Alcherio Martinoli and his team at the Federal Institute of Technology in Lausanne, Switzerland have achieved a breakthrough in automated driving.
In the future, when cars are connected digitally and able to move autonomously, the so-called automobile will finally do justice to its prefix 'auto' (Greek for 'self') . Humans will play a subordinate role behind the steering wheel – if any at all. This development will take a few years to become a reality. In the meantime, automated vehicles will have to share the road with conventionally driven models steered by humans manually.
This mix between automated and manually driven cars bears serious risks, since the automatically driven car operates exclusively within its own digital world. For this reason, the first fully automated vehicles took to the streets in convoys; one vehicle led the convoy, while the other driverless cars followed behind in one line, like a row of ducks. Although this setup has proven to be safe, a convoy like this cannot be scaled ad infinitum. The structure is also not flexible.
Every vehicle observes its surroundings
Researchers at the Federal Institute of Technology in Lausanne, Switzerland (EPFL), led by professor Alcherio Martinoli, have now tested an innovative method inspired by swarm behaviour observed in nature, as a model to orient numerous automated cars driving at the same time. Martinoli and his team have developed a new algorithm, which eliminates the necessity for a group leader. Instead, each vehicle graphs its own surroundings using sensors, numerous assistance systems, and GPS. The data obtained by each vehicle is then shared with the other cars on the road.
As the automated vehicles can exchange data regardless of their speed and position in the swarm, the vehicle convoy can span several lanes. Also, the vehicle doesn't need to continuously 'follow the leader' in a rigid convoy or reorganise itself every time a new car enters or another leaves the swarm. Nor does it matter if the vehicle in question is an autonomous or a manually driven vehicle; all that is required is an appropriate network and a human-machine interface. Each individual vehicle reacts independently to the behaviour of the other members of the swarm, adjusting its speed and position according to the new surroundings and changing lane if necessary. The individual vehicles also have no difficulty re-arranging themselves if the shape of the swarm changes. The algorithm was developed by Martinoli and his team as part of the European Research Programme AutoNet 2030, dedicated to researching the implementation of automated driving.
When vehicles cooperate
Alcherio Martinoli began dealing with the subject in 1995 when writing his dissertation. At first, he worked with small robots moving in a simulator. The next step was to work with actual mini-robots moving together as a group. "Essentially, what we have is an algorithm that allows agents, which are not particularly intelligent on their own, to cooperate with one another in an intelligent way. The result is a complex and harmonious group behaviour."
Essentially, what we have is an algorithm that allows agents, which are individually not particularly intelligent, to cooperate with one another in an intelligent way.
To do this, the algorithm uses information collected from the sensors of the individual vehicles to direct the swarm in real time. Finally, building on the success of robots tested in a lab a simulator for autonomous driving was developed. Here, Martinoli and his team were able to test the algorithm developed in Lausanne.
Last year, the first test drives took place. "There are two advantages to the new system over previous ones; one is that control is shared evenly without there being a leader, meaning we can dynamically integrate other vehicles. The other is that we have integrated swarm intelligence into this self-organising system." And: "Our approach takes a heterogeneous traffic situation into account, including not only automated and semi-automated vehicles, but also older models without connectivity features. These are unable to integrate into an automated moving swarm and are thus regarded as obstacles, which the swarm must take into account," Martinoli explains.
Real-life test successfully completed
Martinoli and his team performed test drives in Sweden, using an automated truck, an automated car, and a conventionally driven passenger car connected to the other vehicles through a human-machine interface. "With these test drives, we were able to prove for the first time that the results we obtained from the simulator also work in a real-life setting," Martinoli explains. “Furthermore, the number of vehicles in the convoy does not affect the complexity of the control mechanism."
In order to retrofit the 'non-sighted vehicles', a corresponding sensor system, GPS, cameras, and a new human-machine interface are required," Martinoli's colleague Guillaume Jornod explains.
The vehicles connected in the 'swarm' select their own speed and position on the road
For the test drives, the manually driven car had this exact equipment. The vehicles connected in the 'swarm' select their own speed and position on the road. This fleet can also use the lanes on the motorway at will and adjust their position if another vehicle joins the group. Furthermore, this model allows each car to benefit from the 'eyes of its neighbour' meaning that in future, it will be possible to create 360-degree all-round vision. A further advantage of this approach is that, as Martinoli explains, "there is no limit to the size of the swarm because each driver has his or her appropriate position and adheres to the rules."
We have shown that our concept works.
So, what is next? How will the project continue? "We have shown that our concept works. Now, we hope that industry will pick up on the idea and we can further develop this technology," explains Guillaume Jornod. Martinoli estimates that will be at least another 15 years before most vehicles on our roads are driving autonomously.
German citizens view of autonomous driving in a positive light, a survey by the digital association Bitkom has revealed.
Most people in Germany see the self-driving car as a positive development. While travelling in the dark or in congested traffic, they can imagine leaning back, chatting to their fellow passengers or reading a book, and letting the electronic pilot do the work. In a representative survey for the digital association Bitkom, 66 per cent of the participants felt that self-driving cars will bring more advantages than disadvantages.
Participants in the survey named improved traffic flow (44 per cent) and lower energy consumption (40 per cent) as key advantages. 34 per cent were also of the opinion that self-driving cars will improve road safety and that the autopilot function will better protect passengers (27 per cent).
The best way to use the time saved?
And how would German citizens like to make use of this new degree of freedom? A quarter of those asked would spend their time in the self-driving car doing other things. Many drivers (44 per cent) would still prefer focussing on the traffic, as they do currently. Only a minority (24 per cent) can imagine watching films, working (15 per cent), or sleeping (7 per cent). "The car that drives itself was something that, just a few years ago, sounded like science fiction. Today, we are thinking about how one could best use the time after the autopilot takes control of the vehicle," Achim Berg, Vice President of Bitkom, points out in regard to the survey.
Most of those questioned gave carmakers high marks. 76 per cent believe that German manufacturers are at the forefront of development and are comfortably ahead of internet corporations and other companies. "German car manufacturers have a very good starting position in the race to lead the world in self-driving cars," explains Achim Berg.
Clear rules for the use of data
In spite of all the enthusiasm about the new technology, 63 per cent of German citizens are concerned about technical problems arising. Hacker attacks (61 per cent) are seen as the most serious issue. "For self-driving cars to be accepted, we have to guarantee the highest possible degree of safety and, that's why we need clear and transparent rules on who is allowed to use which data, when, and for what purposes," Achim Berg explains.
With regard to data usage, German citizens are split; according to the survey, 48 per cent are prepared to provide their data if, for example, improved traffic flow results or crimes can be resolved more quickly. In order to ascertain the cause of accidents more effectively, three quarters of those asked would be in favour of installing a black box. At the same time, 73 per cent want the manufacturers (including software providers) to be liable. Eight out of ten German citizens also demand a binding definition on how the autopilot should react when an accident is unavoidable.