Video Friday: Digit Learns to Dance—Virtually Overnight

2026-04-03 · Source: IEEE Spectrum · Field: Technology & Digital — Robotics & Autonomous Systems, Artificial Intelligence & Machine Learning · Depth: Intermediate, short

Summary

This "Video Friday" compilation from IEEE Spectrum Robotics showcases recent advancements and applications across various robotics domains, including upcoming conferences like ICRA 2026 in Vienna and RSS 2026 in Sydney. Key highlights include Agility's Digit robot learning new whole-body control capabilities overnight through sim-to-real reinforcement training, and Generalist AI's GEN-1 model achieving 99% success rates on simple physical tasks with only one hour of robot data. Unitree has open-sourced its UnifoLM-WBT-Dataset for humanoid robot whole-body teleoperation, while MRReP introduces a mixed reality interface for drawing robot paths. Other innovations feature Mirrorbot for fostering human connection via dynamic reflections, PAL Robotics' VR teleoperation for TIAGo Pro, and Sanctuary AI's hydraulic hands demonstrating dexterous manipulation. The compilation also covers a Sanyuan Aerospace satellite with a flexible robotic arm for in-orbit refueling, Tokyo Robotics' humanoid robot with RL-trained control policies, ABB robots for 3D printing railway buildings, and Humanoid robots testing in automotive manufacturing logistics.

Key takeaway

For AI Engineers and Research Scientists developing robotic systems, these demonstrations highlight the increasing viability of general-purpose AI models and advanced control techniques. You should investigate integrating reinforcement learning with sim-to-real transfer for faster skill acquisition in your robot platforms. Consider leveraging open-source datasets like Unitree's UnifoLM-WBT-Dataset to accelerate development in whole-body teleoperation and explore mixed reality interfaces to improve human-robot collaboration and path planning in shared environments.

Key insights

Advanced robotics are achieving new performance thresholds in control, learning, and human interaction.

Principles

• Sim-to-real transfer accelerates robot skill acquisition.
• High-quality datasets are crucial for robot learning.
• Mixed reality enhances human-robot path planning.

Method

Agility uses raw motion data from mocap, animation, and teleop for sim-to-real reinforcement training. MRReP enables users to draw Hand-drawn Reference Paths (HRP) on physical floors using hand gestures.

In practice

• Utilize sim-to-real for rapid robot skill development.
• Explore VR teleoperation for precise remote manipulation.
• Integrate cobots for automating precise manufacturing tasks.

Topics

• Robot Learning
• Humanoid Robotics
• Teleoperation Systems
• Mobile Robot Navigation
• Space Robotics

Best for: AI Engineer, Research Scientist, Robotics Engineer, Machine Learning Engineer, AI Scientist

Related on AIssential

• Training a Unitree G1 to Walk w/ Reinforcement Learning — Achieving sim-to-real transfer for robot locomotion requires meticulous sim-to-sim fidelity and addressing hardware-specific observation gaps.
• RoboDream: Compositional World Models for Scalable Robot Data Synthesis — A generalizable embodiment-centric world model synthesizes photorealistic robot demonstrations by decoupling motion from environment synthesis.
• How robots learn: A brief, contemporary history — AI models and large datasets are revolutionizing robot learning, enabling adaptation and complex task execution.
• Video Friday: Digit Learns to Deadlift — Robotics advancements span humanoids, swarms, and soft actuators, pushing boundaries in coordination, autonomy, and real-world deployment.
• Benchmarking Action Spaces in Reinforcement Learning for Vision-based Robotic Manipulation — Joint velocity action space significantly improves sim-to-real performance for vision-based robotic picking and pushing tasks.
• Video Friday: Humanoid Learns Tennis Skills Playing Humans — Robotics research is advancing dexterous manipulation, sustainable design, and complex locomotion through AI and novel hardware.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by IEEE Spectrum.