Finding the human role in self-driving car technology
At AutoSens 2018 in Brussels, high-ranking experts discuss the future of driverless car technology – and the great unknown factor: the human.
While making their way to the future of mobility, visitors passed by the golden era of mobility: vintage cars like a Benz from 1915 or the historic Ford Model T. It’s all part of Autoworld, a Brussels museum that was home to AutoSens 2018, one of the world’s leading technical summits for advanced driver assistance systems and autonomous vehicle perception technology. Set in a rather impressive location, international industry and academic experts discussed what lies ahead for self-driving car technology – and what part humans will play in it.
This not only covers driverless vehicle passengers, but also their developers. While artificial intelligence seems to be all the rave, it still needs guidance. Henning Lategahn, CEO of 3D mapping company atlatec, explained in his speech that his company relies on “human guided AI”. The raw mapping data still needs human annotations before artificial intelligence can create the 3D map. “Especially when the lane marking quality is poor, the degree of human guidance increases,” Lategahn said.
Anantha Kancherla, Head of Self Driving Software at ride-hailing service Lyft, detailed another specific human problem: shortage of talents. According to Kancherla, Lyft heavily draws on deep learning technologies to develop their algorithms. The big challenge connected to that is finding the right mix of scientists who know the theoretical models and software engineers who can implement them. “You need both mindsets,” said Kancherla, admitting that it is not always easy to align those two backgrounds. “Alternatively, you need to find those unicorns who can do both.” However, they are a rare species, and thus, an expensive one.
Build your own self-driving car
Thankfully, the conference also offered the opportunity for industry leaders to meet some aspiring newcomers. They just can’t be hired yet – because they are still in high school. A group of students from Portland, Oregon, excited the audience with the self-driving car they developed entirely on their own. As part of the non-profit High School Autonomous Vehicle Project, they used an RC car, a camera sensor and an Intel processor to build and program a self-driving vehicle.
Their first major challenge was one that every company knows: organization. “We had no defined structure,” says Vikram, one of the team members. “We had to establish rules and organize ourselves.” So, they formed groups: one was in charge of the steering of the vehicle, one of the image processing and one of acceleration. They also came across two major roadblocks that sounded familiar to many of the conference visitors from the industry. “Just because we had successfully tested a scenario on our simulator, that didn’t guarantee that it also worked on the road,” remembers Vikram’s colleague Matthew. Furthermore, they soon had to find out that processing power of their car was not endless – and come up with a solution to the technological limitations.
At AutoSens, the car was already making its rounds across the carpet. But it won’t stop there for the students. “Thousands of students don’t get the opportunity we had,” they explain in a video about the project. This is why they want to develop a low-cost, customizable self-driving car kit to enable students around the world to make similar experiences.
People are people, not just users
One of the most insightful panel discussions of the conference also dealt with that crucial relationship of humans and machines. “How can we get customers attracted to enjoy cruising with robots?”, it asked the participants. “People are more than users. We should approach them as multi-faceted human beings,” answered Kubra Zehra Kasikci, an environmental psychologist who researches human-centric design in automated vehicles at the University of Surrey. “Humans will have a dynamic relationship with autonomous vehicles that depends on the mood, the ability and the cultural background of the person.” She urged the industry to broaden their research focus. “At the moment, they ask how we can manipulate consumers into using driverless cars. Instead, we should determine why people would actually want to use those vehicles.”
Her discussion partner Tobias Schneiderbauer, Engagement Manager at McKinsey, noted that the main reasons that prevent consumers from switching to autonomous vehicles vary with regards to the cultural background. In a recent survey, McKinsey found that a big concern among German consumers was the loss of driving for pleasure. In the USA, fear of machine malfunctions was a much bigger issue. According to Schneiderbauer, the two fatal accidents by Tesla and Uber vehicle have made U.S. consumers warier of safety aspects.
Driverless cars: gentrifying mobility?
Schneiderbauer also cited increased convenience and a decrease in costs as possible incentives for consumers to switch to autonomous vehicles. When it came to the question of costs, the discussion took an interesting turn. A conference participant expressed his fear that autonomous vehicles could lead to the “gentrification of mobility” – a dystopian scenario where only wealthy individuals can afford the convenience and safety of self-driving vehicles while lower income classes will have to stick with conventional cars and transport systems. Kasikci noted that inequality of mobility was already an issue today, “depending on where you live.” In her view, the political sphere will play a central role in ensuring that the necessary infrastructure will also be available to countries that are less developed and rural areas.
When all was said and done, and all experts and geeks passed all those vintage cars to leave, one major finding seemed evident: despite all technological progress, the decisive factor will still be the human factor.