Offline Reinforcement Learning for Visual Navigation
This addresses the logistical challenge of online learning for real-world robots by enabling scalable, generalizable navigation from existing data.
The authors tackled the problem of enabling robots to navigate to distant goals using user-specified reward functions without online trial-and-error, by developing ReViND, the first offline RL system for robotic navigation that leverages existing datasets. They demonstrated that it can navigate off-road to distant goals using only offline training and exhibit reward-dependent behaviors.
Reinforcement learning can enable robots to navigate to distant goals while optimizing user-specified reward functions, including preferences for following lanes, staying on paved paths, or avoiding freshly mowed grass. However, online learning from trial-and-error for real-world robots is logistically challenging, and methods that instead can utilize existing datasets of robotic navigation data could be significantly more scalable and enable broader generalization. In this paper, we present ReViND, the first offline RL system for robotic navigation that can leverage previously collected data to optimize user-specified reward functions in the real-world. We evaluate our system for off-road navigation without any additional data collection or fine-tuning, and show that it can navigate to distant goals using only offline training from this dataset, and exhibit behaviors that qualitatively differ based on the user-specified reward function.