Ocado

Contact: David Sharp <david.sharp@ocado.com>

Possible idea to explore:

it was interesting to hear about your new mobile turtlebots being introduced into Amanda Prorok’s course. This opens up some potential for “Sim to real” experiments / reinforcement learning /projects.

It would be great if we could encourage the Cambridge Engineering and Computing robotics research teams to connect together to create an activity in reinforcement learning for legged robotics to endeavour to create a rival to “Anymal” http://www.rsl.ethz.ch/robots-media/anymal.html from ETHZ and the MIT Cheetah https://techcrunch.com/2019/03/04/mits-speedy-mini-cheetah-robot-learns-to-backflip/

Feedback: We will think about how your sim-to-real concept might be adapted to the course, and what kind of robot hardware could be used. Past experience is that CS students have only limited hardware construction skill - we usually have to arrange for them to have assistance from a technician. There also seems to be a slight tendency for them to be scared off from robot projects, rather than stimulated. Last year we had to cancel the following two projects for lack of interest, despite the fact that they both seemed fairly entertaining to me: Visual Pick and Place Robot Death Watch