top of page

What is Amazon Deep Racer? Why is it the Best Starting Project for Learning Artificial Intelligence?

Updated: Feb 14, 2021

Have you heard of professional race car competitions where hundreds of race cars drivers train for years just for a spot on the line to drive a few laps around a track? This has been a cornerstone of individual sport competition and one where a high level of training and skill is required to compete.

Amazon DeepRacer takes the same idea: a race car, a track, a competition, . . . but no driver.

The competition was first hosted in 2019 and is now approaching it's 3rd year in 2021, where thousands of participants have a chance to train their own DeepRacer car to compete for the Championship. The car drives fully autonomously, meaning it makes real time decisions while racing around the track. You might wonder: how can a piece of metal that does not have a brain reach this high of a level of caliber in thinking? Here is where the groundbreaking invention of artificial intelligence becomes essential to this car knowing how to drive on its own.

The DeepRacer car is like a child who needs instruction on what to do or what not to do; it needs to learn what is good and what is bad. For example, it needs to understand that following the dotted yellow lines and turning its wheels accordingly is the correct thing to do, opposed to just pressing the gas full speed and crashing into our washing machine in our garage (which it has done in relentless oblivion).

Based on repeated instruction, feedback, correction, and reward, the car begins to learn what to do and drive around the track based on the previous training you gave it. This cycle of stimulus, instruction, response, action, and feedback is called Reinforcement Learning. Reinforcement learning is one of the 4 types of machine learning, an "application of artificial intelligence (AI) that enables systems to learn and advance based on experience without being clearly programmed".

Check out this link for a more detailed and concise explanation of different types of machine learning and types of AI.

The central idea is that through training and repetition of doing the correct actions such as following the dotted yellow lines and getting rewards based on its behavior, the DeepRacer will soon be on its way to laying down laps with unmatched speed.

Why is this project such a quintessential and appropriate way of jumping into the arctic waters of AI? Well, simply put, you can not only see but also experience working with artificial intelligence and machine learning in a very tangible way. DeepRacer allows you to fully be a part of every step of the car's learning process and understand the basics of how humans train robots to make decisions on their own. This is all based on the combination of stimulus, instruction, action, and feedback, in which you can quickly learn that there are a million ways to approach it: ranging from what instructions you give to how much reward is given, you can adapt an effective method to making the car more fast and accurate (and beat other people's models on the way). By seeing the physical manifestation of AI in a cohesive tangible project such as an autonomous car racing around a track you can set up in your own garage, it makes learning AI 1000000 times more fun. It connects the idea of artificial intelligence to something you can interact with, engage in, learn alongside of!

So, if you are the least bit intrigued about how a piece of metal (designed by the expert people at Amazon of course) can seemingly charge in oblivion around a garage sized track all on its own, you can start making your own model today! Only if you want though

131 views0 comments


bottom of page