Advancements in Robot Learning and Manipulation

The field of robotics is rapidly advancing, with a focus on improving robot learning and manipulation capabilities. Recent developments have led to the creation of more efficient and effective algorithms for tasks such as inverse kinematics, trajectory planning, and grasping. One notable trend is the use of simulation data to supplement real-world data, allowing for more rapid and cost-effective training of robot models. Additionally, there is a growing interest in using human feedback and preference-driven approaches to guide robot learning and improve task performance. Researchers are also exploring new methods for adapting world models to real-world dynamics, enabling more accurate and reliable robot control. Overall, these advancements have the potential to significantly improve the capabilities and autonomy of robots in a variety of applications, from manufacturing and logistics to healthcare and nuclear decommissioning. Notable papers include: IKSel, which proposes a novel seed selection strategy for fast numerical inverse kinematics iterations. Next-Best-Trajectory Planning of Robot Manipulators, which develops a strategy for effective observation and exploration using the Next-Best-Trajectory principle. Empirical Analysis of Sim-and-Real Cotraining, which investigates the use of simulated data to improve performance in real-world tasks. Can Visuo-motor Policies Benefit from Random Exploration Data, which examines the use of random exploration data for training visuo-motor policies. ZeroMimic, which presents a system for distilling robotic manipulation skills from web videos. HACTS, which introduces a novel teleoperation system for robot learning. AutoEval, which proposes a system for autonomous evaluation of generalist robot manipulation policies. Sim-and-Real Co-Training, which derives a simple yet effective recipe for utilizing simulation data to solve vision-based robotic manipulation tasks. Enhancing Physical Human-Robot Interaction, which leverages intrinsic robot tactile sensing capabilities to recognize digits drawn by humans. Teaching Robots to Handle Nuclear Waste, which presents a teleoperation-based learning approach for nuclear waste handling tasks. 8-DoFs Cable Driven Parallel Robots, which presents a novel master controller for high-DoF teleoperation. Slot-Level Robotic Placement, which proposes a modular system for teaching new tasks to robots using human demonstration videos. Information Gain Is Not All You Need, which proposes a heuristic that reduces backtracking by preferring candidate states that are close to the robot, but far away from other candidate states. Preference-Driven Active 3D Scene Representation, which introduces a novel framework that integrates expert operator preferences into the active 3D scene representation pipeline. Adapting World Models with Latent-State Dynamics Residuals, which proposes a latent-state autoregressive world model pretrained in simulation and calibrated to target environments through residual corrections of latent-state dynamics. Estimating Scene Flow in Robot Surroundings, which presents an approach for scene flow estimation from low-density and noisy point clouds acquired from miniaturized Time of Flight sensors.

Sources

IKSel: Selecting Good Seed Joint Values for Fast Numerical Inverse Kinematics Iterations

Next-Best-Trajectory Planning of Robot Manipulators for Effective Observation and Exploration

Empirical Analysis of Sim-and-Real Cotraining Of Diffusion Policies For Planar Pushing from Pixels

Can Visuo-motor Policies Benefit from Random Exploration Data? A Case Study on Stacking

Disambiguate Gripper State in Grasp-Based Tasks: Pseudo-Tactile as Feedback Enables Pure Simulation Learning

ZeroMimic: Distilling Robotic Manipulation Skills from Web Videos

HACTS: a Human-As-Copilot Teleoperation System for Robot Learning

AutoEval: Autonomous Evaluation of Generalist Robot Manipulation Policies in the Real World

Sim-and-Real Co-Training: A Simple Recipe for Vision-Based Robotic Manipulation

Enhancing Physical Human-Robot Interaction: Recognizing Digits via Intrinsic Robot Tactile Sensing

Teaching Robots to Handle Nuclear Waste: A Teleoperation-Based Learning Approach<

8-DoFs Cable Driven Parallel Robots for Bimanual Teleportation

Slot-Level Robotic Placement via Visual Imitation from Single Human Video

Information Gain Is Not All You Need

Preference-Driven Active 3D Scene Representation for Robotic Inspection in Nuclear Decommissioning

Adapting World Models with Latent-State Dynamics Residuals

Estimating Scene Flow in Robot Surroundings with Distributed Miniaturized Time-of-Flight Sensors

Built with on top of