Learning local linear Jacobians for flexible and adaptive robot arm control

Abstract

Successful planning and control of robots strongly depends on the quality of kinematic models, which define mappings between configuration space (e.g. joint angles) and task space (e.g. Cartesian coordinates of the end effector). Often these models are predefined, in which case, for example, unforeseen bodily changes may result in unpredictable behavior. We are interested in a learning approach that can adapt to such changes-be they due to motor or sensory failures, or also due to the flexible extension of the robot body by, for example, the usage of tools. We focus on learning locally linear forward velocity kinematics models by means of the neuro-evolution approach XCSF. The algorithm learns self-supervised, executing movements autonomously by means of goal-babbling. It preserves actuator redundancies, which can be exploited during movement execution to fulfill current task constraints. For detailed evaluation purposes, we study the performance of XCSF when learning to control an anthropomorphic seven degrees of freedom arm in simulation. We show that XCSF can learn large forward velocity kinematic mappings autonomously and rather independently of the task space representation provided. The resulting mapping is highly suitable to resolve redundancies on the fly during inverse, goal-directed control.

Goal babbling with unknown ranges: A direction-sampling approach

“…It is inspired by early goal-directed movement attempts that have already been observed in newborns [7]. Several recent studies have shown the success, speed, and scalability of this approach in robotic coordination tasks [8], [9], [10], [11], [3], which also supports the significance of neonates' early goal-directed movements.…”

“…The complemen- tary question is how they can be achieved, which is necessary to eventually identify the ranges of the achievable workspace. The direction sampling approach is generally compatible with different implementations of goal babbling [9], [10], [11], [3] that have recently been suggested. This paper implements this idea on the basis of the algorithm in [8].…”

Filtering Sensory Information with XCSF: Improving Learning Robustness and Robot Arm Control Performance

“…We focus on the XCSF system (Wilson, 2002;Butz, Lanzi, et al, 2008) and its potential for learning the forward velocity kinematics of a robot arm for inverse control Butz and Stalph, 2011;Stalph and Butz, 2012). The considered XCSF setup is based on the available implementation and the detailed descriptions available in the literature Butz, Lanzi, et al, 2008;Stalph and Butz, 2012).…”

“…The considered XCSF setup is based on the available implementation and the detailed descriptions available in the literature Butz, Lanzi, et al, 2008;Stalph and Butz, 2012).…”

“…The key idea is that the estimation of the state change, which is required in the Kalman prediction step, can be obtained using the velocity kinematics mappings stored in XCSF. Stalph and Butz (2012) have developed a system that successfully learns to control robot arms with up to seven degrees of freedom, using an evolutionary function approximation system called XCSF. XCSF can be described as a derivative (Wilson, 2002) of the XCS learning classifier system (Wilson, 1995).…”