There’s an old saying that the most dangerous part of any car is 'the nut behind the wheel.' Although traditionally applied to road safety, it could just as well be used to describe any driver influenced behavior, from how much energy the vehicle consumes in traffic to how fast it can lap a race circuit.
So, what happens when the human driver cedes part of that control to a machine?
To an extent, of course, we already do. A modern car makes hundreds of decisions every second, from gearshift strategy to thermal management. But automated driving – even in a relatively limited scenario such as adaptive cruise control (ACC) – poses a whole new set of challenges.
As the level of autonomy increases, it could radically redefine the role of the powertrain engineer. It’s been suggested that the motor (or the combustion engine) could essentially become a peripheral, with the most significant impact on performance and economy now coming from decisions taken at a vehicle level. After all, a more efficient inverter design might reduce energy consumption by 1-2%, whereas an autonomous vehicle that drives less aggressively or picks a shorter route could easily deliver 10 times that benefit. Salman Safdar, director of sales and business development at Ansible Motion, explains how this influences traditional vehicle developments:
You [would] typically have a powertrain group and an ADAS or autonomous group working in parallel. There used to be a very distinct divide between those two departments but that’s disappearing now. The new trend is technology groups that have multiple disciplines working together. And once you’re looking at higher levels of autonomy, it’s the decisions from the ADAS that drive your powertrain development.
The link between ADAS and powertrain functions is just one part of an increasingly complex integrated system. With more data flowing into the car, powertrain engineers are better placed to manage the requirements of the vehicle as a whole.
In theory, there's no difference between, say an autonomous vehicle controller easing off the accelerator because it's detected a red traffic light ahead, and a human driver doing the same thing. In reality, we already know that automated systems tend to drive more efficiently in relatively simple cases. But as Nathan De Kerpel, lead engineer for vehicle energy management at AVL notes:
Adding a system like this introduces a lot of complexity to the story, considering the sensors
that are needed and also the control algorithms that have to be integrated. A Level 4 vehicle can have
multiple dozens of sensors [in its perception system], which you have to simulate or test in the loop.
One of the most fundamental concerns [in simulation] is capturing sensor data exactly as it would be perceived in real life. Matt Daley, technical director at rFpro, states the following:
We've reached the stage where we can fool people into driving in exactly the same way in a
simulator as they would in real life. Whatever they do that affects the powertrain . . . we can capture that.
The level of detail required to do this is formidable. Driver-in-the-Loop (DIL) simulation – something of a misnomer in fully autonomous scenarios – can be a powerful tool for studying the human side of autonomous driving. And it applies equally well to assisted and partially autonomous driving. Safdar continues:
Every time I go to China, I see a new car brand on the road. You jump in this car as a passenger and you form your first impressions - but a lot of that can be down to the driver. You feel safe in the car not because of its crash test performance but because the driver imparts a sense of confidence. You could design a very safe or efficient car and drive in a way that's not safe or efficient. The same applies to autonomy. All of a sudden, this becomes the biggest difference between the brands.
As ever, the importance of the brain behind the wheel should not be underestimated.
Read the full article in the June 2024 issue of Automotive Powertrain Technology International.