Hi! We are Romana Challans, Scott Anderson, and Reid Honan – and we are the Karma Engine team. The Karma Engine is actually a Strategy Engine, and is designed to advise, (for this race) what are some good decisions to make based on a complex set of data – where to stop to get maximum points, for example, or when to have (and not to have!) passengers in the car. In the long term, it may even be included in the car itself, and help the car to make intelligent choices about speed etc (but don’t worry, that is a way off yet, and requires a lot more research and development).
We decided we needed a system that was probably cleverer than we are – but that is hard to build overnight! So we decided to use Artificial Intelligence (AI) in the long term, but in the short term, go for a Machine Learning (ML) approach. We know, people find this a bit confusing – AI is about intelligent machines, whereas ML is just about computers being able to learn from whatever data they can find/we give them.
Part of creating this system required us to understand what information would be put into the system. These inputs include everyday values like weather, (such as cloud cover, wind speed and direction), through to more advanced inputs like the tracking of power usage over different environments. We also need to understand some maths, like probability and statistics – so that earlier study has actually paid off!
Our team faces multiple challenges to get this engine up and running. We have lots of different systems to talk to a common point – the Karma Engine. Not only this, but we only have so much processing power to splash around – not like we can lug a supercomputer with using the boot of the car – though Intel has helped us with the generous donation of some really wonderful NUCs (a small-form-factor personal computer designed by Intel)! So we have to train our engine (like a prize fighter!) in advance, and then feed it during the trip to top it up.
We have all the sensors and data we can get our hands on – lots of data, but some of it is in a range, some of it is known data points, some is estimation or prediction (like weather), and some of it is a guess. It all feeds into this engine – and at the end, tells us the best action to take over the race. Where and when to stop, when to pick up passengers (and when to get them out), when to floor it (legally, we assure you!) and when to take it easy.
Sounds simple right? Let’s find out….