■書誌情報
Yanzhou Jin, Adam J. Spiers, Yusuke Iwasawa, Yutaka Matsuo, Yaonan Zhu: Reinforcement Learning with a Low-Cost Parallel Linkage Gripper for Static and Dynamic Planar In-Hand Manipulation, Advanced Intelligent Systems, 2026 (corresponding author)
■概要
Parallel linkage grippers provide a low-cost and compact alternative to anthropomorphic hands for planar manipulation tasks. However, their closed-chain kinematics introduce actuator coupling, singularity-related constraints, and position-dependent manipulation capabilities, making closed-loop control challenging. We present D-PALI-E (Dual PArallel LInkage Enhanced), an open-source, predominantly 3D-printed parallel linkage gripper developed through hardware–learning co-design. The gripper incorporates passive-joint limiters inspired by the overextension-limiting function of the human finger volar plate. During contact, these limiters act as motion constraints and load-bearing elements, excluding mechanically undesirable configurations, contributing to load transmission during pinch grasping, and reshaping the operating region for policy learning. Reinforcement learning is used to achieve quasi-static pose adjustment, including reorientation, repositioning, and full pose rearrangement, as well as dynamic pinch grasping of rolling balls with different sizes and approach directions. Sim-to-real transfer is enabled through system identification and domain randomization. Experiments demonstrate pose adjustment of regular geometric objects with errors within 5 mm and 5°, together with an average success rate of 72.5% for pinch grasping of rolling balls. These results show that combining mechanical redesign with reinforcement learning enables both quasi-static planar pose adjustment and dynamic pinch grasping on low-cost hardware.