Repo for motion synthesis code and experiment results from analyzing unified manipulation on arm-hand combinations.
Paper was accepted to IEEE ICRA 2023! Corresponding email: v.patel@yale.edu.
@inproceedings{patel2023unified,
title={An Analysis of Unified Manipulation with Robot Arms and Dexterous Hands via Optimization-based Motion Synthesis},
author={Patel, Vatsal V and Rakita, Daniel and Dollar, Aaron M},
booktitle={2023 IEEE International Conference on Robotics and Automation (ICRA)},
year={2023},
organization={IEEE}
}
The aggregated results as shown in the paper for both the KUKA and UR10 arms on all 5 experiment tasks are as follows.
Optimization weight versions tested:
- Arm-Hand-Config α prioritizes object pose accuracy
- Arm-Hand-Config β prioritizes joint and object motion smoothness
Additional weight versions that prioritize other motion attributes were also tested, and those results are below!
The results for the 7-DOF KUKA arm with the diffrernt hands on all 5 experiment tasks are as follows.
The results for the 6-DOF UR10 arm with the diffrernt hands on all 5 experiment tasks are as follows.
The results for each experiment task averaged across the 2 robot arms (KUKA and UR10) are shown here. The alpha weight versions are shown here for brevity.
Results from other optimization weight versions tested are below:
- Arm-Hand-Config γ prioritizes avoiding singularities and joint limits
- Arm-Hand-Config δ prioritizes reducing joint torques
- Arm-Hand-Config ε prioritizes collision avoidance
