– Article originally published on July 14, 2018 and updated with video on August 15 –
YOU COULD say that the autonomous 1965 Ford Mustang at the Goodwood Festival of Speed had a bit of a shocker. In fact, a number of websites have, after the self-driving system developed by Cranfield University, in collaboration with Siemens, appeared to dart towards the straw bales on at least one of its runs.
And it did the same when we took a ride in it up the hillclimb earlier today. So yes, the system is… less than perfect, shall we say.
There are a number of things that should be known about the car before it’s described as a complete failure, though.
The power steering broke
The ’65 Mustang came with optional power steering, and the autonomous pony car is one that had it installed. Unfortunately for the Cranfield team, they told us, a power steering pipe sprang a leak before the afternoon Goodwood run on Thursday, resulting in sudden changes in pressure. The software had to try and fight that, and clearly didn’t do a great job of it, so Dr James Brighton, who was sitting behind in the driver’s seat, had to make sudden corrections.
It was designed to be wayward
We were baffled when told that the car was programmed to weave along the course. Why? Because the TV production guys thought it would look better on camera if we could see the steering wheel moving on its own. The trouble with this, which was obvious to almost everyone after the first run (and arguably should have been obvious long before that), was that it just made it look like the Mustang couldn’t go in a straight line.
Did they regret this strategy? Well, that’s a moot point once the runs are under way, apparently, as changing the programming last minute would have simply added more uncertainty and potential issues.
It was programmed in a very short amount of time
Six weeks. That’s how long it took the person in charge of the software was given to program it. He spent a lot of time doing it after hours. And he was a masters student, not a professor.
It wasn’t using radar or lidar
Although radar sensors are installed on the car, these were switched off for the Goodwood run as they were seen as unnecessary. The technical lead believes that another car crashing and punching a hole in a straw bale could have altered the environment and confused the system.
Instead, it used two GPS antennas — one front, one rear -— and the signal was known to drop out under tree cover around the flint wall section. To compensate, the car used an “inertial navigation unit” — a combination of accelerometers and gyroscopes. By recording previous movement, these are designed to help work out where the car is and where it should be going in the future.
The most advanced autonomous cars have much more sophisticated laser scanning, radar and camera technology to ‘see’ their path in real time. Inevitably these would have helped the car navigate its way, but again… six weeks.
The TV feed could have been a factor
The ‘stang was fitted with a live feed camera for the TV crews and a whacking great antenna to transmit the signal from the car. This apparently stopped working on Friday and was fixed, then the signal “turned up to 11”, according to the Cranfield guys. As we stepped out of the car, following our definitely-less-than-perfect ride up the hill, they had a lightbulb moment— was it interfering with their own signal? To find out they vowed to get the feed killed for their second run of the day.
It was never meant to be perfect
This was a project designed to get young people excited about engineering and software design. We have a skills shortage in this area and anything that can be done to enthuse young people into science should be encouraged.
This may not have been the best way to advertise this career path but it’s important to remember that they didn’t set out to build a perfect autonomous car; they just wanted to get people excited about the technology. And it got us talking. Maybe give them a break?