Learning Control from Raw Position Measurements
This addresses a practical issue for robotics and control engineers by simplifying MBRL implementation in velocity-limited systems, though it is incremental as it builds on prior work.
The paper tackles the problem of applying Model-Based Reinforcement Learning (MBRL) to mechanical systems where velocities cannot be directly measured, by proposing VF-MC-PILCO, a velocity-free algorithm that uses Gaussian Process Regression and particle-based policy gradient. Results on simulated and real systems show it achieves similar performance to a previous method but eliminates the need for designing state estimators.
We propose a Model-Based Reinforcement Learning (MBRL) algorithm named VF-MC-PILCO, specifically designed for application to mechanical systems where velocities cannot be directly measured. This circumstance, if not adequately considered, can compromise the success of MBRL approaches. To cope with this problem, we define a velocity-free state formulation which consists of the collection of past positions and inputs. Then, VF-MC-PILCO uses Gaussian Process Regression to model the dynamics of the velocity-free state and optimizes the control policy through a particle-based policy gradient approach. We compare VF-MC-PILCO with our previous MBRL algorithm, MC-PILCO4PMS, which handles the lack of direct velocity measurements by modeling the presence of velocity estimators. Results on both simulated (cart-pole and UR5 robot) and real mechanical systems (Furuta pendulum and a ball-and-plate rig) show that the two algorithms achieve similar results. Conveniently, VF-MC-PILCO does not require the design and implementation of state estimators, which can be a challenging and time-consuming activity to be performed by an expert user.