Deterministic Policy Gradient Based Robotic Path Planning with Continuous Action Spaces

Somdyuti Paul, Lovekesh Vig; Proceedings of the IEEE International Conference on Computer Vision (ICCV), 2017, pp. 725-733

Abstract


Path planners for robotic manipulators often require precise target object locations based on which inverse kinematics return the required joint-angles for approaching the object. This limits their use in real domains with dynamic relative positions of objects not being readily available. We present a deterministic policy based actor-critic learning framework to encode the path planning strategy irrespective of the robot pose and target object position. This reinforcement learning (RL) agent uses two different views of the environment for planning a path to reach a given target from a random pose. On a physics based simulated environment the proposed planner yielded a 100% success rate from 100 different robot poses, with relatively fewer steps required to reach the target. The approach does not require conventional feature matching and triangulation based localization which is often inaccurate, and solves inverse kinematics and depth estimation using only the scene information

Related Material


[pdf]
[bibtex]
@InProceedings{Paul_2017_ICCV,
author = {Paul, Somdyuti and Vig, Lovekesh},
title = {Deterministic Policy Gradient Based Robotic Path Planning with Continuous Action Spaces},
booktitle = {Proceedings of the IEEE International Conference on Computer Vision (ICCV) Workshops},
month = {Oct},
year = {2017}
}