Reinforcement Learning Methods to Enable Automatic Tuning of Legged Robots
نویسندگان
چکیده
Bio-inspired legged robots have demonstrated the capability to walk and run across a wide variety of terrains, such as those found after a natural disaster. However, the survival of victims of natural disasters depends on the speed at which these robots can travel. This paper describes the need for adaptive gait tuning on an eight-legged robot, which will enable it to adjust its gait parameters to increase the speed at which it navigates difficult and varying terrains. Specifically, we characterize the robot’s performance on varied terrains and use the results to inform the implementation of a finite-difference policy gradient reinforcement learning algorithm. We compare the robot’s performance under hand-tuned policies with the performance under the reinforcement learning algorithm, and finally, suggest improvements to the presented policy search process.
منابع مشابه
A 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 کاملReset-free Trial-and-Error Learning for Robot Damage Recovery
The high probability of hardware failures prevents many advanced robots (e.g., legged robots) from being confidently deployed in real-world situations (e.g., post-disaster rescue). Instead of attempting to diagnose the failures, robots could adapt by trial-and-error in order to be able to complete their tasks. In that case, damage recovery can be seen as a Reinforcement Learning (RL) problem. H...
متن کاملRobot Learning
Robot learning consists of a multitude of machine learning approaches, particularly reinforcement learning, inverse reinforcement learning, and regression methods, that have been adapted su ciently to domain so that they allow learning in complex robot systems such as helicopters, apping-wing ight, legged robots, anthropomorphic arms and humanoid robots. While classical arti cial intelligence-b...
متن کاملUsing BELBIC based optimal controller for omni-directional threewheel robots model identified by LOLIMOT
In this paper, an intelligent controller is applied to control omni-directional robots motion. First, the dynamics of the three wheel robots, as a nonlinear plant with considerable uncertainties, is identified using an efficient algorithm of training, named LoLiMoT. Then, an intelligent controller based on brain emotional learning algorithm is applied to the identified model. This emotional l...
متن کامل