Robotic Peg-in-Hole Insertion with Tight Clearances: A Force-based Deep Q-Learning Approach.
Timo MarkertElias HoernerSebastian MatichAndreas TheisslerMartin AtzmuellerPublished in: ICMLA (2023)
Keyphrases
- robotic manipulator
- contact force
- reinforcement learning
- tactile sensing
- function approximation
- position control
- cooperative
- lower bound
- multi agent
- robotic systems
- optimal policy
- model free
- upper bound
- mobile robot
- worst case
- state space
- stochastic approximation
- learning algorithm
- real robot
- reinforcement learning algorithms
- object manipulation
- control scheme
- action selection
- real time
- robotic arm
- dynamic programming
- learning rate
- deep learning
- end effector
- force control
- manipulation tasks
- neural network
- search algorithm
- vision system
- potential field
- monte carlo
- markov decision processes
- master slave
- robot arm