Self-driving cars could be steered towards business use cases

The real value of autonomous vehicles could lie in industrial applications. We take a test drive around Milton Keynes to find out

The writer after a short ride in a Renault Twizzy autonomous car, Milton Keynes, October 2016

Milton Keynes’ Station Square is hosting a very small fairground, with a tent, a few flags and just one ride: the Catapult Lutz Pathfinder autonomous vehicle. Like all the best fairground rides, the self-driving car – a tiny two-seater electric Renault Twizy, painted in white and purple, and emblazoned with the logos of organisations – plays with your mind rather than putting you in physical danger. The only crash is software-based, and happens before we start.

A quick reset later, the “driver”, who has driven just 700 metres of the 30-40km he has spent behind the wheel over the past few days, taps on an iPad and releases the handbrake. The wheel turns itself clockwise, steering us into one of Milton Keynes’ many pedestrian underpasses and along a shared pedestrian and cycleway.

Taking part in the first UK trial of self-driving vehicles in a public space starts as unnerving, but quickly becomes fun, not least in watching the surprised reactions of pedestrians. The short route avoids roads but has seen the project’s “Selenium” software coping with pedestrians, dogs, cyclists and pram-walkers. It does so by jamming on the brakes, producing the ride’s only moderately thrilling moments.

This is because one of the car’s two systems, the low-level computer, is designed to bring it safely to a halt unless the main vehicle computer has a convincing plan to move it. Paul Newman, BP professor of information engineering at Oxford University’s department of engineering science, describes the low-level computer as a “doubting Thomas”, one of a range of features that makes the software capable of driving autonomous vehicles reliably.

The software does not require signals from mobile networks, satellites or beacons, although it can connect to 3G networks for fleet management. Instead, it relies on the vehicle’s cameras and light detection and ranging (Lidar) sensors to drive, meaning it could work indoors or underground.

Newman, who has been working on the project since 2010 with Engineering and Physical Sciences Research Council funding, adds that this demonstration prioritises safety, and that the software is capable of much faster speeds. Today’s trial is satisfying nonetheless. “You’re tapping away on some code with some great colleagues, and then that text file gets turned into something that moves metal around a city waiting for a child to get out of the way. And the public come and poke at it and ask questions. That’s an engineer’s paradise, right?”

The Lutz Pathfinder has been funded by the government-supported Transport Systems Catapult and a company, Oxbotica, has been spun-off from Oxford University by Newman and others to market Selenium. He says that the hardware – both the car and its commodity computer equipment – is unimportant compared with the Linux-based software, which his team at Oxford aimed to make both highly stable yet open to innovation.

“You have to build an architecture that can be sympathetic to an invention that’s going to happen six months in the future, and it’s going to slot right in,” he says, as well as maintaining high standards of code hygiene, coding styles, inter-process communications, security including use of encryption and “a huge amount of effort” on testing.

Selenium has been designed to consider three basic questions: “Where am I, what’s around me and what should I do? Each of the application domains has each of those three questions coming up in different ratios,” he says. A fork-lift manufacturer may not have to worry too much about what’s around the vehicle, as it will normally operate within a warehouse, but it may have to cope with poorer lighting conditions and more enclosed spaces.

Newman believes entirely autonomous cars that don’t need steering wheels on public roads are “a long way away”, but that there are lots of intermediate applications. One is providing transport as a service, and Milton Keynes Council, which is hosting this trial, hopes to offer autonomous vehicle rides from the railway station and other transport hubs to city centre locations from the end of 2017.

“They are a last-mile solution,” says the council’s director of strategy, Geoff Snelson. “You can dial up one of these pods on an app, it will pick you up and take you where you want to go, drop you off, and then tootle off and pick someone else up. It starts to remove the need to bring lots of cars into the city centre.”

Milton Keynes suited to autonomous vehicles

The council wants Milton Keynes to be a test bed for autonomous vehicles. Its grid pattern of streets makes it easier to shut roads temporarily to things. “It’s a little bit more straightforward than trying to run things in the middle of Oxford,” says Snelson, who notes that much of the land is owned by the public sector which makes it easier to install sensors. Jaguar Land Rover, Ford and Tata also plan to test autonomous cars in the city.

Milton Keynes Council is also creating its own three-dimensional map of the city with a local company to help autonomous vehicles navigate, as well as to integrate them with public transport. “You need to allow for flexibility and different sorts of futures, the integration of multiple solutions and multiple providers,” says Snelson. “What you don’t want to do is have Google or Microsoft own your town, come in and own the digital map of the place and the technology, making you beholden to them. You need to be in control of this stuff.”

Self-driving vehicles could raise productivity in mines

Like Milton Keynes, the first commercial customers of autonomous vehicles are providing a supportive environment to help them work. In September 2016, Volvo carried out what it believes were the first tests of an autonomous truck in an underground mine, which included its chief technology officer Torbjörn Holmström standing in the way of a truck to see if it stopped (it did).

Boliden, the operator of the Kristineberg mine in northern Sweden, had installed a wireless communications infrastructure and implemented a site management system, planning vehicle movements and tracking staff. “That’s why we selected that mine to start with,” says Christian Grante, technical specialist for preventative safety and automation at Volvo Group.

The trucks use underground beacons as part of a floating vehicle concept where they are all tracked. Using the mine’s data, a truck can go faster in areas with no other vehicles or people, although four on-board Lidar detectors can apply the brakes if necessary.

Volvo Group says automation is particularly valuable to underground mining, as it can cut the high per-person safety costs of working underground and allow productivity to increase.

“The hauling system in the mine stops for shift changes and breaks,” says Grante. “If you run automated, you run 24/7 without the people in the mine. They are on the surface instead.” The trial at Kristineberg suggests the mine could double its productivity by moving all staff to the surface, automating both vehicles and the mining process itself.

Volvo Group also plans to offer autonomous vehicles to other closed-site operations, including open-pit mines and quarries, harbours and logistics hubs.

“You can change the rules and regulations overnight, and also adapt the infrastructure,” says Grante. Potential customers can prepare closed sites by introducing site management, and dispatch systems and supporting network infrastructure. “If you have a site that is digitised, where the information is digitised and you do the planning digitally, then you have the core parts,” he says.

But on public roads, while 80% or more of long-haul driving is relatively straightforward to automate, Grante adds that unforeseen situations are “almost impossible to automate”, meaning that such technology is likely to support rather than replace lorry drivers. It will improve safety by helping them to avoid mistakes or falling asleep on motorways, letting them focus on difficult but stimulating urban driving. This could allow regulatory changes so they can drive or supervise autonomous vehicles for longer continuous periods.

Agricultural applications could put more money in farmers’ pockets

Some autonomy is already a reality in agriculture. Many tractors and combines steer themselves, although operators control the speed and vehicles will not operate without someone in the driving seat. This allows operators to monitor agricultural operations, such as seeding, which can run through 20 tubes at the back of the vehicle. Blocked tubes mean less of the field will produce crops, and a second pass would damage what has already been sown.

Letting operators pay more attention to the tubes therefore boosts yields. “This is where the money can be made for the farmer,” says Michiel Jochims, Real Time Kinematic manager for Europe, the Middle East and Africa at CNH Industrial, the parent company of agricultural vehicle-makers New Holland and Case IH.

Many modern tractors and combines can also automatically operate implements by switching them on and off at field boundaries so they avoid wasting materials such as fertiliser, thereby reducing costs and environmental impact. Making operations easier for farmers also allows them to harvest crops in less time, minimising weather damage.

Agricultural vehicles have relied on satellite-based systems such as GPS, but this technology’s level of accuracy to within a few metres can involve significant waste. To improve this, CNH Industrial is installing fixed-location base stations across Europe. Its Real Time Kinematic positioning system involves the fixed stations comparing GPS readings to actual locations, allowing the company to generate error corrections every 10 seconds. These are sent via GSM networks to customers’ on-vehicle equipment, increasing accuracy to a couple of centimetres. In the UK, CNH is using more than 110 stations run by Ordnance Survey.

Jochims says that, at present, semi-autonomous agricultural vehicles focus on locational accuracy rather than safety, as they still have operators who can look out for animals or people in the way. But CNH is considering fully autonomous, cableless vehicles, currently at the conceptual stage, which would add sensors.

“The machine would drive itself 24/7,” he says, from a farmer’s yard via public roads to a field. It would operate implements, identify and avoid obstacles while working at the field, then return to base.

“As a farmer, you can control the machine at a distance, you can give it tasks, you can focus on other duties,” says Jochims. But while the technology to achieve this is available, there are issues around insurance, regulations and reliability. “It’s all about who’s going to take the hit if something goes wrong,” he adds.

Despite the focus on driverless vehicles on public roads, these concerns, along with the difficulties of driving in urban environments, mean that autonomous vehicles are likely to be adopted more quickly on closed sites. But operators willing to adapt such sites to work for autonomy could reap significant improvements in productivity and safety.

First published by ComputerWeekly.com, 18 November 2016