Proximal Policy Optimization Algorithms
This addresses the need for more efficient and practical reinforcement learning algorithms for researchers and practitioners, though it is incremental relative to prior methods like TRPO.
The paper tackles the problem of improving policy gradient methods in reinforcement learning by proposing Proximal Policy Optimization (PPO), a simpler and more general approach that enables multiple minibatch updates per data sample. The result shows that PPO outperforms other online policy gradient methods on benchmark tasks like robotic locomotion and Atari games, with better sample complexity empirically.
We propose a new family of policy gradient methods for reinforcement learning, which alternate between sampling data through interaction with the environment, and optimizing a "surrogate" objective function using stochastic gradient ascent. Whereas standard policy gradient methods perform one gradient update per data sample, we propose a novel objective function that enables multiple epochs of minibatch updates. The new methods, which we call proximal policy optimization (PPO), have some of the benefits of trust region policy optimization (TRPO), but they are much simpler to implement, more general, and have better sample complexity (empirically). Our experiments test PPO on a collection of benchmark tasks, including simulated robotic locomotion and Atari game playing, and we show that PPO outperforms other online policy gradient methods, and overall strikes a favorable balance between sample complexity, simplicity, and wall-time.