Continuous control with deep reinforcement learning
This work addresses the challenge of continuous control in robotics and simulation for AI researchers, offering a novel method that is not incremental but builds on prior deep learning advancements.
The paper tackled the problem of applying deep reinforcement learning to continuous control tasks by adapting Deep Q-Learning ideas to continuous action domains, resulting in an algorithm that robustly solves over 20 simulated physics tasks and achieves performance competitive with planning algorithms with full domain knowledge.
We adapt the ideas underlying the success of Deep Q-Learning to the continuous action domain. We present an actor-critic, model-free algorithm based on the deterministic policy gradient that can operate over continuous action spaces. Using the same learning algorithm, network architecture and hyper-parameters, our algorithm robustly solves more than 20 simulated physics tasks, including classic problems such as cartpole swing-up, dexterous manipulation, legged locomotion and car driving. Our algorithm is able to find policies whose performance is competitive with those found by a planning algorithm with full access to the dynamics of the domain and its derivatives. We further demonstrate that for many of the tasks the algorithm can learn policies end-to-end: directly from raw pixel inputs.