Safe Policy Search for Lifelong Reinforcement Learning with Sublinear Regret
This addresses the issue of non-vanishing regret and unsafe policies in lifelong learning for versatile agents, though it appears incremental as it builds on existing lifelong and policy gradient methods.
The paper tackled the problem of lifelong reinforcement learning by developing a safe policy gradient learner that operates in an adversarial setting to learn multiple tasks online with safety constraints, achieving sublinear regret for the first time and validating it on benchmark systems and quadrotor control.
Lifelong reinforcement learning provides a promising framework for developing versatile agents that can accumulate knowledge over a lifetime of experience and rapidly learn new tasks by building upon prior knowledge. However, current lifelong learning methods exhibit non-vanishing regret as the amount of experience increases and include limitations that can lead to suboptimal or unsafe control policies. To address these issues, we develop a lifelong policy gradient learner that operates in an adversarial set- ting to learn multiple tasks online while enforcing safety constraints on the learned policies. We demonstrate, for the first time, sublinear regret for lifelong policy search, and validate our algorithm on several benchmark dynamical systems and an application to quadrotor control.