Self driving cars have exploded into the news as of late–from integrated sensors to increased computing efficiency, the self-driving car craze has hit the mainstream. Tesla, in conjunction with NVIDIA, released a vehicle which could drive itself and perceive the environment around it. NVIDIA has taken in-car computing to a new level.
The NVIDIA ‘Xavier’ AI-on-a-chip is capable of more than 24 TOPS (trillion operations per second) in deep learning operations. The chip is designed to incorporate data from 12 car mounted cameras, LIDAR, Radar, and ultrasonic sensors to determine which operations it should execute and when. Xavier incorporates AI to learn from its owner’s driving patterns.
How does this change self driving cars
Self-driving cars have traditionally relied on beaming information elsewhere for processing and then receiving the post-processing information. The tenths of seconds or more that this process takes have led to negative consequences, when automobile accidents can occur and be finalized in a few tenths of a second.
NVIDIA has succeeded in bringing a computer with processing power equal to 150 MacBook Pro computers into the vehicle to process all of the information onboard. By reducing computation times to only time which a centralized computer has to process the information, NVIDIA is bringing significant improvements to the viability of self-driving cars.
How can this impact the mortality rate of drivers?
NVIDIA has partnered with Bosch, Audi, Tesla, and other automotive manufacturers to bring self-driving car technology to the market in force. The CES 2017 showing by NVIDIA provided a glimpse into the capabilities of an AI-based self-driving car with an incredibly powerful onboard computer.
The computers in question consist of between one and four 12-core processing units. They can process 8 teraflops and 24 TOPS of deep learning operations per second. How energy efficient is a computer of this capacity? It operates on 250W of power. The lowest average desktop computer power supply starts at ~200W of power.
How does this affect the driving economy?
Nearly 30,000 drivers died due to automotive accidents last year, NVIDIA is working to eliminate the most ineffective mechanism from vehicle operation, the driver. By replacing a human driver with limited capability to identify and react to stimulus as well as reactions such as panic, NVIDIA and its partners are poised to lower driving fatality rates significantly.
Beyond human fatalities and automotive repair costs, NVIDIA is also helping to reduce resource costs due to traffic. Traffic wastes not only fuel and engine wear, it also wastes time. Time is one of the most valuable resources humans have, we cannot turn back the clock. Though with modern research, we may be able to reverse some of the effects of aging and extend life, we do not know the long-term implications of these technologies.
How will this alter traffic patterns?
Self-driving cars will likely eliminate most if not all traffic once they reach the point of critical acceptance. When the significant majority of drivers are using self-driving cars, the likelihood of accident will decrease, alongside the number of cars needed on the road at any given time. Fewer cars yield to less traffic, as does more consistent driving and automated route planning.
Self-driving cars are likely one of the most significant leaps in human efficiency since the personal computer. By reducing the amount of vehicles on the road, time in traffic, and freeing up humans to carry out tasks such as eating and working while in transit, self-driving cars have the potential to advance most coming technologies significantly. The average citizen spends 42 hours in traffic per year, some spend more than that.
NVIDIA and its partners in this endeavor are ready to put self-driving cars with supercomputers onboard into commercial and public use. The long-term effects of increased acceptance include reduced traffic and resource consumption, less time in traffic, and fewer accident-related deaths per year.