The proposed process allows for simple training and robustness to out-of-sample data. We apply our method to the task of vision-based autonomous driving in multiple real and simulated environments and show its generalizability.