The intelligent race is on with AWS Deepracer

14th July 2022

Sheryl Miles

0 0

Scientists, engineers, and developers are just some of the participants who partake in the AWS (Amazon Web Services) Deepracer Challenge. It is, however, a machine learning race that is open to everyone who would like to get involved.

Participants programme a 1/18 scale model autonomous vehicle using AI technology, and the basic aim is for participants to race the vehicle around a track in the fastest time possible.

AWS Deepracer consists of four tracks, and participants can race their vehicle in a simulator or on a track – or both, and it has the option for league racing. Of the four tracks the longest is their Summit Speedway measuring at 25.236m long.

However, the real challenge lies in taking the vehicle and using Reinforcement Learning (RL). Reinforcement Learning means participants ‘train’ a machine, in this case a scale-model car, to make split second decisions whilst racing around the track, which means the cars must make choices in an uncertain and complex environment. The technology uses trial and error to optimise the best possible outcome.

But once the vehicle is racing, the only control participants have over it is the ability to control its speed.

Dtsl organiser David said: “In DeepRacer someone is controlling the speed, but the car is doing all of the rest of the navigation and that is all powered by the AI developed over the last few days. The idea is that you go as quickly as possible around the track. There are countless different approaches you can take.”

The vehicle is programmed to knows its objective is to win the race, and RL uses a penalty or reward system to encourage a machine to make its own decisions. The objective is for the machine to get more rewards than penalties.

This type of challenge is a wonderful way to make RL fun, interesting and engaging.

Recently, participants from the Defence, Science, and Technology Laboratory (Dstl) took part in the race and came close to the world record time with a top virtual time of 7.865 seconds and a top track time of 8.069 seconds. To add pressure, participants had only three days to programme their vehicles.

Participants from the Dstl found the differences from the virtual race to the track race, such as lighting and feel of the track, effected their vehicle in unexpected ways.