CORL: A Continuous-state Offset-dynamics Reinforcement Learner
This work addresses a foundational challenge in robotics and AI for domains with complex, real-world dynamics, though it is incremental as it builds on existing fitted value iteration methods.
The authors tackled the problem of reinforcement learning in continuous state spaces with stochastic, switching dynamics, such as robot navigation over varying terrain, and demonstrated that their algorithm can efficiently solve such problems with polynomial sample complexity in state-space dimension, as shown in an experiment with a robotic car.
Continuous state spaces and stochastic, switching dynamics characterize a number of rich, realworld domains, such as robot navigation across varying terrain. We describe a reinforcementlearning algorithm for learning in these domains and prove for certain environments the algorithm is probably approximately correct with a sample complexity that scales polynomially with the state-space dimension. Unfortunately, no optimal planning techniques exist in general for such problems; instead we use fitted value iteration to solve the learned MDP, and include the error due to approximate planning in our bounds. Finally, we report an experiment using a robotic car driving over varying terrain to demonstrate that these dynamics representations adequately capture real-world dynamics and that our algorithm can be used to efficiently solve such problems.