Robot Athletic Intelligence Lab · Chalmers University of Technology

Our Vision

Next-generation versatile automation
with highly dynamic robots.

We develop physical and athletic intelligence for underactuated robots, legged machines, and humanoids — bridging optimal control, multi-body dynamics, and machine learning.

Our Research →

Photo: Lisa Jabar

Principal Investigator

Shivesh Kumar

Assistant Professor (Tenure Track)
Department of Mechanical Engineering
Chalmers University of Technology, Gothenburg

PhD summa cum laude, University of Bremen (2019) · Team Leader: Mechanics & Control Team and Founder of Underactuated Robotics Lab, DFKI Bremen (2019–2023) · SSF Future Research Leader (2025–2030) · New Generation Star in Robotics Research, IROS 2024

Google Scholar ORCID Chalmers Profile Email

Our Motto

Athletic Intelligence

Robots that move with the agility, efficiency, and resilience found in nature — underactuated, dynamic, and certifiably robust. An AI that is physically grounded yet connected to cognitive aspects of intelligence.

Research Themes → Funded Projects → Teaching →

Grants & Recognition

Awards & Funding

🏅 SSF Future Research Leader Grant · 15 MSEK · 2025–2030
🏅 WASP PhD Project Grant · 4 MSEK · 2024–2029
🏆 EuroGNC 2024 Best Paper Award
⭐ New Generation Star in Robotics Research — IEEE/RSJ IROS 2024
🤖 No. of Robots: 7

All Projects → Publications →

Join the Lab

Open Positions

We welcome motivated PhD students, postdoctoral researchers, and master's thesis students passionate about underactuated robotics, optimal control, reinforcement learning, and open reproducible science.

Meet the Team → Get in Touch →

7Robots

22.5MSEK Grants

1Postdoc

4PhD Students

8Master's Students

About

The Robot Athletic Intelligence Lab (RAIL), led by Dr. Shivesh Kumar, at the Department of Mechanical Engineering in Chalmers University of Technology develops next-generation physical and athletic intelligence for robotic systems — underactuated robots, legged machines, and humanoids. Our research spans multibody dynamics, optimal control, reinforcement learning, generative AI, and the co-design of hardware and algorithms. Our aim is to perform cutting edge research and educate the next generation of roboticists and physical AI researchers in Sweden's west coast.

Research

See all projects →

⚙

Series-Parallel Hybrid Robots

HyRoDyn framework, whole-body control in actuation space, trajectory optimisation, and co-design. Elsevier book on this class of systems.

🏃

Underactuated Systems & Benchmarking

Open platforms (pendulum, acrobot, AcroMonk, CloudPendulum) for reproducible benchmarking of optimal control and RL on physical hardware.

🔬

Optimal Control & Co-Design

Simultaneous optimisation of mechanical design, motion planning, and controller. Region-of-attraction maximisation, MPC for legged locomotion.

🌐

Applications

Rehabilitation, space, medical robotics, and quantum optimal control — demonstrating generality of our methods.

Selected Publications

All 87 publications →

RA-M 2025 Reinforcement Learning for Robust Athletic Intelligence: Lessons Learned From the Second AI Olympics With RealAIGym Competition. DOI →
IJCAI 2024 Reinforcement Learning for Athletic Intelligence: Lessons from the 1st AI Olympics with RealAIGym Competition. DOI →
IROS 2025 Cloud Pendulum: Democratizing Hardware Access for Robotics Education and Research.
Elsevier Biologically Inspired Series-Parallel Hybrid Robots: Design, Analysis, and Control. DOI →