Abstract
One of the critical challenges in applied robotics is transferring controllers synthesized in simulation to real-world platforms. This talk presents our effort to bridge the sim-to-real gap by adapting TurtleBot 4 robots to perform a collective foraging task inspired by leaf-cutter ants. As a baseline, we developed a finite state machine framework that integrates multiple perception and control technologies. These include on-board sensors (LiDAR, light sensors, infrared), real-time robot tracking, closed-loop control, and preliminary swarm coordination. Together, these components enable the robots to accomplish the foraging task in the physical world. This work will form the foundation for the future seamless transfer of various controllers, such as Artificial Neural Networks or Behaviour Trees, from simulation to reality.
About the speaker
Raj Ammu is currently working towards his MEng degree in Mechatronics and Robotic Engineering at the University of Sheffield in the UK. He is currently working with Paolo Leopardi on transferring controllers synthesised in simulation into the real world. He is hosted at the University of Konstanz this summer through the DAAD RISE Germany program.