Picking up speed on the road to fully autonomous vehicles
04 May 2016
Author: Guy Bird
I'm whizzing down a Californian freeway at 60mph with US-sized pick-up trucks hurtling past on all sides, in the back of an adapted Nissan Leaf with a man in the driver's seat who seems to have an aversion to holding onto the steering wheel.
At all. He stays calm, though - barely touching the controls throughout the 20-odd minute autonomous vehicle test drive - so eventually I am too. And we survive. Welcome to the world of autonomous driving 2016-style.
The car in question on this early media test event is Nissan's latest 'Piloted Drive' prototype, which is a lot like a regular electric Leaf, aside from a boot stuffed with extra computer hardware working in conjunction with exterior-mounted cameras (in the roofline), laser sensors (each side), radar (in the bumpers) and GPS satnav, to keep the car heading north, south, east and west, and knowing when to brake, accelerate, turn or stop completely.
The test route is as multi-layered as any normal global commute might be - taking in smaller neighbourhood roads, navigating complex crossroads and filtering onto real-world, multiple-lane highways - and aims to show just how far Renault-Nissan's technology has come on its path to full autonomous driving.
The fact that the car pretty successfully drives itself (see 'Hands-free on real roads - really', for caveats) is incredible enough, but the timeline to making these cars showroom-ready for the wider public is perhaps more surprising still.
In a speech at the Renault-Nissan Alliance's Silicon Valley Research Center in early 2016, charismatic Renault-Nissan chairman and CEO Carlos Ghosn said cars from the group able to drive autonomously along highways with 'single-lane control' will debut as early as this year, even in stop-go heavy traffic.
Indeed, BusinessCar has already heard Nissan Europe's chairman Paul Willcox saying on record that a Qashqai able to do just that will launch in Europe from 2017. By 2018, Ghosn says 'multiple-lane control' autonomous vehicles will begin launching worldwide, able to change lanes and avoid hazards on highways, and by 2020 he says Renault-Nissan will have cars with 'intersection autonomy' that can traverse cities in complex urban areas, including pedestrians, cyclists and more.
This 2020 pledge was reiterated in a mid-April announcement from the Alliance, with the added icing of having "the ambition to become the first to offer 'eyes-off/hands-off' technology on mainstream vehicles at an affordable price".
The last element is one of a few sticking points, as Ghosn admits: "Scale affects price for mass market and we know what is acceptable to the consumer, so we've set targets for our engineers. We're not there yet in terms of cost. This prototype is 100 times more than the regular car."
Understanding cultural driving differences will also need to be considered for the technology to work worldwide, as director of Renault Innovation Silicon Valley Serge Passolunghi wryly acknowledges: "A too-shy robot in Mumbai would be stuck in a traffic jam all day. but artificial intelligence can solve this."
Ghosn concedes there are other technical issues to solve too, including how the car 'reads' the road in torrential rain or heavy snow, where sensor glass might get wet and blurry and where road markings and signage can be obscured, but he, and Renault-Nissan, remain bullish about its publicly declared timeline, pending global legislators allowing such cars to exist on regular roads beyond these limited trials.
"By 2020, we'll have the car," Ghosn says. "Whether you'll be able to buy and drive one is another matter. But you need to prepare the public to accept this technology." Over to you, Mr Global Legislator.
Flying along on Tesla Autopilot
While Renault-Nissan is working on technology, another company leading the way on autonomous driving is one that has shaken up the motor industry in a whole series of ways: Tesla, writes Paul Barker.
Other brands, prestige ones in particular, have probably had enough of Tesla.
The Model S executive car has driven a host of innovation for the industry, from EV range expectations and infotainment to the basic ability to upgrade a car via wi-fi in a way more common with mobile phone apps.
So it shouldn't really be a surprise that the firm now has its own new self-driving technology, called Autopilot.
Available now, either as a £2200 option on new cars or retro-fitted for £2600 to cars already delivered, the new software was developed in the company's version 7.0 software update.
Using a forward radar and camera, 12 long-range ultrasonic sensors scanning 16ft in all directions, and a digitally controlled electric assist braking system, Autopilot has the ability to steer within a lane, manage speed using active traffic-aware cruise control and, most impressive of all, change lanes - all with the car controlling steering, brakes and acceleration.
Tesla advises Autopilot should only be used on motorways at present, although it will work on other roads. It also advises the driver to keep their hands on the steering wheel at all times, yet only when changing lanes does the car insist that hands are on the wheel.
The rest of the time it is possible to sit back and relax - at least in time, with more experience of using the system. During our test drives we found ourselves hovering an inch away from the steering wheel and ready to hit the brakes in case the car missed something. Spoiler alert: used properly, it didn't miss a thing.
Activated by a double-tug on the cruise control stalk, the steering wheel icon turning blue is the indication that the car has assumed control. Anyone who has experienced adaptive cruise control automatically slowing a car when it senses a slower vehicle ahead will be aware of the basic sensation involved in letting the Tesla take control, except this is on a much wider scale.
It senses whether the vehicle ahead is a lorry, car or even motorbike, and can use another vehicle as a back-up to judge where the road goes if the white lines are faded too much for the sensors to read. The system will operate at up to five miles an hour above the speed limit.
The steering function takes a bit of getting used to, and the car does tend to sit closer to the edge of the lane than a human would position it. But it successfully negotiated the curved junction 5 of the M25 with no driver input as part of a run round the bottom section of that motorway, where we didn't have to override the Autopilot controls once.
The cleverest bit, though, is the lane-change function. Indicate and the car will check for other traffic before changing lanes on its own; as long as the driver is resting their hands on the wheel, the car will complete the manoeuvre independently. It is an odd sensation, as it switches, generally seamlessly, from one lane to another and then back again when you've passed the other vehicle and indicate to move left again.
It's more cautious than a human and won't pull out when it detects a car coming from behind, even if the vehicle is far enough back to have time to get out. That can cause confusion and hesitation from other motorists, but can be avoided through well-planned moves and not indicating to pull out when other cars are bearing down. Fundamentally, the Tesla won't filter into a line of cars, and won't recognise that another driver has left a gap or 'flashed' you out, but drive in a manner that avoids such scenarios and there's no problem.
Essentially, the system works very well, and for a first development is impressive. Modifications, enhancements and expansions will obviously come in the future, and there are some creases that need ironing out.
My personal feeling was that the car could be more blatant about when it has relinquished control. That's a major piece of information, and there were times where I wasn't sure if I'd made an input on the steering wheel that had overridden Autopilot. The steering wheel icon changes colour, but a more obvious alert would help, in particular, those new to the system and the feeling involved with handing command, and to some extent personal safety, to a machine at high speed while surrounded by other drivers.
It feels like that is the point where any incident could occur: in the transition between Autopilot and human control.
Don't tell Tesla, but we also tried it on single-lane roads, where the car comes a little unstuck trying to work out where the edge of the road is. It does a decent job, considering it's not ready for that level of interaction with the world around it, so it will sooner or later be caught out.
However, traversing several junctions of the M25 without having any input is a revolutionary experience, and especially relaxing in traffic where every lane is plodding along at similar speeds. Rush hour, where regular lane changes on busy motorways are required, aren't its forte, and in that situation it's easier to do the lane changes yourself. That aside, the biggest problem is relaxing enough to trust the Model S do its thing.
TRL talks fleet benefits
Autonomous cars could make driving for business much safer than it is currently, Nick Reed, academy director at Transport Research Laboratory, tells Daniel Puddicombe.
According to Reed, TRL has done "a lot work" looking into the risks for fleet drivers, adding that even when the higher mileage that these drivers undertake is taken out of the equation, business users are more likely to be involved in a collision when driving for work than when driving in their own time.
Reed says there any many reasons why fleets are more at risk than private motorists, including business workers being more likely to be driving under time pressure coupled with a greater chance they are driving on unfamiliar roads, not to mention the increased likelihood of having to take an important phone call while driving.
So how will driverless cars help? "What you can imagine with automated vehicles is that they deal with a lot of those challenges - you would be able to engage with phone calls, and not have to worry about finding your destination, because it is pre-programmed, and you would be aware of your time of arrival because [the car] would have a very clear understanding of its route and the time taken to get to its destination," Reed states.
"All of those things would make driving for business much safer, so I can imagine that when these vehicles become available, purely on a health and safety basis, companies would be very keen to use them, and if I'm thinking about it even more broadly, employees could use the time travelling as you could do on a train - productively," he adds.
Reed also claims that driverless pool vehicles would be cheaper to run on a mile-for-mile basis than one driven by a human, due to efficiency savings, including being able to share vehicles more effectively.