The new technology with potential to revolutionise F1's spectacle

Society has been through many revolutions over the years. We tend to think of the industrial revolution which arose in 18th century England as the birth of the technology-based life we enjoy today, but there have been many others. The development of ever more durable materials,
 the discovery of electricity and the move from nature-based motive power to science and engineering-based systems were ultimately enablers for the sport we love today.

The rise of cheap computing power has changed many aspects of our everyday life and has a profound effect on the design and operation of race cars and race teams. The time of the intuitive engineer making relatively simple calculations longhand has been replaced by highly trained individuals with enormously capable systems which can analyse any problem to the nth degree.

One might argue this has taken away the innovative creativity of great designers such as Colin Chapman or Gordon Murray as the emphasis switches to marginal gains rather than breakthroughs brought about by lateral thinking. Equally, one might argue that creativity in today’s motorsport is just altered by regulation. Today performance often comes from acquisition of data and interpretation and exploitation of that data.

This thought might bring us to considering the next revolution already touching our lives in so many areas and, while its use in motorsport is currently just beginning, will change our sport as well. The technology I refer to is machine learning, a specific implementation of artificial intelligence or AI.

Some years ago in this column I wrote a somewhat pessimistic view of where the application of artificial intelligence techniques may take the enjoyment of sport. I argued that if these techniques were infallible then they could only lead to a singularity of solution – and therefore if each designer fed an artificial intelligence system with a set of Formula 1 rules, we would end up with 10 identical cars. Equally if each team approached the race guided by similar algorithms then the jeopardy intrinsic to sport would be lost.

This is something to keep an eye on but in the first instance we’ve been using structural optimisation software for many years now and yet not every suspension component or brake calliper looks the same. Now, optimisation software is slightly different in that it takes an iterative approach to improving a design rather than a deep understanding which could, theoretically, get it right first time.

Driver feedback is being used alongside computer modelling tools to decide set-up and strategy

Photo by: Steve Etherington / Motorsport Images

However there is a branch of AI that can, and indeed is, adding to our enjoyment of sport. This is machine learning or ML. It’s worth understanding the difference. Artificial Intelligence is a generic term for software which operates much as the human brain does. It will perform complex tasks and learn from them.

Machine learning is a subset of AI that uses algorithms trained on existing historical data to produce adaptable models which can then be used to predict outcomes of a series of events. The key factor is ML needs data to work on and therefore cannot innovate in the same way AI can.

ML is already assisting many of the insights we see on the TV feed. 70 years of data has been stored and used to train models. Yes, the data from early F1 races is sparse and has far less relevance than that of the past few years, but the secret of ML is to feed models with as much data as possible – history has a habit of repeating itself.

The data is now used in many ways. Let’s consider one simple example, the cut-off time required to get into Q3. The simplistic approach would probably be to wait until everyone had done one run in Q2, look at the time of the 10th car and declare that as the cut-off time.

An ML system would look back at years of data and realise that a driver will generally improve on each run; it will understand, from previous data, how the track temperature may change and the effect that will have on lap time; it will know about different engine modes and maybe, in the future, it may even be able to look at driver biometrics, their heart rate, core temperature and stress levels to see whether the driver is having a ‘good day’ or a ‘bad day’, and how this may affect his performance.

If predictive modelling reaches the level where it has a 99% chance of forecasting the outcome of the race it’s likely to turn off a large number of casual fans

Overall, machine learning has the potential to revolutionise all sports by providing insights into performance and probable strategic and tactical outcomes. For many this adds a level of understanding which boosts the enjoyment but it should be treated with  caution.

Jeopardy is a factor that makes sport what it is. The unexpected result, particularly if it involves a level of ‘giant killing’, is attractive no matter what the sport. If predictive modelling reaches the level where it has a 99% chance of forecasting the outcome of the race it’s likely to turn off a large number of casual fans.

It’s likely that the first to really exploit machine learning are the teams, and to some extent this is already happening. They collect vast amounts of data. If human inspection were the only analysis tool available then only a small fraction of this data would be analysed. Automatic analysis tools are developing from being mere reporting devices to being decision makers. This ability to see complex patterns and not just draw conclusions, but to go a step further and make recommendations, will reduce errors in set-up and strategy.

Overall, machine learning has the potential to revolutionise the world of sport but particularly Formula 1 since it’s so rich in data. This will make F1 faster, safer, and more exciting than ever before. As the technology continues to develop, we can expect to see even more innovations that will change the way we enjoy our favourite sport.

F1 has only scratched the surface of what artificial intelligence and machine learning can achieve

Photo by: Steve Etherington / Motorsport Images