Why London is a self-driving nightmare for the Nissan Leaf

Hint: it's got something to do with roundabouts and zebra crossings.


For the people sitting in the lobby of the Aloft ExCel, a premium hotel in the heart of London's Docklands district, it was a typical day. Some were there on business, others to see the city. A handful of parents were patiently waiting to be ushered in for their weekly mother and baby swimming classes. However, in one corner, just out of view, were some of the top executives from Japanese car maker Nissan. They were busy making last-minute preparations for a self-driving vehicle that has been quietly traversing London's streets as part of the first ever European tests. These secret experiments -- the first Nissan has conducted in Europe -- could shape how Brits travel in the coming years.

Over the past seven days, Nissan has welcomed media from all over Europe to sit inside its prototype Leaf and see for themselves how it fares on a pre-set route around east London. I was there to ride shotgun with Tetsuya Iijima, General Manager of Autonomous Drive Development at Nissan, through the streets of Beckton, onto the A13 dual carriageway and back to the ExCel, a short distance from London City Airport and across the river from The O2.


It's not the first time the company has conducted a trial with the car -- they've tested it in Japan and Silicon Valley before -- but European roads pose a unique challenge thanks to an abundance of pedestrian (zebra) crossings, roundabouts and the unpredictable nature of London's drivers. Before fully autonomous Leafs can be sold in glitzy showrooms, Nissan needs to prove it can master each of these obstacles without error.

From a distance, the autonomous Leaf looks like any other Leaf you'd see on London's roads. Up close, however, it's a different matter. Dotted around the car are four lasers, five radars and 12 cameras, most of which have been seamlessly integrated into the bodywork, that provide the car with 20 centimetre accuracy. The two enclosures located on top of the car, which almost look like ears, are obvious but it took me almost an hour to recognise the sensor built into the Nissan badge just underneath the bonnet.

Nissan ProPilot

In the boot -- the entire boot -- lies the processing power. Stacks of computers connected with numerous cables collect and analyse the data generated by the car's sensors. It shows that while much of the technology is available to car makers now, minimising its footprint and integrating it into a production vehicle will take some work.

In the front are three displays. One offers a typical satellite navigation interface and another feeds back footage from the front camera. The final screen shows where the car is in comparison to the environment around it, highlighting pedestrians, vehicles and obstacles as it encounters them. Before we set off, Iijima-san explained that the displays are meant to help passengers; they're not necessary for the Leaf to perform its duties.

On the road

Although the UK is moving swiftly to introduce new laws governing the use of self-driving cars, Nissan's tests must adhere to broader legislation. It means that while Iijima-san doesn't need to touch the steering wheel or the pedals during our 25-minute journey, he must sit in the driver seat and be ready to take back control should the car fail to operate as it should. Nissan notified Transport for London (TfL) and the Metropolitan Police of its plans before the tests and fitted independent front- and rear-facing cameras to capture footage (to present alongside its own data) in the unlikely event of an accident.

"The technology is separate in two parts: common and regional. For the UK, zig-zag lines and lighted crossings need to be understood before we can bring a car to the market," said Iijima-san as we first entered the car. "We need to expand the area of the testing before we meet the regulations and rules of the area. That's why we're here today."

When the time came to depart the hotel car park, the Leaf's autonomous mode was enabled and Iijima-san removed his hands and feet from the controls in an eccentric manner, perhaps playing up to the fact we were paying little attention to the road. Then, to underline the UK's love affair with roundabouts, Nissan decided that the first part of my journey must incorporate no fewer than six of them.

The humble circular intersection was first introduced in America in 1907, England and France followed a couple of years later and made them important parts of their current travel infrastructure. They now host over 10,000 and 30,000 roundabouts respectively, while the US has around 3,000. Roundabouts pose a particular challenge to self-driving cars as no two are the same; some have more than one lane and cover large areas, others are simply a tiny white circle in the middle of a country road. It's Nissan's job to detect their various forms and treat them all in the same manner.

When we approached the first roundabouts, the Leaf was cautious but navigated them with ease. In fact, the first third of the journey was largely uneventful, mainly owing to the fact that the ExCel and the surrounding area aren't especially busy at 2:30pm on a Wednesday afternoon.

That calm, however, was soon punctuated by an accident. Not one involving us, but a collision between a car and an Ocado delivery van on a slip road joining the A13. It gave me the perfect opportunity to see exactly how a machine deals with an unexpected event between vehicles of differing sizes on a fast-moving filter road.

Nissan Self-Driving Car

Approaching the incident, the Leaf recognised that the van and the car were stationary on the right side of the two-lane slip road and kept left in order to pass the obstruction. The radar soon picked up other cars attempting to merge in front of us and the software intelligently gave way. It didn't take very long to clear the accident, but it was clear that Nissan's current programming is designed to be overly cautious. A human driver might not have been so accommodating.

There were no causes for concern on the 50mph A13. The car piloted itself without incident and then exited the carriageway to tackle the final residential stretch of Prince Regent Lane. Here, the street's varying road markings, parked cars and an abundance of pedestrians provide a different challenge.

Bus stops, traffic bollards and zebra crossings weren't enough to phase the Leaf, but two young men crossing the road were enough to make it pause for thought. The duo were a couple of metres ahead of the Leaf and were heading towards the convenience store on the other side of the road. They had barely stepped off the pavement, on the side that was opposite our lane. The Leaf was moving at such a lick that they posed no threat -- by the time they reached our side of the road, we would have been long gone.

It immediately made me ponder what a human would do in that situation. In my experience, a human driver would have acknowledged their presence and identified that they didn't pose a safety issue because there was a significant distance between them and the car. Also, the chevrons in the middle of the road would have provided a natural place for them to pause during their crossing. I posed this question to Iijima-san and he simply said: "This is one of the reasons we are testing in Britain."

Questions that need answers

We returned to base without incident, allowing me to reflect on what I had learned from my brief trip in the Leaf. In April 2016, I tested Tesla's semi-autonomous Autopilot feature in the Tesla Model S and it's impressive to see just how far these technologies have developed in under a year. Nissan's Leaf is already demonstrating a basic mastery of English roads, but its executives admit that its system is prone to mistakes, as a reporter from The Guardian found out. It is a prototype after all.

By 2020, Nissan wants its cars to have mastered single and multi-lane carriageways and intersections before it can deploy its fully autonomous platform ready for the Tokyo Olympics. The company will start by integrating single-lane autonomy into the Leaf and the Qashqai, its two most popular vehicles, later this year. It will then expand the full autonomous ProPilot system to eight more models over the next three years.

To get to this stage, Nissan will need outside help. First, governments around the world need to specify new rules of the road for self-driving car makers. Then, it'll need to direct insurance companies so they are able to cover drivers and the autonomous systems inside their vehicles. The hardest task it faces, however, is convincing the public that their systems are safe and necessary additions to the future cars they will buy.