UST 031

26 Dossier | EasyMile EZ10 autonomous shuttle when it should. False positives, he says, are much safer than false negatives and are definitely preferable as the industry evolves. Enabling the vehicle to adapt to the vagaries of life out on the road also took a great deal of time and effort, Pairot says. For example, when snow ploughs push snow to the edge of the roads they often effectively narrow the road by a metre on each side, and the vehicle reacts to the remaining snow as if it were an obstacle and stops. “A human would see this and adjust accordingly, but it is a technical challenge for an autonomous vehicle,” Pairot says. “The more we deploy, the more rules we find that are human-centric and that are embedded in all our cities – speed bumps, plants in containers, you name it.” While dealing with such issues through the software is possible, if time- consuming, another approach is to negotiate the problems away by talking to the authorities responsible for the roads. In Sweden for example, the snow problem was solved by persuading the local council to clear a further 50 cm on either side of the road. In Switzerland, a large container of plants placed in a narrow street as a traffic- calming measure to force drivers to agree between themselves who would yield was removed to allow the EZ10 to operate – no autonomous vehicle can interpret or make the often vague informal signals that human drivers use in these situations. It might seem odd to regard speed management as a challenge for such an avowedly slow vehicle, but choosing the right speed is an important part of minimising the number of unnecessary slowdowns and stops in city traffic. At the time of our visit, in late February, this was a core focus of EasyMile’s teams and was approached by working to build a better understanding of the vehicle’s operating environment, according to Pairot. Eventually, the company wants all EZ10 operations cleared to run without a safety operator aboard, and is currently working to prove that they can safely be withdrawn in increasingly complex types of environment. The first experiments on open roads without a safety operator are due to take place later this year. Level 4 autonomy goal The aim in terms of autonomy is Level 4, as defined by the Society of Automotive Engineers. The software can drive the vehicle under limited conditions and won’t operate unless all of those conditions are met; unlike Level 3, Level 4 software will never ask an operator to take over – the EZ10 will stop instead. Level 5 removes all the limiting conditions, defining a system that can drive the car everywhere in all circumstances. “The autonomous vehicle industry has been dominated by hype, particularly in the media, feeding the public’s imagination and excitement with predictions of fully autonomous cars driving around on public highways from as early as 2018,” Pairot says. “No-one is at Level 5 yet though. It’s the holy grail of autonomous driving but it could be 50 years away.” Recently, however, things have calmed down and expectations have shifted away from imminent, big changes in driverless technology, he adds. “Level 4 is where it’s at.” These days the EZ10 operates only in pre-mapped environments, and its capability sits somewhere between Levels 3 and 4. Mapping the environment April/May 2020 | Unmanned Systems Technology Set at a convenient height for wheelchair users, the external ramp button brings the electrically powered ramp out from under the floor of the vehicle, and the suspension on the ramp side is lowered to ease access The more we deploy, the more rules we find that are embedded in all our cities – speed bumps, plants in containers...