Neural Control Student Projects
Join our F1TENTH Inivincible race car team
Last Update: 03.12.2024
Status: OPEN
Type: bachelors/masters (short/long)
Contact: Florian Bolli bollif@ini.uzh.ch and Prof. Tobi Delbruck (tobi@ini.uzh.ch)
___________________________________________
We are using our Inivincible F1TENTH race car to test ideas about adaptive nonlinear neural control. PhD student Florian Bolli leads the team and carries out experiments with Marcin Paluch and several bachelor and masters students. The Inivincible teams cooperates with the ETH PBL Forza team and does experiments every Monday in the Dubendorf hanger track area.
Last Update: 03.12.2024
Status: OPEN
Type: bachelors/masters (short/long)
Contact: Prof. Tobi Delbruck (tobi@ini.uzh.ch)
___________________________________________
RC sailing is an enthusiast activity where people enjoy sailing remote controlled sailboats around ponds, for example Irchel pond. The aim of Argois to use modern machine learning technology to autonomously sail such a boat, for example, with the objective of a normal sailboat race or to pass a chosen waypoint.
The DragonForce 65 Argo sailboat has a solid state wind speed and direction sensor, GPS, IMU, and embedded RPi linux computer to control the two servos that determine rudder and sail position. We have sailed Argo and collected data on the Irchel pond and during the CapoCaccia neuromorphic workshops (see Argo photo album).
Possible projects:
Collect end to end training data by manually sailing the boat under various conditions.
Develop an interface to a sailboat simulator to collect training data, to explore transfer learning and reinforcement learning approaches in a simpler model-based environment. The model would then need to be validated on Argo to demonstrate it works.
Train and test a sailing controller on Argo.
Contact Prof. Tobi Delbruck (tobi@ini.uzh.ch , 044 635 3038)
Learning to control labyrinth robot
Last Update: 03.12.2024
Status: OPEN
Type: bachelors/masters (short/long)
Contact: Xiang Deng (dxiang@ini.ethz.ch) and Prof. Tobi Delbruck (tobi@ini.uzh.ch)
___________________________________________
The labyrinth robot's task is to navigate a ball through a complex maze using only two servo motors that control the tilt and pan of the maze plane. The system comprises an event camera positioned for a top-down view as the perception input, the labyrinth maze (as shown in the accompanying image), and two servo motors. Achieving this requires rapid and precise control responses guided by real-time perception data.
Student projects have accomplished three main results from Tobi's initial prototype:
Robust frame-based CNN ball tracker - by 2 gymnasium students
Offline Trajectory Optimization with Online PID Feedback Control – by Zhuoyi Liang (master semester project).
Real-time Model Predictive Control (MPC) along with Simulation Model Development – by Gawain Marti (bachelor thesis).
Looking ahead, we aim to integrate our neural controllers with the existing MPC-based developments, which have the potential to improve control and contribute to a publication.