Meta parameter optimization of robot force controller
Bachelor/Master Thesis at group TAMS
Motivation
Force control has been validated as a powerful tool for deal with contact-rich tasks in the context of robot learning. There have developed different kinds of force control strategies in the very recent years. However, these force controllers require the mate parameters to be manually set, which is usually non-optimal. Optimization techniques could be used to refine these meta parameters, such as Covariance Matrix Adaptation Evolutionary Strategy (CMAES), Particle Swarm Optimization (PSO), and Reinforcement Learning (RL), thus to obtain optimal performances.
Requirements
- Interesting in this topic
- Basic knowledge in robotics
- Basic knowledge in control
We have research assistant positions available. This topic could also be suitable for successful applicants. Please feel free to contact
Chao Zeng if you are interested.
