It’s the holy grail of grand prix racing: driver and car effortlessly threading their way around a sinuous ribbon of asphalt. Inch-perfect. Every apex kissed. Brakes never locked. On the limit of adhesion, yet under control. Faster than every rival.
Of course, achieving this pacesetting synergy between human and machine is anything but effortless. The lead time for a modern grand prix car is at least 18 months, encompassing countless hours of research, design, testing and engineering.
And even when a new car is finally born – unleashed on track – there is no guarantee it will succeed in the fiercest of proving grounds.
But what if you could master the ultimate human-machine relationship in the pursuit of a performance advantage?
It sounds like the stuff of science fiction, but artificial intelligence (AI) could make it reality by 2050.
Designing a grand prix car sees the generation of a wealth of data from a host of sources: simulation, computational fluid dynamics (CFD), wind tunnel, and not forgetting the track.
However, these data sets do not dovetail elegantly. Each has its own nuances. They are different virtual representations of reality. And the grand prix teams that enjoy the most success are those that best understand how all these data sets interact.
But human identification of these relationships can only go so far.
Take a new front wing design, for example. Wind tunnel testing shows that it’s going to bring about a marked increase in downforce. You bolt it onto the car in free practice one, but it doesn’t deliver the performance expected. The data from the track clearly at odds with that from the wind tunnel. The two don’t correlate and that’s a big problem.
How do you find the reason for this, and even avoid it from happening in the future?
The answer is AI.
Its ability to identify hidden correlations, patterns, trends, or behaviours between data sets will make it a key player in the successful design and engineering of the future’s racing machines. It can learn the different relationships between every data set – something which is very difficult for a human to do. For example, mathematically linking wind tunnel data to a simulator is no mean feat, but with AI you can model what’s happening in both and see how they relate to each other.
Grand prix teams will race to develop the most intelligent and flexible algorithms, powered by seemingly limitless computing resources, to pull together these data sets and create a single representation of how their race car behaves. The combined data set would be so rich that teams could run a variety of simulations in different configurations, making it an incredibly powerful tool to design and build the car.
Codifying human ingenuity
When you design a car, a significant amount of it is human-led. But how do you replicate human ingenuity? That creative spark which has been pivotal to every successful design? From Leonardo da Vinci to Adrian Newey, they all have a vision, a personal preference.
Personal preference extends to the person driving the car. It must be compliant and predictable out on track. The numbers from the wind tunnel and CFD may be good, but if it doesn’t inspire a driver with confidence, it’s about as useful as a chocolate teapot.
The science of designing a grand prix car is not purely an optimisation problem. It’s also an art form. The constant attempt to balance the two is where the inherent beauty of these racing machines lies.
Currently, when we look to optimise a part of the car such as front wing or suspension geometry, there is a distinct and significant amount of human input from engineers and designers. We end up creating something borne out of human ingenuity and optimised by an algorithm.
However, in the future we could get to the point where human ingenuity is replaced with an AI algorithm.
Machine learning would see human preferences and decisions, as well as our domain expertise and instinct, captured. Take enough examples of our creative processes and outcomes, and this could be codified into an algorithm which would enable AI to make creative decisions consistent with those of a human counterpart.
Couple this with a model which focuses on understanding and identifying the relationships between data sets, and you have two very powerful models which could interact with one another in a closed loop: the optimiser and the human preference.
The resultant reduction of human interaction in the design process would lead to massive efficiency gains for grand prix teams, and rapidly decrease design lead time and cost.
The race for the best AI
Should such a step change in machine learning take place, a shift in focus from hardware to software to obtain performance advantage is highly likely. The race to create the best AI will be paramount to succeed on track, especially if there were to be more hardware standardisation in the sport in 2050.
Extracting the maximum performance from standardised hardware will rely on driver skill and the best software. The algorithms embedded in the car will become a key performance differentiator.
One of the most exciting things about this, is that it’s very hard to copy software innovation. Currently, if one team innovates a new front wing, the rest of the grid could have it on their cars within three races. If a team devises an innovative piece of software that delivers a performance advantage, rival teams wouldn’t necessarily know the software is even there.
The benefits associated with AI go beyond the creation of a grand prix car. Its applications far-reaching and relevant to every industry. But crucially, grand prix racing can position itself as the ideal innovation lab for such technology, and this will encourage an influx of companies wanting to develop their software in the toughest proving ground.
The ultimate human-machine relationship
In 2050, complex machine learning algorithms will represent one of the most exciting development paths for the sport, as well as wider industry and society going forward.
Grand prix racing will see an increasingly blurred boundary between driver and car as the two interact in unison, be it on the track or off it. Human and machine continually evolving based upon each other’s behaviour.
Ultimate performance will remain firmly in the hands of humans, as the best drivers and teams will not only develop the most powerful AI, but crucially, know how to use it to their advantage. Shattering the notion that machine learning could render the human element of the sport obsolete.