TU-TESL

• 22

According to foreign media reports, although drivers and pedestrians should abide by traffic rules, this is often not the case. If the road is clear, pedestrians may run the red light, and drivers may lose patience in slow driving, so that they can cut into the next lane as soon as they see the gap. (photo source: Phantasma labs official website) although today's cars have made great progress in using sensors and cameras to avoid traffic accidents, there is still a long way to go to accurately predict what the unpredictable human may do. Some experts say that such technical limitations will slow down the widespread application of autopilot in city and city environments. However, some start-ups are now trying to solve this problem by establishing data sets of human behavior, which will help autopilot systems to predict and understand human behavior more accurately. Phantasma labs, a start-up based in Berlin, Germany, is such a company. At present, the company is building human behavior datasets based on mathematical rules and using virtual simulation, and is negotiating with China's automatic driving car manufacturers to adopt the technology. Phantasma labs was founded in 2018 by Maria Meier, a software engineer, and Ramakrishna nanjundaiah, a simulation expert, after completing a three-month project in Berlin organized by enterprise first (EF), a talent investment agency. Last year, Phantasma labs raised 1 million pounds (8.6986 million yuan) in an ef led round of financing and is now negotiating with investors to raise more money. The company mainly provides simulation services for autopilot cars, and constructs a virtual world with different cultural backgrounds. The company has built a platform to select real humans as city participants to interact in real time, and then human behaviors are captured and annotated composite data sets are automatically generated. Now, Phantasma Labs has been selling its dataset to autopilot to earn revenue. Its main competitors include perceptive automata, an American start-up that uses behavioral science and technology to describe how human drivers understand others' thinking States, and then trains deep learning models to achieve the same capabilities. In addition, the Japanese auto giant also launched the guardian automatic safety system this year, which will take over the driver's work when there is a life-threatening situation on the road.