Game AI Tournaments (GAT2023) AI Mini 4WD Competition Event Diary

First published 18 March 2023. Last updated 18 March 2023.

Overview

This was the first time we entered the ミニ四駆AI大会 (AI Mini 4WD competition) of the 8th Game AI tournaments (GAT2023). The event was held on 18th and 19th March 2023. It was organised by 電気通信大学 University of Electro-Communications in Japan. We were very honoured to be accepted to take part in the event. We joined the event remotely from UK.

This year's race format was 10 minutes endurance race. The team that could get their car to go around the circuit with the most laps would be the champion. Pit stops e.g. battery changes, repairs and even track out were allowed, but the time would be included in the total time.

Our Competition Car

Car: Kumamon (Bear Car)

Chassis: MA

Motor: Hyper Dash PRO 2020

Gear ratio: 4:1

Wheel diameter: 2.61cm (standard)

Battery: Eneloop 1900mAh

Setup: Front and rear carbon fibre stay with 9mm rollers. Full ball bearings.

Brakes: Electronic speed control only. No mechanical or friction type brakes. 

Speed limiter: 6 m/s

Mass: ~193 grams

The Race Track

The official race track (FT track) is a Tamiya Junior Circuit (3 lanes) with a slope section.

Unfortunately, we don't have a Tamiya Junior Circuit, but we do have a Tamiya Oval Circuit (2 lanes). We also got a set of wave sections and a set of slope sectors. We hand-made 9 sections of straights.

Based on what we have, I built a track that mimic the the obstacles and length of the official FT Track.

FT track length: ~23.25m

Our track: ~22.87m

It was also the first time I used a lap counter and timer app called LapTimer.

The Runs

I tuned the AI algorithm to give a good balance of speed, consistency and endurance. I set the speed limiter to 6 m/s to conserve battery power. 

I started with the Shirokumakko GT car which had the latest mark 6b technology. During testing and improving the AI algorithm, unfortunately our Shirokumakko GT car had a hard landing after a jump which broke the rear stay.

Since I didn't have any spares with me, the best option was to move the Mark 6b tech to the Kumamon car which had the Mark 5 tech.

I had to opt for a one stop strategy due to the limit of the on-board 32KB memory.

After a few more stable short runs, I decided to do the official run.

The Result

After 10 minutes of nail-biting run with a quick pitstop half way, the car managed to complete 80 laps! Well done Bear Car. Certainly, I could try to get a better result by tuning up the AI algorithm and make it take a more aggressive and risk taking strategy. But since I didn't have any more spare rear stay and it was getting dark (too dark for filming). We will try to get a better result next time. 

Although the result might not be accepted as a formal entry due to the fact that we don't have the same race track, to us taking part is the goal.

Link: https://youtu.be/XEjEDJCRnro

Launch mode: push start

Blue indicator: geolocation beacon

Red indicator: braking

The event was won by a non-AI car, in fact a B-MAX compliant car, by 伊藤研 Ito Ken of 大阪産業大学 Osaka Sangyo University with 81 laps. The second place was won by the same team with an AI car with 69 laps. Congratulations to the winner of this year's event. The current record holder is 99 laps.

Reflection

This was actually the first time we tested our Mark 6b car and it performed amazingly well. We were very pleased to see that it managed to complete the 10 minutes endurance race and the result was very respectable. We noticed the lap time slowed down by about 1 second from start to end due to battery drain and perhaps heat.

We have gained a lot by participating in this event, firstly it was the first time we tested our Mark 6b car; secondly our car has proven that it can run consistently for 10 minutes without any issue; thirdly the data we collected from the event would enable us to improve the car further; finally and yet the most important aspect was that it was great to be accepted as part of the AI Mini 4WD community. I need to brush up my Japanese now.

Thank you to the organisers of the event once again for the opportunity. Looking forward to the next competition and perhaps meeting fellow racers.

Hope you enjoy reading this article. Please feel free to leave a comment on our Facebook page @Mini4Science.