We took an autonomous taxi through the busy streets of London: the robotaxi will still be waiting


We went to London to discover what could become a robotaxi, circulating autonomously in the streets of the English capital. The project, supported by Nissan, has already learned a lot… but there is still a lot to do.

At first glance, the idea of ​​being transported in a completely autonomous car, robotaxi type, is not really reassuring. We keep hearing from Elon Musk, and other manufacturers, that the machine is more reliable than the human. Except that, of course, we still doubt it. We think in particular of all the possible bugs and the various publicized accidents. What could be better then than to concretely test one of these vehicles, in a lively district of London, to judge the progress made?

Nissan offered us to board the Nissan Leafs of the Servcity project. A way for the teams in charge of this experiment to tell us more about the path already covered in 3 years of study, and what still remains to be done. It’s also an opportunity for us to check if lines of code, lasers and cameras can really outperform an experienced driver. We spoil you a little, the answer is: not yet!

Servcity self-driving vehicle pulls through heavy, inhospitable traffic

As Servcity is in the experimental phase, you should know that operators are always present on board the vehicle, to take over in the event of a system failure. The presence of the operators removes some of the natural apprehension one may have of getting into a car that has no driver, as a passenger. However, it is the car that managed its journey without intervention.

The passage screen as in a robotaxi // Source: Raphaelle Baut

We had the opportunity to observe how the autonomous system behaves in the streets of London on loops of just over 4 km. It’s one thing to test this type of vehicle on an occasional circuit, it’s quite another to find yourself stuck in the middle of traffic in a busy part of London. The map of the test area is still pre-recorded in the vehicle to facilitate traffic. It was not a matter of blind navigation in an unknown sector. It’s a bit of cheating, we agree.

In the environment in real conditions, we rub shoulders with other vehicles: cars, trucks, buses, as well as pedestrians. They do not even notice that an autonomous car is driving in their immediate environment. The car is however discreetly lined with electronic sensors and cameras:

  • 4 laser scanners,
  • 1 radar,
  • 9 cameras,
  • 1 V2I antenna (vehicle-to-infrastructure),
  • 6 central units also occupy the trunk.
Radars and cameras on the Nissan Leaf servcity // Source: Raphaelle Baut
Radars and cameras on the Nissan Leaf servcity // Source: Raphaelle Baut

This test was ultimately a good summary of the daily life of this district:

  • Parked buses;
  • The bulk delivery truck;
  • The heavyweight that deviates from its lane to encroach on ours;
  • The car that crosses the roundabout without anyone being able to anticipate its direction;
  • Pedestrians who cross at red (pedestrian) to get to their metro more quickly;
  • Cyclists riding on the roadway between two cycle paths;
  • Cars that overtake and fall back, forcing the passage.

For an average driver, it’s a journey like any other in everyday life. For a computer, this is an enormous amount of information to analyze and to which it is necessary to react as closely as possible to the behavior of other vehicles: without generating any obstruction to traffic, or without risking creating an accident by an unexpected maneuver by other motorists.

An autonomous car is programmed to follow driving rules, to which, with experience, developers will add thousands of exceptions to manage the unexpected and special cases. A puzzle that, like the AIs, will take time to become sharp enough to really fit into the landscape.

Operator screen in the autonomous car // Source: Raphaelle Baut
Operator screen in the autonomous car // Source: Raphaelle Baut

Several behaviors of the autonomous car are quite exciting: the behavior of pedestrians is well managed, at no time have we really feared for them. But, in discussing with the project managers, we realize that there is still a lot to do. Driving is smooth enough to make you forget that the car drives by itself… except for a few details.

The system lacks some more human reflexes, but not for everything!

Although as a driver our reactions might have been different from those of the self-driving car, the results are quite good. For example :

  • We would have anticipated the arrival on a traffic light at red by lifting the foot of the accelerator quite early, exploiting the regenerative braking of the Leaf, rather than maintaining speed and braking to mark a stop.
  • The system does not see the indicators of other vehicles, it will not understand the change of course of certain motorists and facilitate its passage, unless the system knows that a lane is blocked (by a bus for example). We would have left space and signaled to the driver that he could pass.
  • The system is very careful when entering a crossroads, so still a bit slow, at least more than we would have been in the same place.

The car didn’t make any particular mistakes, it didn’t scare us by being dangerous in traffic. Come to think of it, the autonomous car had a behavior quite close to a driving school student who had already reached a certain number of hours of driving.

There is still a long way to go to achieve anything other than a research project on very marked roads. Servcity is not yet mature enough to move to open customer service, but progress is on the way.

The advantage of an autonomous robotaxi service is that the vehicle will not have the faults of some human drivers. The autonomous vehicle will not start honking, insulting other drivers, or retaliating, which is not always the case with some of our taxi driver friends. On this aspect, we can’t wait to see this new robotaxis competition arrive in cities or on links with airports.

What is the difference with other autonomous cars already in operation?

Many experiments with autonomous cars and taxis are developing around the world, especially in the USA. Except that Europe is probably one of the toughest challenges for this technology. An autonomous system designed to circulate in large American cities could not necessarily adapt to the circulation of our European countries. Moreover, it is certainly not for tomorrow that Tesla’s Full Self-Driving will be able to be used in complete safety and in complete confidence in Europe.

One of the Nissan Leaf self-driving test in London // Source: Raphaelle Baut
One of the Nissan Leaf self-driving test in London // Source: Raphaelle Baut

Major European cities, with a few exceptions, should more easily relate to the challenge of self-driving in the crowded streets of Chinese cities. In Europe, in addition to complex traffic rules, we add all the obstacles linked to the historical constructions of our cities: narrow bridges, tunnels, crooked streets, roundabouts which are not…

Autonomous driving experiments in Europe are also limited by the regulatory constraints of the various countries. The support provided by the British government also made it possible to facilitate the implementation of the tests in real conditions.

Autonomous car test control center // Source: Raphaelle Baut
Autonomous car test control center // Source: Raphaelle Baut

For the purposes of the study, Servcity relies on the city’s infrastructure, cameras, connected traffic lights, to ensure good test conditions. This is how 24 km of roads have been equipped, more than 200 points are monitored and more than 300 cameras are accessible live to follow the progress of the test vehicles.

The Servcity research project has already covered 1,600 miles – or around 2,500 km – of testing in London. If the Servcity experiment proves successful, it could lead to a potential robotaxi service in major UK cities.


If you liked this article, you will like the following ones: do not miss them by subscribing to Numerama on Google News.



Source link -100