Summary of the Dynamic Self-repair and Omni-Pi-tent projects in 2020/2021

Its been a while since this blog was last properly updated,the last post was in January of 2020 with many modules on the point of being produced. As the first of these was assembled a number of bugs were discovered in the mechanical design, exacerbated by the changes in part tolerances caused by switching from Robox printers to Stratasys.

The first assembled of the main batch of robots

 

Screwing together and then testing the first of the production modules found it scarcely able to drive, despite no significant change in module weight compared to Robot 2. And with two full robots, this one and the second prototype, a proper test of the full hinge system was now possible, it failed abysmally, stalling under MUCH lower torque than calculated. And so, just at the point when a redesign and reprint of certain major elements was becoming necessary, the government panicked. R.H.Peck, for one, considers that Spring 2020 saw disproportionate and disasterous damage to mental health, civil liberties and the economy, all while failing horrifically to provide the protection to the vulnerable who most desperately needed it. A precedented, that is Sweden like, strategy focused on preserving normality amid rigorous hygiene and immense support for the vulnerable, he feels, would have worked rather better, even if not perfectly, both while waiting for effective vaccines and now that they are thankfully available.

 

Thanks to heroic efforts of our Department's Technicians we resumed work in mid-summer 2020, with R.H.Peck, for one, requiring physical presence to continue the project. Summer, Autumn and Winter have been spent working very hard to make up for those 3 months of time lost to the very worst excesses of authoritarian intrusion. Hence, this update comes over a year since the previous one. As R.H.Peck's PhD project, the Dynamic Self-repair and Omni-Pi-tent work also involves the production of a thesis, which has taken further time and exacerbated the delay of this post.

 

A series of posts will examine the various different aspects of this work, both in hardware and in simulation, but to summarise:

• The wheel mechanisms have been optimised for improved performance
  
• The hinge mechanism has been completely redesigned to use involute rather than worm gears, more powerful 12V motors have been fitted and all the electrical alterations this required will be posted about later on
  
• The hook mechanism has been optimised for better fit
• We have a journal paper in progress on Self-assembly strategies
• Converting self-assembly code from lua in the V-REP sims to C for the hardware has been done, and small-scale demonstrations of multi-robot in-motion docking and self-assembly performed with hardware (more to come)
• The initial implementation of Dynamic Self-repair has been programmed, and is currently being compared against other self-repair methods in simulation
• Work is under way on further improvements to Dynamic Self-repair, ensuring it will soon be able to cope with a wider range of failures
• The compass navigation system has been redesigned to bypass the internal calibration routines of the BNO 055 chip, which were causing problems
• We have found a way to easily run bulk V-REP simulations on a cluster, this too will get its own post
  

 

The first production robot, upgraded with new hinge and wheel mechanisms, with the second prototype in the background

 A video, below, shows the new robot running, note the improved manoeuvrability compared to the prototypes.