Since antiquity, when sea routes and land routes were perilous for people and goods, having up to date and highly accurate maps have been a requirement for safe travel. These maps were patchworks of military sketches, captain logs and merchant updates, and over the centuries became more and more detailed. In an era of increasing vehicle automation, having up-to-date and highly accurate maps is equally critical to ensuring safe vehicle operation both on the highway and on surface roads. These maps work as a virtual on-board sensor connected to other vehicle sensors in other vehicles in realtime, allowing on-board sensors to understand the context of the world around them and provide a reference model for decision-making beyond the sensor range.
Typically a vehicle system will have a preloaded map database on the vehicle that can support localization (understanding where the car is at any point in time), path planning (drawing a course onto the local road network in realtime to get to your destination safely) and vehicle operation (triggering actuator actions that will move the vehicle to the planned location precisely). If there are new maps available and the vehicle is connected to the internet, new maps will be downloaded and updated locally, then used to update other vehicle functions. Ideally this will be done over mobile internet, but as few vehicles are connected today, this is usually accomplished via USB at the dealer or in some cases over wifi.
So today we need to build high definition maps that can be read quickly and efficiently by machines. Yet map building remains an exhaustive and highly manual process, meaning realtime high definition map updates often propagate through to the user with a delay of months or years from the date when they were collected, rendering them useless beyond limited driver assist functions. Furthermore data resides in silos across industries, and also within industries, such as within OEM or fleet management IT infrastructures, making the sharing of data difficult. With the explosion in the number of vehicle sensors, ECUs and increasing richness of output data, the exponential growth of data availability has not led to an exponential growth in user benefit or revenue. Increasingly data is being seen as cost rather than an opportunity. As a result, each dataset is limited to often stale data releases focused on certain areas in certain countries where each of these companies offer their services.
Traditional high definition map sourcing and development requires the developer to license data in the $100s per vehicle lifetime (depending on the number of vehicles), while implemention typically requires millions of dollars for map and map system customization that is often not reusable across vehicles. Road coverage remains sparse, hugging the main highway arteries in a handful of countries. The combination of minimal coverage, stale data and massive cost effectively restricts the benefit of automation to a small percentage of developers in the largest companies with the largest R&D budgets.
In addition, due to increasingly strict regulation around the management of data, processing, storing and extracting value from data is increasingly onerous on the developer. And many more tasks are necessary to provide safe map provision to the vehicle including but not limited to: building maps from data, machine learning for data cleansing, integrating new data with original data, designing IT infrastructures to cope with ups and downs in data throughput, and devising different map update schemes for different vehicle platforms. Very few of these processes are automated, and very few of these processes can provide value to a business’ bottom line. It is time for a new paradigm.
Our new vision for the future of mapping is called the Automated Map Platform or "AMP". It is an open software platform that is based on a contribution model: participating developers accept that vehicles deploying their application or software contribute anonymized sensor data to the platform. In return, every developer has easy, safe, open and sustainable access to high definition maps from across industries, fleets and car makers. One-stop-shop open APIs that allow developers to focus on building software. No need to worry about specific map implementations and maintenance, just pull down the data needed whenever it is needed. Updated maps will use mainstream camera data, the map update and building process will be increasingly automated, and satellite imagery will be help build out a global road network where vehicles are not present in sufficient numbers to support a data pipeline.
We hope this new vision will help startups accelerate their development of automated driving applications into the market. It will help developers and city planners increase the efficiency and automation of city center mobility, a massive issue across the world especially in developed as well as emerging economies. It will help support the revitalization of rural mobility where the economics of running public services have led to rail and bus lines shutting down in Japan as well as many other countries. This cuts off rural communities from critical social infrastructure from food to healthcare and education. And it will help increase the benefit of lower levels of automation (otherwise known as advanced driver assistance systems or ADAS) to help relieve driver stress, as well as help prevent cases of mistaken acceleration, mistaken entry into the highway, and collision prevention in case of vehicle misoperation or misinterpretation.
Yet AMP is still a vision not a product. And day to day we continue working on improving traditional mapping systems for next generation automated driving systems to launch in 2020 and beyond. But a shift towards automating map production which will lead to fresher maps, lower costs and broader coverage including surface roads in the future. If all this sounds exciting, join us! We have started hiring GIS veterans, machine learning and image processing engineers, dev ops specialists, data scientists, and mapping experts into our new Tokyo office. Ideally you’ll have international experience in both large and startup environments. Japanese language skills are nice to have but optional, English is a must. Most importantly we need your passion to help accelerate this vision into reality. And achieve the vision of mobility for all, no matter where you are.