Here we are with ten days to go before the UK micromouse competition and Decimus is still far from ready. All the basic software is there but there are still issues with the basic motion control. I have all the turns calculated and the parameters have been tuned to a good first approximation. This takes hours of tedious repetitive work as the parameters are altered…
For each turn, there is an average speed, a difference speed that sets the basic turn radius, a turn distance that determines the angle, and a lead-in and lead-out value that tells the controller where the turn starts and finishes in relation to a fixed point – usually the cell centre. It is important that the turns do not interact with each other. For example, it would be no good having a 90 degree turn that finished so far in to the next cell that it was not possible to have an other turn immediately after it.
There seems to be no way round the tuning process as the mouse is not perfect. Calculation will get you close but in the end, it has to be finished by experiment. Here is a bit of video of the testing:
The worst thing is that I seem to have some kind of bug that means I just cannot make the turns reliable. Every now and then a turn will be a good few degrees out and a crash is likely. Reliability and repeatability in the maze are vital. Today I set up some complete wall following and solving tests to make sure that configuring the turns had not messed anything else up. I don’t think it had but I discovered that the edge detection for the walls is broken. At the end of the video you can see Decimus travelling along at 20mm/s. At it passes a post that may be zero, one or two edge detection events – as signalled by a brief flash on an LED on the right of the mouse. This has to be fixed. I haven’t even started on the diagonal steering and error correction. Nor do I have the smart (Adachi) solver installed. Well, there are still 10 days to go…