Tokyo: urban experimentation and the 2020 Olympics

A history of experimentation

Since the Meiji restoration, and including Akira and Godzilla, Tokyo has always been the locus where the future fuses with the present and where one would expect the automated driving future to be well under construction. Indeed there are so many projects and experiments around vehicle automation it’s hard to keep up. And when one talks vehicle automation, one has to talk about high definition maps. Such maps offer a reference model of reality, allowing vehicles to position themselves correctly within the lane of travel on the road. By knowing where it is located precisely, the vehicle can plan forward, deciding which lane to take next on the way to its destination and safely navigate its surroundings using its increasingly powerful suite of sensors. So the map is the starting point. Yet when I arrived back in Japan in 2014 after a 6-year spell in Europe, the topic from both a technical and national standpoint was far from resolved.

Looking back, Google seems to have kickstarted the change in mood in 2013 with their fully automated driving experiments, including their iconic Firefly design, leading Japan on its new journey towards automation. In fact, Japan had been the early pioneer of automated driving tests back in the 1990s with experiments from advanced driver assistance to truck platooning, yet nothing was commercialized at mass-market level. In response to Google’s successes, Japan moved to accelerate progress by kicking off a mapping project where Japanese surveyor Pasco was asked to map the future Olympic village area of Odaiba in high definition. Using Mitsubishi’s MMS vehicles, originally used for roadworks and surveying, the data produced was super high resolution and super high quality. But that was a problem, firstly, because collecting data is only the start of the process. It then needs to be scrubbed for irrelevant/spurious information, the raw point data needs to be categorized as either road geometry or roadside objects, those objects need to be created and aligned to the road network, then the data vectorized for best performance, and formatted for use in the vehicle. With very high resolution data, there is simply lots of data to go through. Then comes the standardization issue – unless the data is fairly simple in structure, it will be exceedingly difficult to compile and also to update later on.

Secondly, government expectations would be that the data produced would include every little detail along the road as it always has for infrastructure projects and carmakers expected it to do the same so that they had maximum flexibility about what systems to build. Which may be useful as a blueprint for roadworks, but is not relevant for vehicle systems that may just need to know the location of the kerb in relation to the vehicle. With hindsight both parties could have realized that the expectations had been set too high for the task at hand, but at the time it felt like the project was being setup for failure. And while the project did eventually complete, the limits of the super high fidelity approach soon became apparent. Unless there was a significant amount of simplification (or standardization) in the approach to data collection and testing, especially considering the rather simple vehicle systems still being built at the time, accelerating Japan’s progress would remain difficult.

A year later in October 2015, the government requested a similarly scoped project, but this time from a consortium of surveying and mapping companies including Pasco but also other industry leaders such as Zenrin and Aisan Technology. The exercise led to the setting up of the Dynamic Map Planning Company in June 2016, channeling an investment of 3m yen from 6 Japanese mapping and surveying companies and 10 Japanese automakers. Fast forward to 2017 and this company is now Dynamic Map Platform company, a consortium setup to build the next-generation map of Japan for automated driving. At the same time, the government has put together what it calls the “Large Scale Experiments“, offering a fund of $15m to develop the building blocks required for automated driving including HMI and high definition maps. The geographic area covers just under 350km of road network including part of downtown Tokyo and Olympic Village areas, part of Ibaraki prefecture and JARI’s new automated driving test facility there, plus the Tomei expressway heading west out of Tokyo part-way to Nagoya. According to an article a couple of days ago in Response, 20 Japanese companies have been selected to develop these components and work is underway with results expected in late 2018. Hopefully this data will be made available to the public, so that any developer or businessperson can easily develop their own app or service on top of this automated driving baseline data.

Tokyo Olympics on the horizon

So while all this activity had started with Google’s early experiments, 2013 was also the year Tokyo was nominated to host the 2020 Summer Olympics adding to the urgency. During the last Olympic games in Japan back in 1964, Japan was at the beginning of its postwar era boom, having freed itself from US occupation a mere 12 years earlier. Its growing modern expressway infrastructure and world-leading high-speed bullet train “shinkansen” showcased Japan as innovative and technology-driven where the war was in the past and a bright future was under construction. 56 years later Japan will be trying to improve on that performance, and Toyota as exclusive automotive sponsor has all the incentives up to help make that a success.

While originally rather late to the game (according to MIT Technology Review CEO Toyoda was still debating the automated driving topic in 2014), in 2016 the company made a decisive $1bn investment by setting up TMI in Silicon Valley, a company dedicated to accelerating Toyota’s AI and automated driving research. Fast forward one year and Toyota is aggressively ramping up its automated driving hiring plans in both the US and Japan, and promises to demonstrate a number of showcases at the Tokyo Olympics in 2020. They have recently also brought forward their plans for fully automated driving models to the early 2020s, more in line with other industry leaders including Elon Musk’s Tesla, Nissan, GM and the German premium OEMs. But 3 years is not long in automotive years. The challenge will be whether Toyota can recreate a diverse, global, agile culture and flat organizational structure similar to TMI in Japan, while leveraging the core values and ingenuity that made Toyota a global leader in transportation. If it can succeed, they may well become the centre of some of the most amazing innovations leveraging their massive scale at home and abroad.

But there are also many venture companies already powering ahead aggressively. One example is Navya Arma, a French next-gen mobility company from France, which has been very successful in introducing its automated electric minibus prototypes to Japan via SB Drive, the automated driving venture of IT giant Softbank Group. The recent round of Navya Arma test drives from July 17-23, open to the public at Shiba Park in the shadow of Tokyo Tower in the centre of Tokyo, was well-received by local press and the vehicle is now being showcased up in Hokkaido from Oct 14-16. SB Drive has also just received an extra $5m round of funding back in March 2017 (making it effectively a 51-49 JV between Softbank and Yahoo) which will help continue the buildout its base of automated drive experimentation across Japan and we assume continue promoting Navya Arma as their vehicle platform of choice.

Another example is ZMP, a Japanese robotics company, who have been actively pushing their vision of automation and originally partnered up with blog and mobile game giant DeNA before they broke the alliance and went their separate ways. DeNA later teamed up with Renault-Nissan for their sw platform and EasyMile for their vehicle platform, while ZMP continues with its plans to offer robotic taxis in the Olympic village area with their robocar systems developed in-house. Back in July 2017, ZMP teamed up with Hinomaru taxis, one of the largest taxi providers in Tokyo, to start a study group that hopes to morph into a consortium of taxi providers across Japan dedicated to offering highly automated taxi rides to the public during the 2020 Olympics. According to an article in online magazine Motorcars, the taxis will include a human driver in the first tests this year, and progressively increase the availability area from 2018 to 2019 before launching with the service in 2020. The big jump will be to move with a driver in the loop to no driver, so hopefully that will arrive early to ensure all remote operation issues are ironed out well ahead of time. Regardless, with a large enough pool of taxis gathered through this consortium, a common hailing and operations platform powered by an end-to-end technology suite from ZMP, this could very well create a decently-supplied taxi service that would support the massive increase in demand expected during the Olympics.

It is an interesting sharpening in focus: ZMP focusing on pooling taxis and offering automated rides typically in urban areas, SB Drive focusing on automating bus routes for local governments in rural areas. Will also be interesting to see how Uber Japan fits into the bigger picture. What is clear is that Tokyo will continue being a hotbed of urban experimentation especially in the leadup to 2020. After which one can expect the urban focus to shift to business users via logistics and drones (Rakuten and Amazon anyone?), while rural areas will see accelerated mass-market commercialization for everyday passengers. I personally believe that the last item will represent the biggest transportation shift Japan has seen since the post-war expressway network expansion and subsequent Shinkansen network buildout. And it is coming our way faster than any of us will likely expect.