No Really, Are AVs Ready to Share the Road?
To trust or not to trust: that is the question
Just as several AV companies have started small scale deployments, there’s been declining confidence in the future of AVs - “USD100Bn invested, nothing to show for it” has been the rhetoric. Coincidence? We think not.
In the past few months, there have been several reports of AVs malfunctioning to the point of holding up traffic for hours or worse, getting into accidents. This begs the question, are AVs ready to share the road?
Hey AVs: have you trained on this kind of cyclist?(Twitter)
[Fill in the blank] should be its own distinct object class?
In an attempt to safely navigate through mixed traffic, many AV companies, like Argo.ai and Mobileye have proposed a bespoke approach. For example, in order to recognize cyclists, Argo.ai emphasizes “training its models in a way that specifically notes cyclists, cycling infrastructure and cycling laws…” The recommendation is that cyclists should be a distinct object class (TechCrunch). Intel Mobileye tries to classify open car doors specifically. Could there be another way… as there are too many objects for each to have its own distinct object class.
BlueSpace: defines object by its motion
Realizing that all the millions of miles driven and training data collected still haven’t solved the problem, we built BlueSpace’s Independent Safety System (BLISS). Leveraging next-gen 4D sensors, we’re able to arrive in an entirely new location, and have our software detect objects and predict their motion to inform autonomous vehicles to navigate safely in frame zero, Day 1 (of us arriving in NYC) - see our video.
Now back to self-driving trough of disillusionment - BLISS can be standalone L4 solution or an add-on to existing. Some of our partners utilize BLISS enhance their existing autonomy solution. In the process, they are saving significant costs related to data collection, annotation, and training.
Whether in SF, NYC or another location of your choice, BLISS is ready for anything, anytime, anywhere.