September 08, 2015      

Toyota Motor Corp. has promised to spend $50 million over the next five years on research into robotics and artificial intelligence. Both the Massachusetts Institute of Technology and Stanford University will work with the Japanese automaker, which plans to compete with Google by taking a different approach to vehicle automation.

“Our long-term goal is to make a car that is never responsible for a crash,” said Gill Pratt, who will lead this research and who was previously the program manager for the Defense Advanced Research Projects Agency’s Robotics Challenge. Videos of robots failing to conduct tasks in this year’s DARPA Robotics Challenge went viral, and the military-sponsored contest has been discontinued.

Unlike the research into self-driving vehicles by Google, Apple, and others without prior automotive experience, Toyota wants to develop driver-assist technologies to make driving safer rather than to take it over entirely, he said.

“One million people worldwide die in car accidents every year,” Pratt said as part of Toyota’s announcement. “Our goal is to eliminate highway collisions without eliminating the fun of driving.”

Pratt will be working with Daniela Rus, head of the Computer Science and Artificial Intelligence Laboratory (CSAIL) at MIT near Boston, and Fei-Fei Li, director of the Stanford Artificial Intelligence Laboratory (SAIL) in Silicon Valley. Toyota’s support of $5 million per year to each of the two labs should lead to “accelerate these research activities and [their] application to intelligent vehicles and robotics,” he told IEEE Spectrum.

Consumers are more likely to accept semi-autonomous vehicles than fully self-driving ones. Toyota is also being careful to spread the risk of its research between two universities, rather than co-opting a research group, as Uber did with Carnegie Mellon University faculty.

Toyota’s bicoastal strategy is an expansion of existing automotive capabilities, such as proximity warnings, rather than an effort to develop a fully autonomous vehicle’s hardware and software from scratch. While this may initially seem less ambitious, the AI required to distinguish and aid different environments and driver behaviors is still very challenging.

“We hope to develop some of the new algorithms and systems and technologies that will take autonomous driving and robotic mobility to the next level, so that robots can deal with way more complex scenes than what they can deal with today,” said CSAIL’s Rus.

Stanford’s complementary research will focus on machine perception and learning, reasoning, and communication. Robots can increasingly sense and recognize objects, but existing software isn’t yet up to the task of understanding what they are in context and working with a human in a dynamic environment, said SAIL’s Li.

“Building on Stanford’s expertise with computer vision, machine learning, large-scale data analysis, and human-computer interaction, our team will work to help intelligent vehicles recognize objects in the road, predict behaviors of things and people, and make safe and smart driving decisions under diverse conditions,” she said.

“In parallel autonomy, there is a guardian angel or driver’s education teacher,” Pratt said. “It usually does nothing, unless you are about to do something dumb.” He added that the systems being developed should actually help people be better drivers.

Race to robotic cars

Toyota was one of the first carmakers to offer automated parking back in 2003. In addition, General Motors, Nissan, Telsa, and Volkswagen are working on hands-free systems, according to Reuters. Other automakers developing autonomous vehicles include Audi, BMW, Mercedes-Benz, Ford, and Volvo.

One impetus for expanding driver-assist technology is to enable older motorists in Japan and the West to “age in place,” so that they can retain their mobility.

What if cars could become our trusted partners?” said Rus. The combination of MIT’s research into environmental awareness and Stanford’s research on AI and decision making could result in safer vehicles, but integrating them will be difficult.

“Solving these challenges will require combining our knowledge of data-driven and model-based approaches to decision making and perception,” she said. “Developing a vehicle that’s incapable of having an accident is an ambitious goal, but at CSAIL, we’ve always focused on the moonshots.”

Toyota is supporting its research with sales of midterm stocks and an investment fund with Sumitomo Mitsui Banking and Sparx Group. Toyota has already developed robots for personal care and healthcare assistance.