Dual arm manipulation

Most of the service robots are painstakingly coded and trained in advance to perform tasks correctly. The knowledge of such robots is fixed after the training phase, and any changes in the environment require complicated re-programming by expert users. Such limitations prevent the successful deployment of service robots in human-centered environments.
Lifelong learning and continuous improvement during deployment is the most likely path to handle these limitations. In IRL-Lab, we mainly focus on few-shot, lifelong, transfer, and continual learning techniques for robots.
Our dual-arm robot is now fully functioning and can learn about different objects over time. This way, it can easily adapt to new environments and be used in various daily tasks, such as serving a beer!
Check out how our robot can continually learn about different objects on-site and how it serves a drink for the guests!
IRL-lab: www.ai.rug.nl/irl-lab
RUclips: • Dual arm manipulation