Conversation with Alberto Broggi, General Manager VisLab
Automated vehicles during the Coronavirus pandemic
by Domenico Crocco*
During the Coronavirus pandemic, an autonomous vehicle such as those that have been circulating in many countries around the world for some time allows many things: if you are in quarantine and cannot move from home, it can go independently to the pharmacy to pick up your medications, pick up your groceries, and transport latex masks and gloves to hospitals.
Obviously, with technical and scientific spirit, in the international tables of the World Road Association (PIARC) in which we participate, experts in automated driving and smart roads have also evaluated the zero option. They have asked themselves whether the benefits of the automated vehicle and smart roads are greater or less than the costs. They also asked whether we could do without them. Whether it is really worth it. And the main reason they were persuaded to say yes wasn’t due to the driving comfort. It wasn’t the chance to work quietly in the vehicle. It was the dramatic data on road safety. One million 200,000 dead, including 260,000 children and 59 million injured in the last year alone worldwide. 25 thousand victims in Europe, about 3,300 in Italy. Accidents that the automated vehicle could really eliminate. In fact, 94% of these accidents were caused by the human factor: distraction, use of mobile phones at the wheel, excessive speed. For this reason, European directives have made ADAS, Automatic Driver Assistance Systems mandatory, such as: automatic emergency braking, alcohol level control, automatic lane maintenance control, automatic ignition of the 4 arrows in case of sudden braking. The belief is that eliminating the human factor that causes accidents, assisted driving and then autonomous driving can lead to a future with zero fatalities. Of course, only when automated vehicles will be circulating. That’s why in all the technologically advanced world autonomous driving tests are underway That’s why all the most advanced countries are implementing projects to digitize roads, transforming them into smart roads, thus providing the autonomous vehicle and the user with as much information as possible and the best environment in which to travel. That’s why technical experts from all over the world involved in the World Road Association (PIARC), are intensively exchanging technical knowledge on autonomous driving tests and smart roads implementation. Italy is also working in this direction and foresees by 2025, through provisions by the Ministry of Infrastructure and Transport, the progressive digitalization of the main roads (those belonging to the European network). And in this context ANAS, the Italian Authority for national roads, is operating with an ambitious project to transform 3.000km of its network into Smart Roads.
For several months, vehicles with autonomous driving capabilities have been circulating on the streets of Turin and Parma in the midst of ordinary traffic. They proceed on their own, without a driver, even if there is always an expert driver in the car ready to take control of the vehicle. So far no accidents have occurred and the tests are proceeding positively. These first tests of autonomous driving on the road, authorized by the Ministry of Infrastructure and Transport are conducted by VisLab, a company founded as a spin-off of the University of Parma and then acquired by the American Ambarella. VisLab was founded in the mid-1990s in Parma within the university field thanks to the intuition of Alberto Broggi, full professor at the Department of Information Engineering. As a child, his first favourite game was a calculator, then the home computer. He was among the first to write videogames for the “ZXSpectrum” and among these his favorite was a simulator of a Fiat Uno. Broggi conducts in 1998 the first experiment of autonomous driving with a production vehicle and low cost systems and which then remains as a milestone in the history of vehicular robotics: a vehicle crosses Italy in automatic mode following the route of the famous “Mille Miglia“; the test called “Mille Miglia in Automatico” (“One thousand miles in automation”) lasts a week during which the vehicle travels more than 2000 km covering 94% of the journey in automatic mode. In 2009 the VisLab university spinoff was born and the University invested 500 euros. VisLab begins by developing software for the artificial vision of the road environment. It immediately collects certificates of recognition for its innovative research. In 2010 Broggi becomes a pioneer in the field of automated driving, managing to bring his automatic vehicles from Parma to Shanghai.
In 2014 the VisLab group developed Deeva, a prototype equipped with sensors and video cameras able to perceive and interpret what is happening in the road environment in real time and able to decide independently in which direction to move and at what speed. The introduction of this technology can revolutionize the concept of rod transport and increase road safety, thus massively reducing the number of accidents. In 2015 the US company Ambarella decides to purchase VisLab but it is no longer worth 500 euros. Ambarella pays 30 million dollars for the acquisition. The team of engineers led by Broggi is asked to move to California, but the team prefers to stay in Italy. So the company starts by hiring 30 persons in Parma. Broggi will be the one who goes back and forth between Parma and California. And it will be Broggi to request at the Ministry of Infrastructure and Transport for the first authorization to carry out tests in Italy and to obtain it.
Alberto Broggi is also a member of PIARC Italy’s Technical Committee Automated and connected driving which is composed of managers and experts from the Ministry of Infrastructure and Transport, ANAS SpA, the main automotive and telecommunications sectors, motorway concessionaires and universities. In this conversation with Broggi he will give us an update on the perspectives of autonomous driving, the tests already implemented, its regulation, and the optimal road environment to host it.
During the year 2019, VisLab – a company of the Ambarella Group (listed on NASDAQ and based in the Silicon Valley) – obtained from the Ministry of Infrastructure and Transport the first authorizations for automated driving tests under the Smart Road Decree. How do you evaluate the authorization process carried out in Italy?
The authorization process has been complex and articulated, also to allow ministerial experts to make the appropriate assessments for the safety of users. Thanks to this authorization, the Ambarella Group has been able to continue its program of significant investments in Italy and VisLab has been able for more than twenty years to continue its activities of research and development.
And what has been the outcome of the autonomous driving experiments in Italy?
The experiments conducted in Italy by VisLab of automated driving on public roads concerned urban scenarios. VisLab has thus conducted numerous tests that concerned both the perception and the management of the route and implementation on the vehicle. The current ministerial authorisation allows VisLab to conduct trials in the cities of Parma and Turin. Watch the video. The result so far has been very positive.
What types of autonomous vehicles are you testing on the roads? Which brand?
Our technology applies to any vehicle brand or size. We are also agnostic about the type of propulsion, but I think the choice to use hybrid vehicles (fuel and electric) is a winning one as we combine extended automated range with the availability of electric power for when these vehicles are used to develop new features.
What technology do you implement on your automated vehicles?
VisLab perception devices mainly consist of cameras and radar. They are two complementary systems because they work on different wavelengths and therefore with different perceptive capabilities and limits, complementing each other and mitigating each other’s less effective areas.
For example, the cameras allow a greater density of information while the radar, with lower resolution compared to a camera, has the ability to penetrate even the most adverse weather conditions giving back useful information to the system.
The cameras used by VisLab are equipped with dual sensors and therefore allow a stereoscopic vision. While single cameras (i.e., equipped with a single sensor) allow to recognize objects. Stereo cameras also have the ability to return information to the system in relation to the position in 3D, the distance and the spatial relationship between objects.
Unlike most operators and competitors, the system designed by VisLab therefore does not use laser scanners (i.e., LiDAR), and the Italian tests carried out by VisLab have confirmed the excellence of the technological path taken.
An additional difference between the technology used in the Italian trials compared to the one developed by competitors is that the perception tools used by VisLab (i.e., mainly the cameras designed and built within the Ambarella Group) are equipped with internal processing engines for deep learning, which is therefore carried out “on the edge” (inside the camera itself) and with low power consumption.
Moreover, the technology developed by VisLab may also not use information from a GPS, as the localization can also take place with the sole use of visual information from the cameras.
How are you enhancing the results obtained with the testing?
The developments achieved and the results obtained during the trials carried out in Italy were then consolidated and further developed in the United States, in different traffic conditions, infrastructure, and type of roads.
The autonomous capabilities of VisLab/Ambarella vehicles were successfully demonstrated in Las Vegas, Nevada (CES 2020). Watch the video. The experience gained in the United States (Nevada and California) will allow to further extend the areas of the next testing phases in Italy.
To travel in complete safety does the automated vehicle require to be connected to a smart road or is any type of road sufficient?
Our philosophy is that technology must have the ability to drive the vehicle in reasonable safety on any road; on the contrary, designing an automated driving system that needs information from third parties (such as an infrastructure network) for its operation I believe can be limiting. That said, if further information is available (for the benefit of both safety and comfort) via a smart road, all the better. It should also be considered that the collateral risk of connecting an automatic vehicle to a smart road network is the increased possibility of potential intrusion or manipulation in the automatic driving system.
Some international experts say that the autonomous vehicle may find it difficult to determine driving in the presence of works in progress or other obstacles and that the limited vision of the sensors on board may fail to see them in time. The Situation Rooms of ANAS and other motorway concessionaires presently have this “long vision” and the digitalised smart roads will have this vision even more. Would it not be possible to “connect” the vision of the Situation Rooms with the autonomous vehicle so that one can benefit from the other?
Overall, the problem of recognition of works in progress and other unexpected situations is quite common. I believe that the autonomous vehicle will one day have to be able to interpret these situations independently just as a human driver does. In any case, any additional information coming from the communication system will give the possibility to further optimize the driving (for example to modulate the speed before the visibility of the event or even to change the route by choosing an alternative itinerary).
PIARC Task Force dedicated to autonomous driving highlights the need for harmonisation of continental road signs and also the need for technical measures to promote the visibility of signs in different weather conditions. Do you agree? Should the signage useful for the autonomous vehicle be physical or digital? What is the optimal solution for a road operator who wants to avoid spending unnecessary money and prepare for the future?
With regard to signage, as long as we do not only have autonomous vehicles running on our roads, we will have to keep signage that can be interpreted by humans drivers. I therefore think that it must be the vehicle that needs to adapt to the signs in a particular geographical area. Obviously, as in the case of the aforementioned communications, if the signage were active, emitting radio signals such as a danger signal or a traffic light, this information would allow further optimisation.
What do you ask of road operators to encourage independent driving? That the road surface is well maintained? That the horizontal and vertical signs are visible? What else?
The main request to road operators is to continue to check and maintain precise horizontal and vertical road markings. It should also be pointed out that any change to the signage could cause inconvenience to those who use high-resolution digital maps, as they are required to update them.
What is the ‘guaranteed minimum’ that the motorway operator should provide in order to ensure itself also in terms of legal liability in the event of an accident?
With regard to the “minimum guaranteed” that the motorway operator should provide (to ensure itself also in terms of legal liability in the event of an accident), based on the assumption that the autonomous vehicle is designed to circulate on the current road network without other means, I would say nothing more than what it “owes” to users today. Any additional information that the operator will want to provide to the vehicle through the smart road network can only lead to additional responsibilities (e.g. if the information is wrong).
Several car manufacturers consider the connection with the road infrastructure as certainly very useful but “additional”, like an “additional” sensor that could even fail (the connection may not be guaranteed 100% of the time), and for this reason they include fallback solutions based only on optical sensors or HD maps. Could we therefore imagine that the motorway operator, while being incentivized to offer the vehicle the most useful information through the smart road wherever possible, can keep its responsibility only with respect to “ordinary” maintenance (horizontal and vertical signage and road surface)?
This is probably not only a technical-engineering issue, but also and above all a legal one. The principle, I believe, is that the operator is responsible for every information and service that he offers or manages, so, the more services and information that he makes available to users, the greater will be his responsibilities.
Having said that, the basic principles remain, that the maintenance by the operator is very important today for the human driver, and it will also be important tomorrow with electronic drivers. In fact, the vehicle is designed to move in an environment with precise characteristics, both in terms of geometry and signals. Any additional information can still be used, but will involve a responsibility on behalf of the operator to ensure its accuracy.
Is it true that the autonomous vehicle has a visibility of about 100 meters ahead while with the smart road visibility can increase by several hundred meters?
It is true that the smart road can communicate to the vehicle information that is not directly visible, such as the presence of obstacles behind a curve, or traffic information on its route. When provided by the road operator, this information is certainly helpful for driving but the vehicle must be able to move safely even on roads that do not have the ability to offer this information.
What timescale can be envisaged for the marketing of the various levels of autonomous driving 4 and 5 from an automotive point of view?
Forecasts are always difficult, but I don’t think we’re very far off, considering that even a decade or so is a limited time compared to the more than 100-year-old life of the ‘vehicle’ transport system. We are dealing with a real revolution that this new technology will bring not only to the transport system, but also to the urban planning and accessibility of our cities.
A European Platform is working for the European regulation of autonomous driving. In terms of Regulation, is there a country to take as an example? Compared to other countries in the world, what progress has Italy made in terms of regulation and tests on autonomous driving? In this context which countries are to be monitored?
In 2018, Italy adopted a regulation that sets rather stringent requirements for those who intend to conduct tests. Our group is authorized to conduct tests in Nevada, California and Italy. The procedures for obtaining authorizations, as well as the requirements during testing, are considerably more complex in our country and can be a barrier for those interested in testing in Italy.
Based on the number of authorized experimenters, I believe that the world leader for testing is California, where there is a principle of accountability of the experimenter that summarizes everything, and it does not matter whether the applicant is a vehicle manufacturer or system developer.
* Manager Anas SpA, First Delegate and President Technical Committee Autonomous and Connected Driving PIARC Italy