Lunghua University of Science and Technology hosted the annual Taiwan Micromouse and Intelligent Robot Contest on September 15th 2013. One again, I was fortunate enough to be able to attend and compete with some of the best micromouse builders around.
This was my fourth trip to Taiwan and each visit is a huge pleasure from start to end. Once again, we were staying at the Chateau de Chine Hotel in Taipei. I mention it not just because it is a fine hotel but because the name stands out. Apparently, one of the contestants had an immigration official question whether the hotel really existed. Well, it certainly does. I arrived on the Thursday evening so that I might get a head start on missed sleep. The flight over was packed and, although my neighbour was pretty quiet and didn’t need to be getting up much, the person the other side of him only did three things during the entire 12 hour flight. Eat, sleep and talk. Not necessarily in that order. In spite of the best efforts of the airline, KLM, to get us bedded down quickly, I got no real sleep on the flight. I did have a bit of a chat with a couple of the cabin crew who were very friendly and helpful. One of them in particular did a great job of appearing interested in what I was going over for. It turns out that this entire crew would be returning on the same plane as me so I was given instructions to take pictures and report back on the return flight.
In keeping with the non-Chinese name of the hotel, the food had a distinct western bias in the ground floor restaurant. Even so, I had to get detailed instructions form the maitre d’ before I could work out how to select and receive food. In essence, there were two main options available – Mongolian barbecue and Pasta. In each case, the idea was that you selected a plate full of tempting goodies, tagged your plate with a little clip that had your table number and then handed it to a chef. A little while later, the plate was returned to your table by a waiter with everything cooked. There was also a range of salads, fruit and, as is common in Taiwan, plenty of seafood options. Since this was an open buffet, I felt obliged to try as many options as possible for my NTD550 (about £13). Very soon, I was barely able to move and felt I should go for a walk around the area while I still could.
For a UK resident, Taiwan at night is a surprising place. First of all, it was still well over 25 degrees C and quite humid – even at 8PM. In fact I think it did not dip below 25 degrees all night. Second, it was Thursday night and everyone was still working so the incredible variety of small shops, factories and other businesses were in full swing. Then there are the scooters. They must stop sometime because there are almost none about at 5:30 in the morning but there are still a good many all evening. Pretty well all the streets have pavements (sidewalks if you prefer). However, they are pretty well useless because, if they are not being used by the building owner to conduct their business, they are packed with parked scooters – all lined up so close they nearly touch. As a result, you are forced to walk in the street with more scooters weaving around you. It takes a little while to get used to this but the only option seems to be to put on a brave face and get on with it. Street crossings are interesting. These are controlled by lights and clearly marked. Everyone waits patiently for the lights to change and then they set out. However, traffic turning into the street you cross seems to be allowed to do so even while your cross. I found two ways to deal with this. First, don’t look. If you do, you will see a scooter apparently on a collision course. You may then stop and upset the aim of the rider who had no desire to run into you so long as you kept going. The other tactic is to stay close to a local as they cross the road so that you can make best use of their experience. On the whole though, the streets of Taiwan in the evening are never less than fascinating.
On the Friday morning I met up with Dave Otten who had just arrived at about 6:30. We had breakfast, stowed Dave’s gear and then our host, Su Juing-Huei had arranged for a visit to a food market near his home. This was a very large collection of small stallholders selling almost every imaginable variety of food. Strangely absent though were much in the way of herbs and spices. I saw only one spice seller and only a few identifiable herbs. Notable among these was Thai Basil which, if you don’t know it, has a wonderful aniseed aroma. For some reason this is very hard to find in my part of the world which is a shame as I consider it a key ingredient in Thai Green Curry. Seafood sellers were particularly interesting. Not only did they have a huge variety of fish I had never seen before, they also kept some in water tanks so that you could be sure that the fish you bought was as fresh as could be.
The scooters were again in evidence here. The market was a sprawling and busy area taking up several streets. The scooters would wind their way among the pedestrians, stopping right up by a stall to select their goods. Purchase made, they would then make their way to their next, favourite stall. Prices did vary between stalls and I have no doubt there was better and worse among them both for price and quality but it would take a lot of practice to find out which was which.
Lunch was in a restaurant at a nearby shopping mall. The contrast could hardly be greater between the ordered chaos of the street market and the shining, bright and more expensive shops in the mall. Apparently, the food markets are growing rather than shrinking in the face of more competition from the kind of food shopping experience we see in the west. I would certainly rather go out and select exactly what I want from a market than have to choose between the limited commercial brands available in a supermarket. The mall was much like any you would find – full of fancy clothes and generally up-market goods. I was mildly surprised to find a large Mothercare store in there. I still can’t decide if I was surprised to find that all the images in that Mothercare store were of western mothers and babies. Really, I cannot work out the logic.
After lunch, it was off to Lunghua University for a bit of practice on the maze. The venue was abuzz with activity as the small army of staff and student helpers prepared for the weekend’s events. We had the maze almost to ourselves and I was able to get in a some good, solid set-up time. Decimus 4B was behaving well. I found and corrected a few small issues and set about deciding on my best strategy for the contest. The Taiwan rules are much the same as the APEC rules and so there is a search time penalty added to the run. Spending any longer than necessary in searching will cost you a lot on every subsequent speed run. Only five runs are available so you cannot afford to go back to the start until you are ready for a speed run. I had a cunning plan worked out and hoped that I would gain a small strategic advantage from it. In Japan, where the rules are simpler, strategy is less important. Not unimportant though. A key decision is how fast to make your run attempts. Try to go to fast too soon and you will just crash out – potentially damaging the mouse. Start too slow however and you may use all your runs and find that you could still go faster. Under the APEC/UK/Taiwan rules it is almost impossible to improve on the overall score gained in the first of your speed runs. In the end, I was happy enough with the performance of D4B. It looked like there would be little work to do in the Saturday practice session. This was unusual for me as I am normally full of problems right up to the last minute.
Towards the end of the afternoon, we were able to visit the lab where the students did their development work. This is a small but busy place and while we were there, several students were busy calibrating mice on a very neat 4×4 maze segment. We were also shown a very nice student mouse design used for teaching. The mounts and wheels were all custom made and fairly standard 12mm motors with (I think) 10:1 gearboxes were used for the motive power. This looked to be an excellent starter mouse and It would be really good to be able to get hold of a couple for demonstrations. Prototyping of parts was done down the corridor with a very nice 3D printer. the parts we saw that came off this machine were every bit as good as the early parts I have had made by Shapeways. These 3D printers are fast becoming essential tools.
In the evening, we went to a well-known mountain-side restaurant to the North of the city. The Top is very popular and deservedly so. The view out of the city at night is spectacular. The restaurant itself id spread out over a large area with a mixture of covered, enclosed tables and open air seating. It is only possible to reserve the enclosed seating and so the view while eating is more restricted. However, the queue for the open air seats can be a couple of hours on a busy night. The food and the company were excellent. Ng Beng Kiat had arrived and by this time we had been joined by a collection of former students from Lunghua whom we had met on previous visits. It was good to catch up with them and find out what they had been doing. When we arrived back at the hotel, We were greeted by the arrival of Harjit Singh and Pierre Hollis from the USA. There is something slightly surreal about the gradual collection of familiar faces from around the world.
Another early start on Saturday (not that I was sleeping much) and yet another chance meeting with a contestant in the hotel lobby. This happens a lot and this morning, it was Khiew Tzong Yong from Singapore. His room was not ready at that time of day so he joined us all for breakfast and considered his options for the day. The plan was to go on a hike in the mountains a little further North from last night’s hotel. Tzong Yong had not been expected for this but stowed his gear in someone’s room and came along for the ride.
It was a beautiful place. No doubt about it. We had good company, great scenery and famous sandwiches. Really – that is what they were called. They were good to eat as well. To be honest though, I was not expecting the climb. At one point we must have gone up about 1000 feet in less than a mile. Even without a chest infection and without the high temperature and humidity, I might have found it a little challenging since I am some way off fit. Under the circumstances though, I struggled somewhat. In the end, it was well worth the effort with some fantastic views over the city and the countryside around us, all the way out to the sea. I would recommend it to anyone. Anyone, that is, who is in better shape than me.
After a quick visit back at the hotel for a shower and a change of clothing, we made our way back to the University for another chance to practice in the maze. By now, I was a little concerned. I couldn’t think of anything that was giving me real trouble in with my micromouse. Sure, there were plenty of things that could be ‘improved’ but nothing that was standing in the way of actually running in the contest. This was very unusual and unsettling. I ran the mouse through a couple of test mazes, satisfied myself that I had selected the best set of speeds and spent the rest of the time catching up with everyone else. During the afternoon, the Japanese contestants arrived in force and, by tea time (that is English for late afternoon), we were all assembled and more or less ready for the contest on Sunday.
The evening meal was in the Hotel restaurant. There were too many of us this year to all fit around one table so we split up into language-compatible groups which essentially meant all the Japanese contestants on one table and the rest of us on another.
Contest day and – guess what – an early start! Now that were were all assembled it needed minibuses to get us to and from the venue but all was arranged with the customary excellent efficiency by our host. As on previous occasions, we were all to assemble on the stage for the opening ceremony. This time, however, we did not have to stand and say anything. That was a bit of a relief since I had simply ad-libbed last year and was not sure I could do any better this year.
Clearly, I focus on the classic micromouse contest since that is all I enter. However, there is a half-size micromouse event, several line-follower events and a line maze also going on in the same contest. The line maze is exactly what it sounds like with robots following a maze made out of lines taped to the ground. They can use the same sensors as the line followers and yet have similar goal to the micromouse event. The robots have to find their way through the maze to an identifiable goal, return to the start and then run the best route as fast as they can. The best robots are able to cut some corners but it is not entirely clear to me how to tell the difference between legitimate corner ‘smoothing’ and illegal shortcuts.
The line follower event is still a developing contest. By that I mean that each year seems to bring new challenges as course designers and robot builders think of new ways to do things. This year, the Taiwan course designers revealed two ways to make life more interesting for the contestants. Firstly, the course was 60m long. This is the maximum length allowed in the rules and presents some specific challenges. Each entry is only allowed three minutes and three runs with the fastest run being your best score. Generally, this means an exploration run and two fast runs. On a longer course it takes a fast mouse and good organisation to get three runs in. On a course this long it seemed unlikely that most contestants would manage more than two runs. The other peculiarity of this course was the long, large-radius turn. While the rules give a minimum radius, I believe there is no specification for maximum radius.
|A1I-028||中州電機 B 隊||53.99||49.91||X||4|
You can see that only one mouse managed to get three runs in and it is no surprise that he got the best overall score. To be able to do the exploration run in 42 seconds or so implies an average speed of about 1.5m/s. I have no idea if Cartis03 runs at constant speed or not but that is pretty fast on an unknown track. Cartis03 uses a boom carrying the line sensor so I can see how it would be possible to vary the speed during exploration.
The classic micromouse contest in which I was taking part is an international invitational with three of the top local mice included. This is pretty significant competition and I was more than a little concerned about how well I might perform. The field had been reduced a little after it turned out that Dave Otten’s mouse had a sensor problem and would not be able to run. The running order seemed to be determined at random rather than being seeded. either way, I was going to run early and was second up after Zeetah V and Zeetah VI – both by Harjit Singh and Pierre Hollis. A timing system problem confused the score from Zeetah VI but the correct values seem to have been recorded and were displayed later.
Decimus 4B did a grand job of searching the maze. I thought I had done a lot to speed that up but, comparing first run times with other entries, I still have some work to do there. Its first speed run was a bit lucky but achieved exactly what I wanted. It was running at the fastest speed that I thought was sensible to be sure of getting through and it did that. Reviewing the video later, it was clear that I wasted a lot of time just cleaning the wheels before the first speed run. Since all that time gets counted toward the final score, I would have been one place higher up the running order simply by not messing about for so long cleaning tyres. Lesson learned. the next two runs both failed at the next highest speed setting. With just one run left, I chose to go back down to the initial speed run setting to see if I could get a repeatable time and to satisfy myself that it was not just a lucky run. What I should have done is to increase the speed further since I have seen cases where faster can be better for some particular turn combinations. Something else I need to work on. Still, D4B performed a final speed run and I was more than happy with its performance. For now.
The remaining runners were all well experienced and there were few surprises in the main. Japanese mice had a tendency to search for too long. The maze used here was complex with two viable paths, the shorter of these was 94 cells long. That is about 50% more than a typical Japanese contest maze. It turns out that, to get longer paths, the number of options tends to be reduced and, once the mouse has been out and back once, it probably already has the best route. Not all mice took the shortest route and I didn’t find out why that was. Kojimouse 7CL was entertaining as usual. It is a half-size mouse in the classic maze and runs around looking tiny. It has a great party trick of being able to turn around in one cell by executing a small loop. Unfortunately, it had trouble with the walls and did not manage to get a good score in spite of an impressively short fastest run.
Of the runners, only Hulk 1 failed to get a single run time. In the end, the fastest run times were, as you might expect, pretty impressive with the best of all being 8.22 seconds. I can see how I might get my run time down from 12.03 to about 10 seconds but it is not at all clear how I might be able to lose another couple of seconds after that. I shall have to try and find a way if I ever hope to get inside the top five.
Rather than describe the runs in detail, you can see them via Youtube. I have collected the videos, taken by the Lunghua camera, together on this page. Thanks to Ai Jing for uploading them.
Here are the results. In each column, the first number is the score and the number in brackets is the run time. Note that a 3 second bonus is subtracted for a mouse that is not touched except in the start square.
The half-size contest had a smaller set of entries and you can see from the results table that it was rather more tricky to navigate than the classic maze. The half size mice are now quite likely to be miniature replicas of their classic cousins. Gone are the days when these were chunky, awkward devices. Correspondingly, performance has gone up and many of these have very impressive speeds. Note that the scoring here appears to be the same as in Japan. That is, fastest runs are recorded with no search time penalties. As in the classic contest, each person may only win one prize but can enter more than one mouse. The top few run times were all satisfyingly close. Although these mice have similar distances to run, there is somehow more satisfaction watching them run the greater number of cells. One day, I may get to build a half-size. Probably not until I have milked a bit more out of the classic mouse though.
Unfortunately, I don’t have a maze editor that can cope with the half-size maze so a photo will have to do.
Results:[table “20” not found /]
The prize giving gave us invitees one last thing to worry about. At the last moment we were asked to present some of the prizes. Nothing to worry about though, the ceremony was well organised, well rehearsed and went off very smoothly. We watched some of the rehearsals on the Friday and I was disappointed that, on the day, the Taiko drummers only got to perform one piece. In rehearsals, we got a better idea of their repertoire and I had been looking forward to their session.
So – finally the contest was over. I was happy and now feel that my mouse is finally on the right track. Others had mixed fortunes but that is the nature of the event. I may have some difficulty attending the Japan contest in November but I may get another chance in the spring at APEC. My sights are now firmly set.
My flight always leaves very late on the Monday – 23:20 in this case. That makes it a ridiculously long day but at least I do get an extra day to play tourist. On this occasion, Harjit and Pierre checked out very early so we did not see them. Dave Otten and Ng Beng Kiat had the morning free and Khiew Tzong Yong did not leave until the following day. the entire Japanese contingent seemed to have planned for an extra day so they set off for a day out. the rest of us could not resist a trip to the Electric Plaza for a dose of geeky gadgets and components. I did manage to find something genuinely useful in the shape of a LED ringlight but otherwise it was a case of marvelling at the range of things available off the shelf rather than through a catalogue as was inevitably the case at home.
By lunch time, only Tzong Yong and me were left so we set out in search of lunch. Taiwan is blessed with a huge number of small places to eat. However, in spite of my normal foodie bravado, we decided to play it safe and, after a lot of searching found a place not far from Longshan Temple that looked interesting. I had never seen such a thing but the deal was simple enough. You pick a bunch of things from a list – thoughtfully printed in English as well as chinese – and they were cooked up in front of your very seat on a hotplate that formed the rear of the counter. It was all good and, of course, we ordered too much. I did my best though.
After lunch we went to Longshan Temple for a visit. I had been on a previous occasion but Tzong Yong had not. Next we made our way to Taipei Central Station. Here we found a vast food court full of plenty of good looking opportunities for food. We could have save a lot of walking had we thought about it a bit more. So – if you are in Taipei and looking for lunch go to the main station. You are unlikely to be disappointed. they even have a little row of curry restaurants. We took some time over a cup of coffee and talked micromouse for a while and then made our way back to the hotel. I had pre-booked the sauna for a shower and freshen up . This is pretty well essential after a day out in Taipei. While we were there the temperature hit 35 degrees C a couple of times and you don’t want to be embarking on a 18 hour journey after a day out in that.
Unbelievably, it was not until I got to the airport that I managed to get my only bubble milk tea of the entire trip.
On the plane, I was very pleased to find that my new cabin crew friends from the journey out remembered me and we were able to have a chat about what we had all been up to over the weekend. I even got to show off a short slideshow of our weekend. Polite or genuinely interested I may never know but it made all the difference on a 12 hour flight to be able to have something in common with the cabin crew – all of whom by the way were excellent.
The flight back was packed and my neighbour was clearly not interested in any conversation. In fact, she made herself a tent form the blanket, drew up her feet and hibernated throughout the flight.
Here’s looking forward to another chance to visit this fascinating country and take part in another micromouse contest. I am sure I join all the other guests in thanking the Taiwan Ministry of Education for their support, Lubghua Universty for hosting the event and particularly to Su Juing-Huei who is a most generous and excellent organiser and host.