Fabrica AI reposted this
We've got our grouting robot optimized for higher room coverage. Using reinforcement learning, we've jointly researched what's the optimal placement of the extruder for filling the gaps between tiles and what is a good width to length ratio of the grouting robot. The results were quite clear: - Push extruder as much to the site as possible - Aspect ratio does not change much We've applied the changes in the new design of the grouting robot. The extruder position could not be moved all the way to the right, so we moved it as much as possible. This gave us ~7 % larger room coverage when grouting flats in Singapore, almost for free. Next we'll answer questions about drivetrain and sensors: - Is there better coverage if we use mechanum wheels? What would be the impact on average grouting speed? - What is the impact of using less precise tank belts that can climb stairs? - What will be the impact of the quality control system we're thinking of implementing? Should we even develop it? - Is there a better placement for our cameras? What high level robotics questions have you dealt with?
Looks really interesting! Would you be keen to offer projects to Cambridge University students on this?
Ronald Luc got a video of how your robots operate?
Amazing work!
This would be interesting to apply to investigate learning strategies to enable grasping of objects using a passive ‘human-like’ hand mounted on an industrial arm. Here's more info: https://docs.google.com/document/d/1imU7ePApfiRSwDv3-6nvCm1JJvSfcojgWq4JKe2kibg/edit Would it be able to do it?