"HD Mapping" and "Connectivity" (V2V, V2I) are both viewed as indispensable predictive tools for highly automated driving.
In the context of highly automated driving, describing HD mapping simply as “mapping” is misleading.
Today, HD mapping is considered must-have data for highly automated driving. It can provide information just as critical as other sensory data generated by vision, lidar and radar.
HD Mapping (Source: Here)
Liran Bar, CEVA’s director of production marketing, had his finger on it. Asked why mapping is such a big deal, he said, “It’s all about prediction.” HD mapping enables you “to predict before lidars and radars detect that a road is blocked ahead, for example,” he explained.
It’s a hugely important element in the advancement of the robo-car. “HD Mapping” and “Connectivity” (V2V, V2I) are both viewed as indispensable tools for “prediction.” But much of their implementations needs a buy-in from everyone in the automotive industry, and a consensus to build common infrastructure.
We’ve witnessed some progress. But I don’t think we’ve gone far enough. In this column, let me zoom in on the HD mapping issue.
At the CES last week, Here, owned by a consortium of German automotive companies including Audi, BMW and Daimler, garnered numerous partnership deals.
Aside from Intel’s announcement to buy a 15-percent stake in Here, Here and Nvidia also announced at CES an extended collaboration to develop Here HD Live Map into a real-time, high-definition mapping solution for autonomous vehicles.
Mobileye also disclosed that it is now working with Here to provide Mobileye’s own crowd-sourced, high-definition mapping technology called Road Experience Management (REM).
Phil Magney, founder and principal advisor for Vision Systems Intelligence (VSI), explained, “Here brings value by providing the location platform and the infrastructure to support it. Mobileye brings a crowd source of data that collectively will pool with Here data to create a localization platform.”
REM is a technology unique to Mobileye.
As Mobileye describes it, REM is “is an end-to-end mapping and localization engine for full autonomy.” The solution consists of three layers: harvesting agents (any camera-equipped vehicle), map aggregating server (cloud), and map-consuming agents (autonomous vehicle).
The harvesting agents collect and transmit data about the driving path’s geometry and stationary landmarks. Mobileye’s real-time geometrical and semantic analysis, implemented in the harvesting agent, allows it to compress map-relevant information – facilitating very small communication bandwidth (less than 10KB/km on average).
Crowd sourced Roadbook (Source: Mobileye)
Under this scheme, the relevant data is packed into small capsules called Road Segment Data (RSD) and sent to the cloud. The cloud server aggregates and reconciles the continuous stream of RSDs – a process resulting in a highly accurate and low time-to-reflect-reality map, which Mobileye calls “Roadbook.”
Amnon Shashua, Mobileye’s co-founder, CTO and chairman, sees Roadbook as a layer above what Here offers. He stressed, for Mobileye, “Mapmakers are not competitors.”
Aside from Here, Mobileye announced a partnership with Japanese mapmaker Zenrin and an OEM to complete HD-Mapping of Japan. “By October, 2018, we get this done,” Shashua promised. The ultimate goal for Mobileye is to develop “World Roadbook” in collaboration with a variety of OEMs.
HD mapping isn’t the only thing that needs to be shared among car OEMs, however.