Autonomous driving, where vehicles use sensors and computers to automate the driving of a vehicle is perhaps the ultimate use of telematics. What is certain is that since human error is a factor in around 90% of road accidents, the pressure for autonomous driving to help reduce those deaths and injuries will ensure that it becomes a reality in the next 10 years. It could also play a part in reducing CO2 emissions by avoiding harsh acceleration and braking.

We won’t see a sudden switch to autonomous driving. There are still issues to resolve, including technical and legal questions. The expected sequence of events is a gradual rollout with progressively more assistance, before a fully autonomous car becomes a reality.

Motor industry supplier Delphi, which supplies a range of components to motor manufacturers from fuel injection equipment to electrical and electronic components makes many of the sensors and systems used in the driver assistance and safety systems that are at the core of autonomous driving.

Coast-to-coast with 99% autonomy

Earlier this year, Delphi staged a 5,470km drive with an autonomous Audi SQ5 from San Francisco to New York, via, Los Angeles, Phoenix Arizona, El Paso Texas, Dallas Texas, Jackson Mississippi, Atlanta Georgia, Durham North Carolina, Philadelphia Pennsylvania and Washington DC.

Delphi was the first manufacturer to fit radar to a car in 1999. Sales of active safety components (such as electronic stability control (ESC) and active cruise control (ACC) systems are rising and such systems will be fundamental to autonomous driving.

We asked Delphi chief technology officer Jeff Owens what the company had learned from the test. “We did the drive as a learning opportunity. It ended up generating a lot of publicity, but that was really unexpected”, he says. The car completed the journey driven autonomously for 99% of the distance. “That’s a tremendous achievement, but far ahead of where we expected it would be.

“The car has 20 sensors on it and of course you would never do that with a production vehicle,” says Owens. The company wanted to test the various types of sensor – radar, camera, Lidar (laser light sensor) and vehicle-to-vehicle systems, to see which ones worked better in certain situations. 3TB of data were collected during the test, which Delphi is still evaluating. “We found out things that you could only experience on the road,” continues Owens, “The actual real live conditions that you and I as a driver navigate all the time. If you are going to have a computer make those decisions, you have to have that factored into the algorithms.

“For example, we learned that radar works extremely well in all kinds of weather. That’s important. We learned that vision (cameras) is a nice add-on to radar, but vision will occasionally have problems, for instance with a low sun angle and problems in inclement weather because it’s an optical system.

Lane markings

“We learned that lane markings in the United States are quite different in every State. We knew there were differences but they are really different. Sometimes it’s a solid line, sometimes a dotted line – sometimes long dots, sometimes short and almost in all cases they were worn to some degree or dirty, so a little more difficult to see. Some municipalities just have reflective material and don’t even have lines. In any case you rely on your systems to know where the lane boundaries are, so you can position the vehicle correctly. We will take that knowledge back into making our individual sensors more effective.

“A significant part of the learning was the vehicle control algorithms,” says Owens, “You’ve got all the sensor input and you want to bring all this together and evaluate where you are in a total scenario analysis all around the vehicle. To take the place of the driver, you need algorithms that digest all that and then decide, ‘If this happens, I’m going to go here, if that happens, I’m going to brake fast, if this happens, I can’t go here because there’s a guard rail or there’s a car beside me, or whatever.”

All-round monitoring

Production cars for autonomous driving are likely to have a variety of sensors around the car to provide 360° coverage around the vehicle. “Your going to have a long-range radar to detect objects that you are closing on 200m away,” reckons Owens, “When you add vision to that – vision does a few things that radar can’t do and radar does things that vision can’t do, so the two give nice coverage. I think that will be a favourite implementation from manufacturers around the world. Vehicle-to-vehicle input would be a third input and I believe that will be mandated in the United States to come to the roads, probably within four years. That’s another input that you can’t get with radar or vision.

“Then if you go for fully autonomous, you’re going to have Lidar. It’s faster than radar for getting information”, continues Owens, “It gives you some positional information around the car and it also helps you locate exactly where that car is. Today with GPS you can get it down to the size of a laptop, but you need to know precisely where you are to keep you in the centre of the lane or make manoeuvres and that will take something like Lidar, but it’s expensive.”

That will rule it out of the early stages of autonomous driving but Owens thinks that as we progress to the stages where control systems take full control of the car, Lidar will be needed.

Regulation

In technology terms, we have most of the technologies that will be needed to implement autonomous driving but the legal issues are among those that could prove the most difficult to resolve. Owens agrees, “I think that is one of the big problems to be resolved and it’s everywhere, it isn’t just the EU, it’s certainly in the United States and around the world. I think the good news is they want to find a solution.

“Rosekind, the National Highway Traffic Safety Administration (NHTSA) Administrator in the United States, recently called for the automotive industry to make active safety standard equipment on their vehicles. So he fully appreciates that this is a big opportunity to significantly reduce accidents and fatalities. The technology has moved to a performance level where everybody is confident of the performance. The cost has come down. It really is a significant improvement opportunity, but that said, he called for it to be standard equipment, but that doesn’t make it law and that doesn’t set a timetable.

“Everybody is sold on the technology, but the best path to get it in and to navigate some of the regulatory issues and the legal framework, I think that remains to be seen.”

The way ahead could be through NCAP. In Europe only cars with active safety systems can be awarded a five-star rating. The US is following a similar path.