Video Friday: Humanoid Learns Tennis Skills Playing Humans

2026-03-21 · Source: IEEE Spectrum · Field: Technology & Digital — Robotics & Autonomous Systems, Artificial Intelligence & Machine Learning · Depth: Advanced, short

Summary

This edition of Video Friday showcases diverse advancements in robotics, featuring several new systems and research initiatives. Highlights include LATENT, a system from KAIST DRCD Lab that learns athletic humanoid tennis skills from imperfect human motion data, and Sharpa, a robotics company demonstrating dual-dexterous apple peeling using a Mixture of Dexterous Experts (MoDE-VLA) AI model for complex manipulation. Cranfield University presents a Strandbeest-inspired robot, while the Robotics and AI Institute utilizes NVIDIA Isaac Lab for training reinforcement learning policies for their Unmanned Multi-purpose Vehicle (UMV). Other innovations include a Finger-Tip Changer technology from Tesollo and RoCogMan LaB, an operational PR2 robot from Fluent Robotics Lab, and a fully biodegradable soft robotic system developed by Nature. The University of Pennsylvania's Nirby project uses satellite data and drones for precision agriculture, and Boston Dynamics details the industrial evolution of its Atlas humanoid robot.

Key takeaway

For robotics engineers developing advanced manipulation or locomotion systems, explore integrating shared-autonomy data collection with multi-modal AI models like MoDE-VLA to overcome challenges in high-degree-of-freedom control. Your focus on modular design, as seen in the Atlas robot's industrial evolution, can significantly enhance a robot's utility and thermal management, making it suitable for demanding industrial applications. Additionally, consider the environmental impact by investigating biodegradable components for future designs.

Key insights

Robotics research is advancing dexterous manipulation, sustainable design, and complex locomotion through AI and novel hardware.

Principles

• Imperfect human data can train complex robot skills.
• Shared autonomy improves data collection for RL.
• Modular design enhances industrial robot utility.

Method

Sharpa's MoDE-VLA fuses vision, language, force, and touch data with specialist "experts" for stable, effective control in high-dimensional manipulation tasks.

In practice

• Use shared autonomy for complex robot data collection.
• Integrate multi-modal data for dexterous control.
• Consider biodegradable materials for robotic systems.

Topics

• Humanoid Robotics
• Dexterous Manipulation
• Reinforcement Learning
• Aerial Robotics
• Sustainable Robotics

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

Related on AIssential

• Video Friday: Atlas Versus a Fridge — Robotics is rapidly advancing, with humanoids moving to industrial tasks and diverse specialized robots emerging.
• 9 Amazing Robotics News You Should Know About — Scaling human motion data and shared intelligence layers are accelerating robot dexterity and real-world deployment.
• Video Friday: Robotic Motion Discovery Reveals Unusual Behaviors — Robotics innovation spans humanoid loco-manipulation, domestic assistance, space operations, and open-source hardware for research.
• Robots that refuse to fail: AI evolves 'legged metamachines' that reassemble and withstand injury — Modular robots with athletic intelligence can reconfigure, recover from injury, and maintain movement in dynamic environments.
• Intelligent Robots in 2026: Are We There Yet? [Nikita Rudin] - 760 — Bridging the sim-to-real gap and integrating semantics are critical for advanced robot autonomy and value generation.
• Can a Prosthetic Hand Deliver Humanoid Dexterity? — Integrating human-proven AI bionic hands into humanoids enhances factory-ready durability and dexterity.

Open in AIssential →

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