FOCUS INNOVATION : Keolis’ fleet of autonomous vehicles

Clément Aubourg


Clément Aubourg is Head of Autonomous Vehicles for the Keolis Group. After business school, his interest in vehicles, innovation, and technology drew him to the field of self-driving vehicles, initially as a consultant from 2016 onwards, and then as a manager in his own right from 2017. At Keolis, he serves as an interface between technical staff and sales, marketing, and legal teams, straddling these two aspects and seeking to structure the business, which was still embryonic in 2016.

Clément, can you tell us exactly what your job consists in?

My main mission is to upskill the group in this area so that we can deliver autonomous mobility for our clients. I’m based at our headquarters in Paris, but I work for all Keolis Group subsidiaries worldwide: my day starts in Australia and ends in the USA, after having visited Dubai and the rest of Europe! We’re privileged to be addressing a range of different, country-specific problems: extreme cold in Canada, extreme heat in Australia, and so on.

The aim is to improve skills, structure the offering, and test vehicles in complete safety. We started with quite simple use cases; now, we’re capable of launching services with no on-board operators.

One of our goals is to be able to provide driverless vehicles, in particular to address the shortage of drivers in many subsidiaries and of course to make this type of solution financially viable in the long term. To achieve this, both the people and the technology need to be ready; above all, this is still a people-centred endeavour. We still need plenty of humans, including drivers, to master this technology, verify that it’s safe, and offer alternatives, including career opportunities for drivers in particular, who account for most jobs within the Keolis Group.

What’s the latest from Keolis about self-driving shuttles?

We’re in the process of launching two projects in their second phase. The first is in France at the Olympique Lyonnais soccer club’s Groupama Stadium, where we’ll be providing two demand-responsive autonomous Navya shuttles for users in the immediate area, where a whole ecosystem has grown up, extending well beyond match days. We’ll be providing a demand-responsive service between the Grand Large tram stop, the station, and various buildings.

Another project which we’re quite proud of is entering its second phase with our Keolis Canada subsidiary, in Montréal city centre. We’re developing a first/last mile solution in the Plaza Saint Hubert neighbourhood: a mobility solution in a high-density area with a range of challenges – a location where as yet there are very few autonomous mobility alternatives.

Our self-driving mobility test site is located in Châteauroux in France, where we trial these new technologies and services. EasyMile is a ‘no op’ (no operator) project with Level 4 autonomy, in other words a service with no operator on board. The facility is private and off-road, so we don’t have to worry about the limitations imposed by traffic legislation.

To what extent is the legal framework impacting development?

In legal terms, it’s now possible for a vehicle to be on the road in France without an operator on board, provided that there’s somebody capable of taking control of the vehicle; in most cases, this simply means being able to stop it. This individual could be described as a remote operator. They must hold a bus driving licence, but they can be 5, 50, or 500 kilometres away from the vehicle in question. The various parts of France’s Mobility Orientation Act that deal with this issue mean that in the near future, a single remote operator will be able to handle more vehicles at the same time.

By the end of the year, the Vienna Convention will be amended to allow this type of experiment to be conducted in signatory countries. However, the technology itself is not yet fully ready. Legal adjustments aside, the technology must be safe and ready, and that’s something we’re working all-out to achieve.

What experiments are currently underway in France and internationally?

I was very impressed with the Las Vegas project: it was one of the first to be carried out on the open road, in 2017, in traffic. We even had people asking us if they could get married on one of our shuttles! It was a way of testing our working approach and a real challenge when it came to providing suitable help from a distance for technology and vehicles that we use in France.

In Sweden, we’re working with Intel and Ericsson to leverage the benefits of 5G technology, in particular supervision. Having a dedicated network and latency under control makes it possible to provide remote assistance, confirming the decisions made by the vehicle at a distance, and doing so safely.

What changes will 5G be bringing?

It’s currently possible to do everything using 4G, but when you see the amount of data required to manage large fleets of vehicles, it quickly becomes apparent that more bandwidth is required. 4G is rapidly reaching its limits. During one of our experiments, we realised that there could be two to three seconds’ delay between a vehicle connected with a 4G camera and one with a 5G camera. That might not seem like very much, but if you’re confirming a decision to overtake, a lot can happen in a few seconds. 5G allows us to be more responsive.


What about cybersecurity?

Cybersecurity is one of our chief concerns. We’re regulated by the French National Agency for Information Systems Security (ANSSI), which has to grant permission for any experiment we want to perform.

For now, the risk is minimal, since we have an on-board operator who can press the stop button to apply the brakes if there’s the slightest problem. What we need to anticipate is the no-operator phase. The smaller the scope of use of self-driving shuttles, the smaller the risks if the system is hacked. The idea is to ensure that only a very few actions can be performed remotely. We’ll be assigning pre-defined missions to the vehicle, rather than allowing it to be directly operated from a distance. Even manual operation is restricted; in some cases, we can continue to enable the sensors to prevent any misuse. In any case, the access points are very secure.

How have self-driving shuttles gone down with passengers?

We’ve had very encouraging results. We’ve conducted surveys in 9 locations in France, the USA, and Australia, interviewing 13,000 passengers: 98% of those surveyed say they are satisfied. There are several factors explaining this: the vehicles are accessible, and the service is free. Of course, there’s still room for improvement, including speed and driving flexibility: the vehicles don’t take any risks, so they may brake more suddenly; frequency also needs to be improved.

How does France compare with other countries?

France is very advanced when it comes to self-driving shuttles; we’re lucky to have a vibrant ecosystem. We’re home to the three largest transport operators (Keolis, Transdev, and RATP), traditional manufacturers such as Peugeot-Citroen (the Stellantis Group), Renault, and Alstom, shuttle manufacturers Navya and Easymile, and others. We benefit from being able to work together, in particular through working bodies for autonomous public transport systems (known in France as systèmes de transport public autonomes, STPA), where we get together to support various initiatives and further the common good.

Will we see a greater share of autonomous vehicles in the years to come? What are the obstacles and hindrances to the spread of self-driving vehicles?

All the signs are that market share will increase, although as far as I know, no regular route has yet been taken over by AVs. All the routes we’ve operated, including those forming part of public transport networks, have been created specifically for autonomous vehicles.

The aim is to develop vehicles that can supplement the existing offering, at night or during off-peak hours. At present, safety and technology issues are the greatest hindrances. As an operator, we’re already making preparations to ensure that our employees are ready, in terms of operating processes, organisation, and training. At the same time, the technology has to comply with a number of parameters and standards, including ASIL levels, based on the number of incidents per kilometre: these are very stringent and difficult to implement. Today, drivers play a very important role, especially in terms of safety. As a result, manufacturers are working very hard to try and provide appropriate levels of operating safety and security without a driver being present.

To have appropriate services that can deliver, manufacturers need to improve vehicles’ Operating Design Domain (ODD). This refers to the conditions in which a vehicle can be used autonomously. To date, these conditions are quite restricted: we can’t go on motorways, we can’t change lanes in an unsecured environment, and so on. Nevertheless, the technology is evolving, and we can now do things that were unthinkable just a few years ago. The challenge is to move on to the next stage and provide efficient service whilst bringing costs down. For the time being, a 15-seater autonomous vehicle costs as much as a 12-metre bus, and the cost per mile is much higher. However, French, EU, and regional funding is allowing us to experiment and be ready for the future.

Will bus drivers ultimately end up being replaced?

I tend to look at what’s happened with metros. The first automated metro dates back to 1983, but we still have metros with drivers because the approach is gradual; we can’t change all buses to make them self-driving overnight. At present, we’re starting with new routes that still require a driver; in the future, the aim is to have specific routes supplementing the existing offering, and one day, to replace existing routes without the need for a driver on board.

Automated metros don’t have a driver on board, but they do have supervisors who know how to control the train, as well as a mobile team that can intervene on site when there’s a problem. We see things going exactly the same way for self-driving buses. We might have two, three, or four people operating a score of buses. All of the 160 drivers that we’ve trained to operate self-driving vehicles worldwide on a volunteer basis were interested in trying out something new, expressed an interest in the technology, and were keen to transition to this business line. Some of them have become trainers, others have become supervisors, and others again are coordinating different experiments. Other drivers are not in the least bit interested: they enjoy driving, and we’re very happy with that, because all of this will happen in stages.

How do you respond to critics of AVs?

No technology is perfect, but this technology has several advantages that shouldn’t be overlooked. Rather than being seen as a risk, it should be seen as an opportunity to have services that initially supplement the existing offering, such as the ability to serve more rural areas that don’t have enough transport links, or to offer demand-responsive transport.