Part II
Updating the maps via the cloud
It looks also TomTom. Together with the automotive supplier Bosch, Dutch company wants to measure exactly the German motorways until the end of 2015. The cards according to Bosch in the future should consist of several layers: for navigation, localization, and planning. The calculation of the route from A to B with the roads to be ridden, achievements through the navigation layer Bosch explains. Based on the isolation layer, the vehicle detect its exact position in the lane. Furthermore, information about objects that would win over the environment sensors would comparing with the corresponding data in the layer. The planning layer in turn contained data about road markings, road signs and speed limits and road geometry. So choose an automated vehicle, for example, when it changed the track.
TomTom also plans an accuracy in the decimetre range. With such cards the Bosch test vehicles are highly automated in public roads since early 2013, a Bosch spokesman at the request of Golem.de said. Apart from the latest such accuracy is sufficient. GPS signals would only used to locate a vehicle after the start of the system and before an automated drive roughly. This is sufficient for a precision in the range. The GPS signals were later used to plausibility the precise card data.
Road users upload their own sensor data
But what good is the most accurate material, and what often occurs, there is a new guidance when suddenly a construction site blocking the way? 'Information about traffic jams, accidents, hiking site, or speed limits must be the highly automated vehicle currently and timely available. Only then it can choose the best driving strategy ahead", Bosch writes. To detect such short-term changes, TomTom will use not only its own fleet of mapping, but plans according to own statements together with Bosch "use another vehicle fleets with corresponding sensors on board". What fleets should be, is currently still.
A different approach, however, Nokia here with its "living streets" (live roads). The company has developed an interface specification, which defines how sensor-based data of different vehicles in a cloud can be processed. This specification was in June 2015 the automotive industry via a CC license free of charge. "The generated data can be summarized in a standardized form in the cloud, processed and analyzed to design a detailed real time picture of road and traffic", Said here. On the basis of the information collected card information could also match and improve. Moreover, warnings about damage to roads or building sites could be given.
Each light must be accurately recorded
However, the resulting volume of data may be very large. According to a forecast of by market research company SBD, 33 million crosslinkable vehicles will be sold up to the year 2020 per year. This generated more than 163 million terabytes per year of data. Here trust that the use of the vehicles the network technology will have evolved, a spokesman on request. Then, for example, the mobile phone standard 5 G could be ready for the market. Google is betting its autonomous cars that they sign non-mapped traffic signs to the Central.
The requirements for the cards are still significantly higher, if automated cars want to leave the highway and driving around in the city. Since it is required that every zebra crossing and each traffic light is captured and recognized. The cars need to know exactly where and to what extent are the relevant traffic lights. While Google shows with its test vehicles in the United States, that even urban transport is in principle automatically to dominate. However are the approximately 6.6 million kilometers of road in the United States until a year ago, only a few thousand Google mapped been. The entire mapping of the United States was still not "fearsome task", said the head of Google-car, Chris Urmson.
"Cities must change"
In this context, it is not excluded that a manufacturers such as Audi in the United States relies on Google maps. The next generation of the Audi A8 to 2017 with the first highly automated functions like a traffic jam drivers be equipped. "We can now say used navigation data in two to three years in our vehicles for the U.S. market", a spokesman said. At the present time could not even be said, "when we get high resolution maps for the first time in connection with highly automated functions to use". That is to say: may the A8 in the United States first only on the basis of the sensor data drives. In case of a traffic jam pilots that would be but not so crucial, because the car then easily with could swim in the traffic.
Given the problems yet to be solved, German manufacturers assume that in 15 years vehicles in all traffic situations autonomously on the road can be complete. Currently, they are still the view alone with better cards not to achieve this goal. "In an urban environment we expect autonomous cars until the year 2030. The cities must change", BMW development head, said Huber. Traffic lights could be then able to communicate with the vehicles. Google, however, has continued to insist that this step takes place much earlier. Currently eleven year old son should do more no driving test at the age of 16, Urmson said. Google should really achieve this goal, the German competition would have first bad cards.
In a four-part series of articles, Golem.de represents the current state of development of autonomous vehicles. In other articles, let's examine whether and in what way autonomous vehicles can be made more secure by wireless connection to the Internet or other vehicles. Recent hacker attacks have shown that the networked cars can be also a great security risk. In addition, we want to illuminate what political steps on the way to the autonomous driving are still required, unless by legal regulations and standards or help with the infrastructure and test tracks.