Advances in Robot Learning and Manipulation

The field of robot learning and manipulation is rapidly advancing, with a focus on developing more efficient and effective methods for training robots to perform complex tasks. One of the key directions in this area is the development of new reward structures and reinforcement learning algorithms that can handle long-horizon tasks and sparse rewards. Researchers are also exploring the use of visual and spatial information to improve robot manipulation, such as learning to anticipate and plan for future actions. Additionally, there is a growing interest in using temporal logic and stage-aware reward modeling to enable robots to perform complex tasks that involve multiple stages and sub-tasks. Notable papers in this area include ReLAM, which introduces a novel framework for automatically generating dense rewards from action-free video demonstrations, and TGPO, which proposes a hierarchical framework for solving general Signal Temporal Logic tasks. Other noteworthy papers include STAIR, which addresses stage misalignment in preference-based reinforcement learning, and TimeRewarder, which learns dense rewards from passive videos via frame-wise temporal distance.

Sources

ReLAM: Learning Anticipation Model for Rewarding Visual Robotic Manipulation

Bridging Discrete and Continuous RL: Stable Deterministic Policy Gradient with Martingale Characterization

STAIR: Addressing Stage Misalignment through Temporal-Aligned Preference Reinforcement Learning

Mash, Spread, Slice! Learning to Manipulate Object States via Visual Spatial Progress

Towards Tighter Convex Relaxation of Mixed-integer Programs: Leveraging Logic Network Flow for Task and Motion Planning

SARM: Stage-Aware Reward Modeling for Long Horizon Robot Manipulation

TimeRewarder: Learning Dense Reward from Passive Videos via Frame-wise Temporal Distance

TGPO: Temporal Grounded Policy Optimization for Signal Temporal Logic Tasks

RTFF: Random-to-Target Fabric Flattening Policy using Dual-Arm Manipulator

Built with on top of