Zhao, X., Liang, W., Zhang, X., Chew, C. M., & Wu, Y. (2024). Unknown Object Retrieval in Confined Space through Reinforcement Learning with Tactile Exploration. 2024 IEEE International Conference on Robotics and Automation (ICRA), 22, 10881–10887. https://doi.org/10.1109/icra57147.2024.10611541
Abstract:
The potential of tactile sensing for today's dexterous robotic manipulation has been demonstrated by its ability to facilitate nuanced real-world interactions. In this study, the challenge of object retrieval from confined spaces, unsuitable for conventional visual perception and gripper-based manipulation, is identified and addressed. Significantly, a tactile-sensorized tool stick is utilized and a reinforcement learning (RL) agent with a blend of generic continuous actions and parameterized action primitives is proposed and implemented to learn the optimal policy for such manipulation of everyday objects. We hypothesize that a model trained on representative basic shapes with different physical properties can be generalized to a wide range of daily objects. Moreover, to further accelerate the hardware-based training process, a curriculum on terminal goals is formulated. A range of comparative experiments and ablation studies have been conducted to validate the effectiveness, robustness and efficiency of the proposed approach. The experimental results show that the proposed method achieves a high success rate on the robotic object retrieval task.
License type:
Publisher Copyright
Funding Info:
This research / project is supported by the Agency for Science, Technology and Research - Career Development Fund
Grant Reference no. : C210812049
This research / project is supported by the Agency for Science, Technology and Research - RIE2025 Industry Alignment Fund for Pre-Positioning Program
Grant Reference no. : M21K1a0104