The UK Micromouse championships were held at the Technology innovation Centre in Birmingham on Saturday June 27th. A good number of mice ran this year, some for the first time. The morning practice was a nerve-wracking affair which looked like it would result in another win by MouseX. However, a couple of mechanical repairs, some software tweaks and an unusually fortunate maze configuration saw a new champion emerge…
Having lost a lot of sleep over the last couple of weeks, I was convinced I had done nearly all I could to make Decimus ready for competition. There remained several serious weaknesses: I had been unable to adapt the motion control to allow me to search without stopping for simple turns and that was costing a lot in search time penalties – it is time all that got re-written; steering at speed was still a bit unstable in spite of improvements and that was limiting my straight-line performance. On the other hand, the search algorithm was greatly improved. Decimus now has a more Adachi-like algorithm that only searches for unexplored cells on the calculated best route and so it no longer needs to re-visit the centre or the start until it has completed its search and found an optimal solution. In spite of straight-line performance issues, Decimus can corner faster than MouseX. All other things being equal, Decimus probably has the advantage in a maze with many short sections and 90 degree turns whereas MouseX, which is usually configures to reduce the number of turns in its search path, prefers long straights if it can find them. The heats maze proved challenging. It appeared to favour Decimus in its configuration. However, although finding the centre on the first pass was fairly simple, it took Decimus nearly twice as long as MouseX – 75 seconds compared to 45 seconds. Whatever else happened, that would add at least 1 second to the overall score time, putting Decimus at a serious disadvantage. Both mice suffered from the following search. This was a complicated maze with many short wall sections that meant that, as more of the maze was explored, the optimal path changed frequently and it took a long time for the mice to be sure they had a good solution. Eventually a solution was found and Decimus got to do its stuff. The result was less than impressive. Suddenly my mouse seems unable to do the simplest things and would not run at speed with anything like the stability I had hoped for. A reasonable time was managed but Decimus sat in a relatively poor third place behind MouseX and MouseX-squared. I resigned myself to another runner-up prize and wandered off to see if I could change anything. It soon became apparent that the wheels were coming off Decimus – literally. one of the wheels was quite loose and wobbled alarmingly. The gears had even started to chew on the edge of the tyre. A little work with the tools and a drop of superglue as threadlock sorted that out. Small changes were made to the steering constants and the sensors checked for calibration. I tried increasing up the search speed and managed to improve that by about 20% but the in-place turns remained. It was now going fast enough that there was a risk of over-running the decision points in the search. A fresh approach would be needed to make it any faster. The main running parameters were tweaked up a bit so that the first speed run was faster and cornering speed increased as much as I dared. I didn’t want it to crash out on the first fast run but, under UK rules, pretty well nobody improves their score after their first speed run. since I thought there was no chance of coming first, I figured I had little to lose and might as well push it to the edge early on. I was still considering a couple of other changes that might have been possible, if risky, when the announcement came that the finals were starting early. Ready or not, it was all too late now. The finals are run in reverse order of the heats position and I had already missed many of the other mice. The maze had relatively minor changes from the heats and would still take some time to explore. However, there was a clear, short route that would suit Decimus very well so long as it didn’t crash. The run began. After a couple of heart-stopping moments and a lengthy search for alternatives, Decimus finally settled on its best route and began its speed runs. Its shortest time was a little over five seconds and I was pleased to see that the instability was gone and Decimus behaved itself perfectly. Now, what would the Derek Hall- and Jim Chidley-driven MouseX team manage? MouseX2 was next to run and it performed a comprehensive search as expected. There were few surprises and the search was impressively fast. Soon enough mouseX2 indicated that it was satisfied and made its way back to the start for a comprehensive thrashing of Decimus. Except that it didn’t! Instead of taking the nice, short route that Decimus took, it went right round the outside on a path that was probably twice as long. I can only assume that it tried to minimise the number of turns and overlooked the short route. It performed beautifully and managed the route in a very impressive time but it was a bad choice of route and the overall score was worse than Decimus. Suddenly I had a chance. I was going to be second place at least. Now, literally on the edge of my seat, I watched Derek hall step up with MouseX. There was the usual fiddling while the entry card was scanned and the mouse placed carefully in the start square. Tension in the hall was high and the audience all but silent. I watched intently, feeling, with Derek, the nervous anticipation that every mouse builder experiences when they start their mouse. Derek pressed the button. The usual short pause before the mouse was ready stretched out. After a moment, Derek tried again. Nothing. A quick examination and another attempt. Still nothing. Derek announced his mouse to be dead. Some error had caused the last program to be loaded incorrectly and the mouse would not run. It was all over. Decimus was the new UK champion. Part of me was overjoyed while another part was disappointed. This was no way to win. Derek expressed the opinion that his mouse would have chosen the long route just like MouseX2 and so was very unlikely to beat the time set by Decimus but we couldn’t know that for sure. MouseX had to retire but would be given the opportunity to run again later. I had been in Derek’s position before and it is extremely disappointing to put in all that effort and have something simple prevent your mouse from even running. Derek did return with his mouse reprogrammed and ran the finals maze. As he predicted, MouseX took the long route and, while very fast, didn’t manage to better the time set by Decimus. Its final recorded time gave it second place. True enough, if either of the MouseX beasts had chosen the short route, I would probably have not have won. But they didn’t and for the first time in several years, the maze configuration suited my program choices perfectly. UK champion at last! Here is a picture of the maze with the route taken by Decimus as well as a link to the maze file in the usual format:
This Post Has 4 Comments
I obviously missed a very good competition.
Hope to be there next year.
I look forward to it – we missed you.
Don’t forget the November competition – I havn’t fixed the date yet but it has been the last Saturday in November for the past 4 years, so I don’t see that changing – all depends on what space I can get.
Look forward to seeing you there
Congratulations! See you in September in Taiwan!