Optimal control with adaptive internal dynamics models

Optimal feedback control has been proposed as an attractive movement generation strategy in goal reaching tasks for anthropomorphic manipulator systems. The optimal feedback control law for systems with non-linear dynamics and non-quadratic costs can be found by iterative methods, such as the iterative Linear Quadratic Gaussian (iLQG) algorithm. So far this framework relied on an analytic form of the system dynamics, which may often be unknown, difficult to estimate for more realistic control systems or may be subject to frequent systematic changes. In this paper, we present a novel combination of learning a forward dynamics model within the iLQG framework. Utilising such adaptive internal models can compensate for complex dynamic perturbations of the controlled system in an online fashion. The specific adaptive framework introduced lends itself to a computationally more efficient implementation of the iLQG optimisation without sacrificing control accuracy-allowing the method to scale to large DoF systems.