The laboratories Vibration and Smart Structures lab are used both for research and educational purposes. Research topics investigated in the labs cover a wide range of vibration related subjects, from calibration and validation of computational models to measured data to energy harvesting, with applications within wind power and automotive industries, for example.
Vibration laboratory
The vibrations lab holds several data acquisition systems, some readily mobile and others rack mounted, used to measure accelerations, forces and strains over time. Data from such measurements are used e.g. for identification of sources of vibration and fatigue and to calibrate computational models. Applications served are the railroad, automotive, metal cutting and wind power industries, and experiments are equally likely to be performed in the lab as at the machine, sleeper pad or train set exhibiting problems.
Contact person
Smart structures lab
In many ways, the equipment in the smart structures lab resembles that of the vibration lab, but the focus here is instead to minimize vibrations through active/semi-active damping. The smart structures lab also holds a scale model of the drive line of a wind turbine plant, set up to study the effects of wear on bearings caused by such things as axial misalignment.
Robot Athletic Intelligence Lab (RAIL)
The Robot Athletic Intelligence Lab (RAIL) at Chalmers University of Technology develops next-generation physical and athletic intelligence for robotic systems focusing on legged robots such as quadrupeds and humanoids. Inspired by the agility, efficiency, and resilience found in nature, we combine multibody dynamics, optimal control, and machine learning to create robots that can move dynamically and safely in complex, real-world environments. Our aim is to deliver world-class education and research in robotics and physical AI.
RAIL is built around the concept of “athletic intelligence” - the capacity of robotic systems to move with the agility, efficiency, and resilience found in biological locomotion. Our focus is on underactuated robots: machines that, like animals and humans, have fewer controllable actuators than degrees of freedom and must therefore rely on intelligent exploitation of their natural dynamics. This class of systems is at the frontier of robotics research, and mastering it requires deep integration of applied mechanics, control theory, machine learning and system design. Our research program spans four interconnected themes.
The first is the co-design of legged robots: the simultaneous optimization of mechanical hardware, motion trajectories, and closed-loop controllers, supported by formal stability guarantees. This holistic approach, funded by the SSF Future Research Leader grant (15 MSEK, 2025–2030), targets applications in hospitality, healthcare, search-and-rescue, and planetary exploration, and employs generative AI and natural language interaction to make robot design accessible to non-specialists.
The second theme addresses constrained optimal control for bi-manual loco-manipulation, developing AI-driven frameworks for humanoid robots performing dexterous assembly tasks. The project (WASP PhD Project Grant, 4 MSEK, 2024–2029) targets platforms including mobile manipulators and humanoids and combines kinodynamic motion planning and control with learning from demonstration to handle the hundreds of geometric constraints arising in closed-loop robot designs.
Third, RAIL is a leading contributor to open benchmarking in underactuated robotics. Through the CloudPendulum initiative, we provide the international community with standardized environments for evaluating optimal control and reinforcement learning algorithms directly on physical hardware. The AI Olympics with RealAIGym competition series has attracted participants from across the world.
Fourth, we pursue cross-disciplinary applications of our core methods in manufacturing, transportation research, healthcare, space robotics, and quantum computing demonstrating the generality of the mathematical and algorithmic frameworks developed in the lab.
Our infrastructure
Cloud Pendulum
A cloud-connected testbed of physical underactuated systems (such as double integrator, simple pendulum, double pendulum, 5 bar mechanism, coupled oscillator etc.) hosted at Chalmers, 24x7 accessible remotely by students and researchers worldwide. The platform integrates multibody dynamics simulations, live hardware video streams, supporting courses in programming, AI, dynamics, robotics, and control. It is funded with grants from the Chalmers Foundation Digitalization Initiative and the Chalmers Innovation Office.
Robots at RAIL
The robot hardware infrastructure of RAIL consists of a Unitree Go2 quadruped and Unitree G1 humanoid. These systems serve both as research testbeds for validating optimal control and reinforcement learning algorithms and as teaching platforms for the course TRA455 Athletic Intelligence in Robotics.
