Technologies for fully autonomous driving get into gear

News that a self-driving car saved its owner’s life this summer delivered a boost to the autonomous motoring revolution that is quietly taking over decision-making tasks from the human behind the wheel. Many new-car owners are already comfortable with self-driving modes like park assist and emergency braking and are ready to accept extra help to deal with the pressures of driving on today’s roads.

Increasing levels of on-board intelligence allow autonomous driving to advance from occasional use to handling an increasing proportion of the driving from the beginning to end of each journey. Joshua Neally, who suffered a pulmonary embolism during a journey and was saved by his self-driving Tesla Model X, only had to steer the last few metres of the 20-mile journey to hospital, including navigating the complexities of the hospital car park.

Continous progress in precision mapping and V2X communications could soon enable machines to handle even those tasks that have historically been seen as too complex to automate.

Among major projects that are currently active, three involving leading automotive brands seek to combine digital mapping with visual data gathered from multiple vehicles in real-time, such as information on dividing lines, kerbs and other road characteristics. Ultimately this should enable driverless vehicles to handle complex road layouts and respond correctly to changing conditions, without human intervention.

Researchers at Toyota’s Central R&D lab in Tokyo for example have developed a system that gathers road images and position information from cameras and GPS devices on-board large numbers of production vehicles. The information is automatically pieced together, and corrected and updated in real-time in the Cloud to generate live highly accurate and detailed digital road maps. Toshiba is also aiming to map large areas using a similar approach, and is developing advanced image-matching technologies to ensure accuracy to within better than 5cm on straight roads. The system is expected to be introduced on production vehicles by 2020.

Mobileye is also collecting information “from crowd to Cloud”, to develop Road Experience Management (REM) software that runs on its EyeQ processing platforms. REM running on individual vehicles extracts landmarks and roadway information which is transmitted to the Cloud at a very low data rate of approximately 10kBytes per kilometre. Artificial intelligence algorithms running in the Cloud integrate the segments of data received from all vehicles to maintain a live, global map. Mobileye is working with General Motors to integrate REM into forthcoming new models, with a view to equipping driverless vehicles in the next few years. Other interested brands include Volkswagen, which is currently integrating REM into its fleet, and Mobileye announced in 2016 that Nissan will also use its technology.

BMW, Audi and Daimler are taking a different commercial approach by setting up a joint venture that has acquired mapping technology from Nokia. This is providing the foundation for the Cloud-based HD Live Map system which uses tiles and contains several layers of dynamic content. The layers combine static map content as well as temporary information from updates and analytics. Each source provides different details such as lane information, dynamic road-network and situation changes, and speed-profile data. This approach allows any new events to be layered onto the map in small files, without needing to update the entire image. Ten car-makers are already testing HD Live Map in projects underway in the US, France, Germany and Japan.

Critical technology

Connected car technology will be critical for gathering the data needed to keep live maps current to help improve road safety and traffic flow. The IEEE 1609 wireless-access standards and the IEEE 802.11p WLAN amendment to support information exchange in intelligent transportation systems have been in place for some time, but connected vehicles have yet to hit the streets.

Audi has demonstrated the speed and reliability of V2V communications emphatically in January 2015, exchanging information between vehicles at relative speeds up to 500km/h and distances of up to two kilometres. And at this year’s Geneva Motor Show, Harman teased audiences with V2X in a tech-laden BMW i8. The waiting may soon be over, as both Cadillac and Mercedes Benz have announced plans to introduce V2V communication on production models in the 2017 timeframe.

As MIT Technology Review commented, those initial Cadillac and Mercedes models could feel pretty lonely on the roads as they try to find friends and make connections. It could be a decade before a large number of V2V-equipped vehicles are in common use. There is also the small matter of the infrastructure upgrades needed to support V2X. Well, everything has to start somewhere. Once that start is made, one or two safety success stories – in the vein of Joshua Neally’s experience - could provide a catalyst for rapid adoption.