Reinforcement Learning of Robotic Legged Locomotion
نویسندگان
چکیده
Humans and animals show a remarkable level of proficiency in their ways of locomotion. They exploit the dynamics of the whole body to perform a variety of motions such as jumping and running. Hereby, the elasticity in the muscles and tendons carries a key role in enabling robust, dynamic and energy efficient locomotion [1]. At the Autonomous Systems Lab, we have developed the robotic leg ScarlETH [2] as a test bed to study the mechanism that can be observed in nature. The bio-inspired articulated leg is electrically driven by highly compliant Series Elastic Actuators (SEA) [3] in the hip and knee joints. The properties of the springs in the joints can be compared to the elasticity of the tissue in nature. Our goal is to maximize ScarlETH’s maneuver performance and locomotion efficiency by developing controllers that excite the robot in step with the dynamics of the system.
منابع مشابه
Optimizing Robotic Single Legged Locomotion with Reinforcement Learning
Humans and animals show a remarkable level of proficiency in their ways of locomotion. They exploit the dynamics of the whole body to perform a variety of motions such as jumping and running. The nature of legged robots raises big challenges for controlling these systems. High degrees of freedom (DOF) and highly nonlinear non-smooth dynamics (due to interaction with the environment) count among...
متن کاملAdaptive posture control of a four-legged walking machine using some principles of mammalian locomotion
This paper presents an adaptive control scheme for the four legged walking machine BISAM. The task of the adaptive control is to learn sensor based reflexes for posture control. For this purpose, an incremental learning scheme is developed based on reinforcement learning. For the planned trajectory of the CoM the data taken from a goat are chosen as a basis, to investigate the transfer potentia...
متن کاملA cerebellar approach to adaptive locomotion for legged robots
This paper describes a neural learning architecture for control of legged robots inspired by mammalian neurophysiology. Biological studies indicate that the cerebel-lum is a key part of an adaptive control system which enables mammals to display remarkable limb coordination during loco-motion. We present a distributed control system using reinforcement learning methods and mechanisms inspired b...
متن کاملImproved Dynamic Stability Using Reinforcement Learning
Many researchers studying legged locomotion have applied tools from reinforcement learning/optimal control to minimize characteristics of a walking gait, most notably the energy consumption. In this paper, we use similar tools to maximize the region of stability of the controller defined as the set of initial conditions from which the robot maintains its balance for at least 5 seconds. Experime...
متن کاملC-trace: a New Algorithm for Reinforcement Learning of Robotic Control
There has been much recent interest in the potential of using reinforcement learning techniques for control in autonomous robotic agents. How to implement eeective reinforcement learning in a real-world robotic environment still involves many open questions. Are standard reinforcement learning algorithms like Watkins' Q-learning appropriate , or are other approaches more suit-able? Some speciic...
متن کامل