January 19, 2018 – Deep Learning in country-specific, real road traffic plays a central role on the road to autonomous driving. This is shown by Mercedes-Benz Intelligent World Drive, which ended at the Consumer Electronics Show (CES) in Las Vegas after five months.
A test vehicle on the basis of the current S Class completed a challenging study trip on five continents in order to "learn" in automated test drives in real traffic. From zebra crossings on Chinese motorways, turning off right from the left-hand lane in Melbourne, Australia, pedestrian traffic on all kinds of roads in South Africa or a temporary driving ban in the immediate vicinity of stopping school buses in the USA – on every continent the S Class faced challenges which will have an influence on the driving characteristics of future autonomous vehicles. Automated and autonomous vehicles have to know about these country-specific particularities and understand them in their respective context in order to be able to make the correct driving decisions.
The Intelligent World Drive also underlines just how important the international harmonisation of the legal framework for automated and autonomous driving and its infrastructure is, in particular of lane markings and traffic signs.
Intelligent World Drive provides an insight into the complexity of global challenges
With the test vehicle on the basis of a semi-automated S-Class test drives were carried out in Germany, China, Australia, South Africa and the USA. The differences in the countries give a small insight into the complexity of global challenges in the development of automated and autonomous driving functions. In particular the national particularities in terms of infrastructure, traffic regulations and the conduct of other road users place very different requirements on the sensors and algorithms of the vehicle. It also becomes apparent just how important high-resolution maps could become for the development of higher automation. Daimler AG is thus involved in the map service HERE and is working on faster implementation and updating of even more precise navigation data.
Mercedes-Benz started the 'Intelligent World Drive' at the Frankfurt International Motor Show (IAA) to adapt more highly automated driving functions to national user and traffic practices. The main area of interest in Germany is specific driving behaviours on motorways and in traffic jams.
However, the technical side of the Intelligent World Drive - that is to say the systematic compilation of situations, experiences and data - is just one aspect. The social component is equally important: This involves getting into contact with people all over the world, soliciting feedback and opinions, and stimulating discussions about the personal mobility of the future.
The high density of cars, two-wheelers, three-wheelers and pedestrians and the associated traffic behaviour in Chinese cities pose different requirements on automated driving functions than in Europe or the USA. That is why China requires foreign motorists to have a Chinese driving licence. Automated and autonomous vehicles also have to prove their fitness for this market in advance.
In addition, there are road signs with Chinese characters and lane markings, which in China have different or even multiple meanings. For example, short white lines, known around the world as pedestrian crossings, can also be found on motorways. However, they don't denote a pedestrian crossing, but the minimum distance between vehicles. The sensors must be able to recognise this and interpret it correctly. The same is true for speed limits, which can differ from one lane to another. Another challenge: Parking spaces come in many different shapes and frequently are full of obstacles that are hard to detect for sensors.
These special national features show how important it is to gather worldwide insights into real-life traffic on the road to autonomous driving and to adapt automated driving functions to the particular traffic practices and conditions. In the past seven years, Mercedes-Benz conducted about 5100 test drives around the world with 175 test mules for validations of driver assistance systems in the field alone. The performance of the driver assistance systems was assessed on some 9.5 million kilometers in Europe, the USA, China, Australia and South Africa, and more than 1.2 million measurements were made especially in real-life traffic situations for their continuous advancement.
In China the focus of the test drive is on driving behaviour in the dense traffic of Shanghai with its millions of inhabitants.
Turning off to the right from the left-hand lane, flashing speed limit traffic signs, kangaroos jumping over the road: road traffic in Australia presents some very special challenges. Automated and autonomous vehicles have to be aware of these peculiarities. In the third leg of the Mercedes-Benz Intelligent World Drive, the test vehicle – based on the S-Class series-production saloon - is facing up to Australia's unique conditions and situations with automated test drives on highways, motorways and in the city of Melbourne.
Instead of the classic speed limit signs made of metal, Australia's road traffic system is making increasing use of electronic displays with variable speed limits which are designed to improve both traffic flow and safety. They are fitted with a bright white LED display, a red LED ring and a yellow LED warning lamp and are able to display not only speed limits but also simple symbols and letters. In some cases they are positioned one after another, to ensure dynamic traffic flow, and can change what they display within fractions of a second. These signs present a major challenge for the performance of the digital camera and the quality of the digital mapping data, which enable automated driving functions such as Active Speed Limit Assist and Active Distance Assist DISTRONIC to function reliably.
Road traffic in South Africa presents some very special challenges: different road surfaces, wildlife on rural roads and many pedestrians in the city as well as in the interurban traffic who often cross lanes completely unexpectedly. Automated and autonomous vehicles have to be aware of these peculiarities and respond in a reliable manner. In the fourth leg of the Mercedes-Benz Intelligent World Drive, the test vehicle - based on the current S-Class series-production saloon - is facing up to South Africa's idiosyncrasies with automated test drives on the roads of the Western Cape and in the city of Cape Town.
Whether in the city or out in the country – in South Africa, there are many pedestrians on the road. Sometimes they walk on the street and often, they cross lanes completely unexpectedly. In the extremely dense urban traffic in Cape Town, driving is truly a precision task – particularly in narrow streets, where the pavements are mostly overflowing with parked cars on both sides. But even on national roads outside of towns, and on the motorway too, drivers always have to expect to encounter crossing pedestrians. Just as high in number are the accidents which occur. In 2016 some 5,410 pedestrians died in road traffic accidents. That figure represents 38 percent of all road traffic deaths. This pedestrian behaviour calls for an additional, increased level of awareness and thus, also poses particular challenges for the sensor systems of automated and autonomous vehicles. Cameras and radar systems have to detect passers-by and interpret their movement correctly so that the vehicle can react within milliseconds in the event of an emergency.
Further special features include traffic signs which are only found in the 15 Member States of the Southern African Development Community, such as South Africa, Namibia, Botswana or the Seychelles. For example, the no stopping sign shows a crossed-out letter 'S' in a red circle, while the sign for no entry is made up of two black horizontal bars in a red circle. In addition, the road traffic signs in South Africa are often incomplete. Intersections where you have to stop are not always indicated by a stop sign – in some cases they only have wide, white lines across the road surface. Warning signs before the commonly-found speed bumps are also not always present, or are positioned close to the obstacle that there is insufficient time to react.
The lack of signs presents a major challenge for the performance of the camera and radar systems as well as the quality of the digital maps, which enable automated driving functions such as the Active Distance Assist DISTRONIC with route-based speed adjustment to function reliably. Validating the latest digital map material from HERE, particularly with regard to intersections where the vehicle would need to stop and traffic obstructions such as speed bumps, is therefore a particular focus of the test drives on the Western Cape.
When the school bus stops, all other vehicles around have to stop too. Speed limit signs exist that are not found anywhere else in the world. There are dedicated lanes for carpools, laws permitting traffic overtaking on the right, and lane markings consisting of raised “dots” instead of painted lines. Road traffic in the USA makes many specific demands on the sensor systems and algorithms of vehicles equipped with automated and autonomous technology. The aim of the Intelligent World Drive on five continents was to gain worldwide insight into real traffic conditions so that future, more automated driving systems can be adapted to country-specific traffic and user habits.
The automated test drives in the greater Los Angeles area, and subsequently to the Consumer Electronics Show (CES) in Las Vegas, concentrated on the assessment of driving behaviour in dense city traffic and on highways. The focus was particularly on the recognition of school buses, lane markings, and speed limit signs.
An equally demanding task for camera and radar systems is identifying separate carpool lanes, which are reserved exclusively for vehicles in which at least two people are sharing a ride. These so-called HOV lanes (high-occupancy vehicle lanes) can be found on multi-lane interstates and freeways in urban areas. For the sensors and algorithms of automated and autonomous driving technology, it is difficult to recognise these as special lanes and distinguish them from normal or exit lanes. Moreover, the lanes are not always in the same position, and can be on the left, right or centre of the highway. HOV lanes can also be separated from the other lanes by two uninterrupted yellow lines, or by metal guard rails, or they may be identified by painted diamonds. In the future, autonomous driving will also require that the system knows the number of occupants in the vehicle to ascertain whether they are permitted to use the carpool lane. This is because in metropolitan areas, there are also HOV lanes that are only permitted for carpools of three or more (e.g. Los Angeles) or even four or more persons (e.g. New York).