Understanding Cheetah Locomotion using Inverse Reinforcement Learning

Project ID: 2531ac1468

(You will need this ID for your application)

Research Theme: Artificial Intelligence and Robotics

UCL Lead department: Computer Science

Department Website

Lead Supervisor: Amir Patel

Partner Organisation: Mathworks

Stipend enhancement: £1,250 per year

Project Summary:

Understanding Cheetah Locomotion Using Inverse Reinforcement Learning

Why This Research Is Important

Discovering how cheetahs achieve their incredible speed and agility can revolutionize the design of agile robots. This research aims to decode the principles behind cheetahs’ rapid accelerations and sharp turns using inverse reinforcement learning (IRL), leading to advancements in robotics and artificial intelligence.

Who You Will Be Working With

Join Dr. Amir Patel’s Intelligent Robotics Group at University College London (UCL), collaborating with experts in biomechanics, robotics, and machine learning. You’ll also partner with MathWorks, gaining access to cutting-edge tools like MATLAB and Simulink and receiving mentorship from their development teams.

What You Will Be Doing

• Investigate IRL Methods: Evaluate current IRL techniques using simulated mechanical models like inverted pendulums and monopod robots to refine analysis methods for dynamic systems.

• Apply IRL to Cheetah Data: Use existing datasets of cheetah movements to extract reward functions that explain their high-speed maneuvers. Analyze kinematic and kinetic data to determine what objectives cheetahs optimize during rapid actions.

• Implement Findings in Robotics: Test these reward functions on the Unitree Go2 quadruped robot. Develop and train reinforcement learning controllers to enhance the robot’s agility, enabling it to perform movements similar to a cheetah.

Throughout the project, you’ll have opportunities to publish your findings, attend international conferences, and contribute to the future of robotics and AI.

Who We Are Looking For

We are seeking a motivated candidate with:

• A strong background in engineering, computer science, robotics, or related fields.
• Experience or interest in machine learning, especially reinforcement learning.
• Proficiency in programming, preferably with MATLAB or Python.
• A passion for interdisciplinary research that bridges biology and robotics.

If you’re excited about exploring the natural world to drive technological innovation, we encourage you to apply!