Just after the first Westpac AWS DeepRacer Championship, I am still feeling to have the momentum to get into another race. And… here it comes!
Comparing to the re:Invent 2018 track, Fumiaki Loop is much more complex with some sharp turns and downward sections which is inspired by Fuji Speedway. And its length is about 3 times of the re:Invent 2018 track.
I only had very limited spare time to train the models for this race. I mainly want to verify if my reward function used in the first race can be applied well here. For the model of final submission, I only trained it for two hours. It was a bit lucky that I can complete 3 consecutive laps without any off-track. I was ranked 12 out of 111 participants. It is said that a present is on the way. What a gift for the coming Christmas!
