Our research is aimed at eliminating deaths caused by car accidents, for which artificial intelligence is expected to play the key role.
Our relationship with cars has reached a turning point thanks to new technologies, including AI-powered deep machine learning technology and edge computing distributed processing technology. The dream of collision-free cars is about to become a reality, so we interviewed researchers in charge of just such a project at Toyota Motor Corporation and NTT Network Innovation Laboratories.
The Birth of Intelligent Cars
(Photo) Yasuyuki Tamane, Executive Manager, Connected Management Department, Connected Car Company, Toyota Motor Company
Q: At the NTT R&D Forum in February 2016, I saw a demo of a collision-free car in which multiple Prius scale models moving about freely avoided colliding with each other. Apparently the same demo was conducted in the U.S. in January 2016. Can you tell us about your organization, Connect Car Company?
A: The organization was launched in April 2016 and has roughly 500 members. Our mission boils down to making a better car, but instead of the manufacturer deciding what is meant by a ’better car,’ we think that this should be defined by customers.
Given the technological changes occurring today, including smartphones and other devices connecting with cars, we decided to think about cars from the aspect of connectivity.
Q: Connecting all cars externally means you’ll also utilize big data.
A: We are collaborating with organizations with which we have never had relationships to work towards creating new value.
Q: What does the future look like for cars based on the ’connected’ concept?
A: It’s impossible to predict the future with certainty, but the key is, what can we connect to vehicles to make driving more fun.
If we limit this to information-gathering mechanisms, such as access to search engines, social media, etc., then it would be no more than an extension of one’s smartphone. Keeping in mind the goal of creating a better car, in 2015, we decided to invest in PFN*1, a Japanese company researching artificial intelligence.
Q: Does this suggest a future in which automated driving will become common?
A: We don’t yet know for sure if artificial intelligence will lead us directly to the automobile technology that we imagine, but we are at the stage where we think that one of the possible uses would be automated driving.
That is why, at CES 2016*2 exhibition in the U.S., we purposely avoided using the term 'automated driving’ and instead introduced the concept of the 'collision-free car’ representing a combination of three technologies: AI, cars and communications.
(Video) Exhibition booth at NTT R&D Forum on February 18-19, 2016
Edge Computing is the Key
Q: What were your goals and methods for demonstrating a collision-free car that utilizes machine learning?
A: The demo was conducted specifically to give consumers an understanding of cars that use AI. The project is based on PFN’s logic of a vehicle learning how to drive while predicting collisions, which we will attempt to incorporate into cars.
The AI used for this purpose starts with a blank slate and becomes more intelligent over a period of roughly three hours. We thought about starting the demo from the beginning of this process, but visitors who stayed for just the first five minutes or so wouldn’t have understood what was happening. So we used cars that had already accumulated a certain level of knowledge, and we also presented a video of a car gradually becoming more intelligent.
(Figure) Storyboard of video shown at CES
Furthermore, if the demo had consisted only of cars that had been set to avoid collisions, people might have thought that we pre-programmed their movements. Therefore, in addition to the silver-colored models, we also included red-colored models that visitors could operate by remote control. This showed everyone that, even with the red models being operated freely, the silver cars moved of their own volition to avoid collisions.
Q: What kind of reactions did the exhibition receive?
A: Many people posted videos of the demo on YouTube, and we received comments from a wide variety of people, so I think we successfully raised the understanding of this concept.
Q: How did you work with NTT Group on the demo at CES 2016?
A: The capacity of semiconductors installed in cars is improving. We wanted to let the car take care of as many computations as it could and have the remaining computations processed on the cloud. We originally imagined lowering the burden on the required communications. However, it didn’t work out that way, because communication delays frequently caused time lags.
So we focused our attention on NTT’s edge computing technology (described in Vol.2,) which processes some of the data simultaneously while exchanging information at high speed in as lean a manner as possible.
(Figure) In contrast to cloud computing (left) in which data collection and processing are concentrated within a cloud server, edge computing (right) distributes processing among multiple edge servers in a network.
We wanted to know how smoothly data on the state of cars worldwide could be exchanged with data on the cloud. We consulted with NTT regarding such technology and eventually managed to develop our collision-free car.
The only problem was that, originally it was planned to operate the cars via Wi-Fi, but we knew that it was most likely that overlapping signals at the CES venue could cause poor communications and possibly prevent some of the cars from operating properly.
For this reason, the folks at NTT prepared a system for us within a short period of time that would allow uninterrupted data exchange between the cars and the cloud. The plan involved creating a wireless environment, integrating it into the demo cars and having PFN hurriedly perform the necessary tuning, etc. for machine learning.
Retaining Enjoyment while Avoiding Risk
Q: What are the future goals of Connected Car Company?
A: Our research is aimed at eliminating deaths caused by car accidents, and we think that artificial intelligence could play the biggest role in achieving this goal.
However, achieving this goal would be meaningless if we couldn’t make riding in the car enjoyable. So the key is to create an attractive car that people actually want and then work to eliminate driving risks while maximizing riding pleasure.
Q: How will this evolve going forward?
A: When using a computer or smartphone, you connect with other people through a display. In the case of driving a car, however, we have to come up with a safe, convenient interface; one that is more appropriate than a display. But the question is, what kinds of tasks could be performed easily and conveniently? Can we really achieve what we want to do?
Going forward, we will continue our research, including looking at the relationships between driver and passenger behavioral patterns and human psychology. We don’t know what can actually be achieved, but there are many things we want to try.
- *1PFN is a Japanese venture capital firm founded in 2014 and headed by CEO Toru Nishikawa. The company applies edge computing to conduct machine learning through distributed coordination by incorporating high-level machine-learning algorithms in network and edge devices.
- *2CES (Consumer Electronics Show), the world’s largest consumer electronics tradeshow, is held every January in Las Vegas, Nevada, USA. The show, first held in 1967, is operated by the Consumer Electronics Association. This year, more than 2,000 companies organized exhibitions and upwards of 100,000 visitors attended even though the show is not open to the public.