The auto industry is currently experiencing a shift to vehicle autonomy.
New cars with advanced self-driving capabilities are coming out annually, with each model more sophisticated than the last. Delving into what is driving this rapid shift – technology companies are emerging in this sector. These companies bring novel and updated products to the market constantly. As a result, traditional OEMs, which are manufacturing companies, are being forced to scramble and adapt – to be competitive and profitable in the long run, they must match the accelerated timeline set by technology companies. They must bring vehicles equipped with advanced autonomous vehicle (AV) features from “concept” to “market” at an exponentially faster rate than the traditional norm.
Today, the industry is in an era of partial autonomy – specifically, vehicles are capable of driving on their own on the highway (in the presence of a human driver). As mentioned above, technology companies have spearheaded this evolution to highway autonomy, and traditional OEMs subsequently followed suit. The next logical step is for vehicles to navigate on their own in urban areas (again, in the presence of a human driver).
However, city streets, with their dynamic and complex nature, open the door to a plethora of difficult variables to account for in comparison to the highway. First off, cities include road occupants other than automobiles, including vulnerable road users (VRUs). Bikers’ and pedestrians’ unpredictability, in comparison to traditional vehicles, presents a danger for AVs that must be mitigated.
Also, urban settings are filled with traffic lights.
Given that the stage of these devices is constantly changing, the correct phase must be accurately detected no matter the surrounding conditions (inclement weather, bad lighting etc…). Unfortunately, currently, there is no system that has the capacity to enable this capability in AVs.
Lastly, metro areas have an exponentially higher volume of vehicles, resulting in cars being in extremely close proximity. Cities also include constant “stop-and-go” traffic that requires one to be extremely wary of his/her surroundings. Unlike the highway, where speeds are constant, and vehicles have ample room (and thus time) to make any maneuver necessary, AVs in urban settings must be equipped with a platform that allows them to make accurate decisions faster than ever, given the constraints of the environment.
Presently, the industry has reached a barrier with highway autonomy – there is nothing that can facilitate the next step and enable AVs to safely and efficiently traverse metro settings. This is so because current solutions are based on repurposing legacy technology. As a result, these platforms are technologically constrained, preventing them to have the compute and power efficiency necessary to enable urban autonomy.
When we drive, to process the plethora of surrounding visual information in real-time, our brains utilize a datacenter level of compute while consuming less power than a lightbulb. To mimic these capabilities to enable urban autonomy, an AV must be equipped with a system that can generate a minimum of 75 Tera-Operations-Per-Second (TOPS) of compute for every watt of power consumption. This unsolved optimization problem is known as the visual perception problem.
We @ Recogni are developing a novel solution from the ground up.
Our visual perception platform is based on key innovations in mathematics, AI, and ASIC architecture. As a result, our solution generates 100 TOPS for every watt of power consumption. Clearly, we are the only ones on the market capable of solving the visual perception problem outlined above, which is key to enable urban autonomy.
Traditional OEMs must horizontally integrate this solution into their vehicles. In doing so, they will be able to equip their cars with a system that can enable urban autonomy. As a result, they will be able to compete with emerging technology companies to sustain profitability in the long run.
To learn more about Recogni, check out www.recogni.com
Author: By Sidhart Krishnamurthi, Product Management @ Recogni
Statements of the author and the interviewee do not necessarily represent the editors and the publisher opinion again.