Bosch Invests in Mapping Company DeepMap

The corporate venture capital company of Bosch, Robert Bosch Venture Capital has taken a stake in Deep Map, a US based software startup that specializes in mapping localization for autonomous vehicles.

Commenting on the acquisition, RBVC managing director Dr Ingo Ramesohl, sais that for safe autonomy with regards to autonomous vehicles, it is important to have maps that are explicitly designed to be read by machines. Deep Map has developed an approach to mapping that uses embedded software on the vehicle and can be highly relevant highly automated and autonomous vehicles, both within Bosch and for the entire automotive industry.

Ramesohl said, “DeepMap builds upon the growing RBVC portfolio of technologies addressing the autonomous vehicle vertical, next to start-ups like AImotive and Tetravue.”

DeepMap has taken a totally different approach to tackling the mapping challenge and its technology can be used to create custom HD maps, which come with different characteristics and data structures.

DeepMap also develops software which can be embedded on the vehicle to address localization, change detection, data updates, route planning, HD map caching, and data collection.

Using these capabilities, DeepMap’s goal is to come up with “living”, continuously updated HD maps rather than a static map database. It would instead have a full software layer available in the cloud and embedded on the vehicle.
DeepMap is also working on a new model for data ownership based on which the end customer would owns the underlying data.

This strategy enables them to take a more scalable, low-cost approach to HD map making. Instead of owning and operating expensive survey fleets, data can be collected from the customer’s autonomous vehicles deployed in the field.

High Definition (HD) maps are complete and accurate volumetric reconstructions of a 3D environment and provide a critical set of “pre-information” to autonomous vehicles.

The data would include information related to parameters like lane connectivity, drivable areas, crosswalks, signs, signals, traffic restrictions, and other information that can be highly specific for each end customer’s solution.
HD maps are extremely precise, and are typically at the centimeter-level.

In general, autonomous vehicles need very precise instructions on how to negotiate roads in the real world. The greater the precision of the HD map, the simpler is the self-driving software that is needed to obtaining higher performance.
James Wu, CEO and co-founder of DeepMap, said: “Self-driving car development is both a technology challenge and a product challenge. It needs to be safe, affordable, and scalable. In the self-driving era, a map is no longer ‘just a map’.
It’s an integral part of the car’s brain.”

The company has a 75-person team that has extensive experience at launching maps at a massive scale. These include specialists who have had prior stints developing Google Maps, Apple Maps and the Baidu HD Mapping initiative.