HD mapping will do to manual navigation what smartphones did to cordless phones, blueray did to cassettes, and Cloud did to CDs — supersede with a storm and render obsolete forever.
The analogy may not hold perfect, but the underlying features are the same: exponentially boosting functionality, manifold increase in utility and taking it to the next level of innovativeness. With the topsy turvy disruption in the automobile industry brought by autonomous vehicles, HD mapping will reimagine the way mapping is done, perceived and analyzed. It is a lucrative sector that is poised to reach $20 billion by the next decade.
In this burgeoning segment, Carmera is a company that is making significant strides. Its vision is to deploy next-generation street intelligence at low cost and democratize autonomous mobility.
The company develops fully regenerative HD maps which ensure that autonomous vehicles are using the most up-to-date map data to make localization, perception and path-planning decisions.
Carmera maps combine highly detailed HD base maps, with an always-on update feed that is crowd-sourced from their own camera-based sensor network—including from our commercial fleet partners. The main advantage of this approach is that it allows map upgradation without the costly, time-consuming process of rescanning roads with LiDAR and the like. The end result is a more accurate, affordable and scalable HD map product.
Recently, Carmera joined hands with Toyota Research Institute-Advanced Development (TRI-AD) to conduct a proof of concept about developing camera-based automation of high definition (HD) maps for urban and surface roads. This was the first step towards the realization of TRI-AD’s open software platform concept known as Automated Mapping Platform (AMP), which support the scalability of highly automated driving .
Emphasizing the need of active collaborations, flexibility and being optimistic about initiatives like AMP, Ro Gupta, CEO, Carmera, says, in an exclusive interview with Geospatial World, “We have long believed autonomous vehicle technology needs to be flexible, interoperable and agnostic—rather than a walled garden—so it’s great to see leading OEMs validate that belief and show leadership as in areas like mapping.”
Advent of 5G means low latency rate and faster streaming than ever, how would it transform the HD mapping industry?
I believe it’s important to keep in mind that it will take some years before most of the world experiences 5G. We are focused on providing autonomy for all, not just those who are fortunate enough to live in countries like Japan and the United States. Also, in geographically large countries like the United States, 5G coverage may not be ubiquitous till a few more years.
The key is to ensure that technology is able to reap the advantage of the speed and bandwidth 5G promises, while ensuring that it still can provide accuracy and reliability in those areas that do not have 5G connectivity.
Data would be crucial to autonomous vehicles in a lot of ways, as they would both generate and utilize it. Do you think it would lead to new scenarios when it comes to the contentious issue of privacy?
Yes, that’s why there are a few proactive approaches on data security and privacy. We believe more data isn’t necessarily better because it also adds up to the cost, so we have a very selective approach regarding what data we need for availing most signal strength, least noise and keep maps updated. Also, we only utilize mobile sensor data, not fixed cameras which tend to be susceptible to risk. For the data we collect, we do not provide access to raw imagery to deliver our service, since it’s the derived/schematic info that we sell. We follow all the standard norms for blurring faces and license plates, and other practices related to GDPR and the automotive industry, which thankfully have been treated by most in the industry as a prerequisite and not an afterthought.
On the lines of the partnership with TRI-AD, are you looking ahead at other engagements and collaborations with research institutes and automakers?
Collaboration is the key to taking the autonomous vehicle industry to maturity, as it is still in a nascent stage. We’ve long had a network of partners, stretching back to the first time an autonomous vehicle drove on our maps on a snowy day in 2017 at Michigan State University. We build our technology to be I/O agnostic, so that we can partner with any automaker, regardless of their sensor setup or the mapping format they use—Apollo Baidu, DARPA/RNDF, something custom or proprietary, or anything else.
Instead of LiDAR, you are using dashboard-mounted cameras for feature detection and gaining insights. Do you think cameras can have an upper hand over LiDAR?
It’s not a question of which has the upper hand; it’s more about knowing when to use each technology. LiDAR provides incredible detail and precision, so for building HD base maps, it’s the gold standard, and our default choice. But when it comes to tracking change, LiDAR is simply not scalable. That’s why we use commodity cameras, like dashcams or cellphones, to track change—because they give us the ubiquity of coverage needed to really track changes in near real-time, anywhere our clients need.