Goal-Driven Dynamics Learning via Bayesian Optimization
This work addresses the problem of efficient controller design for complex robots in uncertain environments, representing an incremental improvement by integrating Bayesian optimization with active learning for task-specific dynamics.
The paper tackles the challenge of controlling robots in poorly understood environments by proposing a task-specific approach that learns a dynamics model to maximize control performance, rather than modeling true dynamics, and demonstrates its efficacy through simulations and real quadrotor experiments.
Real-world robots are becoming increasingly complex and commonly act in poorly understood environments where it is extremely challenging to model or learn their true dynamics. Therefore, it might be desirable to take a task-specific approach, wherein the focus is on explicitly learning the dynamics model which achieves the best control performance for the task at hand, rather than learning the true dynamics. In this work, we use Bayesian optimization in an active learning framework where a locally linear dynamics model is learned with the intent of maximizing the control performance, and used in conjunction with optimal control schemes to efficiently design a controller for a given task. This model is updated directly based on the performance observed in experiments on the physical system in an iterative manner until a desired performance is achieved. We demonstrate the efficacy of the proposed approach through simulations and real experiments on a quadrotor testbed.