Target Localization for Robotic Arm Grasping Based on Deep Reinforcement Learning

Iterative Grasp Pose Refinement: A Deep Reinforcement Learning Approach for 2D Vision

A reinforcement learning-based framework for robotic grasp refinement, integrating keypoint-based object representations with a Deep Q-Network (DQN), is proposed, offering a scalable and adaptable solution for contact-rich manipulation tasks.

Point and Pick: Bounding-Box Conditioned Diffusion Policies and Offline RL for Target-Specific Robot Manipulation

Stability-Aware Closed-Loop Recovery for Robotic Grasping in Complex Simulation Environments

Language-guided robotic grasping has made significant progress in semantic understanding, but existing methods often rely on open-loop execution strategies and struggle to handle physical disturbances such as object collisions, target displacement, and transportation slippage. To address this problem, this paper proposes a stability-aware dynamic recovery mechanism, named SADR. Based on a multi-threaded decoupled architecture, SADR decouples semantic planning, target tracking, and execution control, and constructs a two-stage stability criterion through pre-closure displacement checking and post-closure force/current feedback verification. When target instability, missed grasping, or slippage is detected, the system performs local trajectory correction and re-grasping based on real-time tracking results, without restarting global semantic planning. Experiments in PyBullet show that SADR significantly improves the grasping success rate under high-density disturbance scenarios while reducing the average task completion time. This study provides an effective closed-loop recovery solution for improving the reliability of robotic grasping tasks in complex simulation environments.

Robot Trajectron V3: A Probabilistic Shared Control Framework for SE(3) Manipulation

We aim to address the challenge of teleoperating robotic arms for high-degree-of-freedom (high-DoF) manipulation tasks, which is cognitively demanding and error-prone, particularly when relying on low-bandwidth interfaces. We propose Robot Trajectron V3 (RT-V3), a probabilistic shared control framework designed for $SE(3)$ grasping tasks. RT-V3 formulates shared control as Bayesian inference by learning a prior over user intent and combining it with real-time user commands to estimate the posterior intent distribution. The prior models user intent as a distribution over future trajectories conditioned on past robot dynamics and visual scene context. The intent prior is parameterized by a transformer-based conditional generative model that reasons over point clouds and candidate grasp poses, together with a factorized translation-rotation representation that improves learning efficiency in high-dimensional action spaces. During execution, RT-V3 continuously estimates the posterior distribution over future trajectories by combining the learned intent prior with a user-command likelihood derived from the observed control input, enabling continuous intent refinement and shared assistance. Comprehensive experiments demonstrate that RT-V3 achieves high accuracy in trajectory prediction and competitive performance in reactive planning. Furthermore, real-world user studies indicate that RT-V3 significantly outperforms baseline methods in terms of success rate and efficiency, while substantially reducing the user's physical and mental workload.

Robust Brachiation on a Life-Sized Dual-Arm Robot Using Waypoint-Guided Reinforcement Learning

Brachiation is a form of locomotion in which primates move primarily using their arms, enabling traversal in environments without footholds. However, this motion requires highly coordinated whole-body movement and precise timing control for bar grasping and release. As a result, achieving robust behavior on life-sized robotic platforms remains challenging. In this study, we present a reinforcement learning-based method to realize brachiation on a life-sized dual-arm robot. The core of the proposed approach is Waypoint-Guided Reinforcement Learning (WGRL), a learning framework for inducing non-linear and complex motions. For high-difficulty tasks where imitation learning data are unavailable, WGRL guides behavior acquisition by sparsely specifying waypoints for the end-effector trajectory, while whole-body motion is generated through reinforcement learning. In addition, by integrating the waypoint-following guidance with rewards based on task success and mechanical energy, and training in an environment designed for Sim-to-Real transfer, the proposed method achieves both forward progression and motion stability. The acquired behavior is evaluated through Sim-to-Sim experiments under monkey-bar environments with geometric variations and hardware experiments, confirming robust brachiation including failure recovery behavior. This study provides effective learning design guidelines for realizing arm-based locomotion on life-sized robotic hardware and expanding the traversable workspace of robots.

Deep Reinforcement Learning-Based Intelligent Control Algorithm for Dual-Arm Robots

This paper presents a review-oriented comparative analysis of deep reinforcement learning (DRL) for intelligent control of dual-arm robots. Instead of focusing on a single control algorithm, it organizes recent studies into an algorithm-task-metric framework and extracts quantitative evidence from representative applications including cooperative grasping, assembly, transportation, obstacle-aware planning, contact-rich control, and sim-to-real transfer. PPO, MAPPO, MADDPG, and SAC are compared in terms of success rate, convergence behavior, trajectory smoothness, force regulation, safety constraints, and transferability. Key design factors and future trends, including reward design, multimodal perception, safe reinforcement learning, sample efficiency, and real-robot deployment, are summarized to provide practical guidance for dual-arm intelligent cooperative control.