The Record - Issue 19: Winter 2020

64 www. t e c h n o l o g y r e c o r d . c om BY A L E X SM I TH The long-held goal of autonomous mobility is coming into sight for the automotive industry, but will require a dramatic shift in the way companies think about vehicle design F E ATUR E O nce the preserve of science fiction, momentum is building behind the development of driverless cars. Semi- automated aids that can take control out of the hands of the driver are already hitting the mar- ket, with development on more advanced sys- tems continuing to gather pace. The autonomy of a vehicle is measured accord- ing to six levels, a scale devised by the Society of Automotive Engineers. These levels range from fully manual driving at Level 0 to fully autono- mous at Level 5. At this higher level of automa- tion, there will be no need for a human to pay any attention to the vehicle, which will be able to go anywhere and do anything that an experi- enced human driver would be capable of. Fully autonomous vehicles that reach this level are currently under- going testing, but most systems currently available in cars are at Level 2, able to control steering and acceleration while a human monitors the environment and is able to take control at any time. To reach those higher levels of automation, ensuring the safety of the passengers who rely on autonomous systems is of paramount impor- tance. With part of the promise of autonomous driving being the removal of human error, every accident is analysed intensely. “Safety is key, because this is a hard problem to solve,” said Mitra Sinha, principal programman- ager of Autonomous Driving Microsoft Azure Cloud at Microsoft. “Humans have so many accidents that we almost take it for granted, but every safety incident with an autonomous vehicle becomes news. And while we see some regional efforts at regulation, there is no worldwide stand- ard. This means that there will be incidents, and that is why we see the emphasis on validation, with both road testing and simulation.” Validation involves fleets of test vehicles driving hundreds of thousands of miles, either virtually or on real roads, to acquire the data necessary to ensure that a vehicle can react appropriately in any given situation. A wide array of sensors must be attached to the test vehicles, all of which produce massive amounts of data. Automakers must find ways to ingest this data, then find the specific datasets that are of interest. Helping to meet these massive data challenges is where Microsoft sees its role within autonomous development. Developing the driverless future