Introduction to Reinforcement Learning
Introduction to Reinforcement Learning
Learning to exploit dynamics for robot motor coordination
Learning to exploit dynamics for robot motor coordination
Online convex optimization in the bandit setting: gradient descent without a gradient
SODA '05 Proceedings of the sixteenth annual ACM-SIAM symposium on Discrete algorithms
Planning Algorithms
Synthesis and control of whole-body behaviors in humanoid systems
Synthesis and control of whole-body behaviors in humanoid systems
RoboCup 2006: Robot Soccer World Cup X
RoboCup 2006: Robot Soccer World Cup X
Efficient gradient estimation for motor control learning
UAI'03 Proceedings of the Nineteenth conference on Uncertainty in Artificial Intelligence
Anubis: Artificial neuromodulation using a bayesian inference system
Neural Computation
Compliant skills acquisition and multi-optima policy search with EM-based reinforcement learning
Robotics and Autonomous Systems
Hi-index | 0.01 |
In this paper, we analyze the insights behind the common approach to the assessment of robot motor behaviors in articulated mobile structures with compromised dynamic balance. We present a new approach to this problem and a methodology that implements it for motor behaviors encapsulated in rest-to-rest motions. As well as common methods, we assume the availability of kinematic information about the solution to the task, but reference is not made to the workspace, allowing the workspace to be free of restrictions. Our control framework, based on local control policies at the joint acceleration level, attracts actuated degrees of freedom (DOFs) to the desired final configuration; meanwhile, the resulting final states of the unactuated DOFs are viewed as an indirect consequence of the profile of the policies. Dynamical systems are used as acceleration policies, providing the actuated system with convenient attractor properties. The control policies, parameterized around imposed simple primitives, are deformed by means of changes in the parameters. This modulation is optimized, by means of a stochastic algorithm, in order to control the unactuated DOFs and thus carry out the desired motor behavior.