Super Computer meets Super Car


 The only thing missing is a cape!

 Beyond traditional F1 racing and the two-year-new Formula e (electric racing) series now comes supercomputer racing – establishing the first racing series with no actual drivers.  And because there are no race drivers, the cars don’t need standard driving systems and will look radically different. Formula CEO Alejandro Agag, is an investor in RoboRace and has been enthusiastically endorsing it. If you can believe it, the basic – know where you are on the map – and sense/plan/act brain of self driving cars barely visible on few roads today, is already moving into deep learning mode.

As much as all this sounds like science fiction, Denis Sverdlov CEO of RoboRace says they will have a running race car by the fall of 2016.  Little has been shared of the design philosophy or the racing mechanics so far but the RoboRace executive has been traveling the world, meeting with executives from companies who can help mould this dream into reality.  What has been shared is the race series will be comprised of 10 teams, each with two cars that are identical to each other and all cars will be powered by an NVIDIA Drive PX 2 brain. The cars are full size, weighing in at maybe 2,200 lbs and having high performance electric motors powering each wheel with speeds on some racetracks maybe approaching 180mph.

The idea is to get the different teams to build and train their own neural networks that will power their race cars. The Drive PX 2 will process all real-time sensory data gathered from the car, but it will have to compute with each team’s network to make sense of the data and take appropriate action under tough racing conditions. This would include making behavioral learning maneuvers at racing speeds that would not be considered for street cars under normal and basic operating conditions.  However information learned from RoboRace cars could be very useful and critical for future driverless street car road safety such as accident avoidance and hazardous road conditions. Since the teams will have identical cars and identical hardware; the team with the most robust neural network will likely have the best chance at winning a RoboRace.

The heart of self-driving is of course, Artificial Intelligence. Technology currently championed by NVIDIA built GPUs. NVIDIA claims that by working with a large community of self-driving car experts, their object recognition, self driving scores and accuracy in this arena is unsurpassed. Also, while along the way, developing their network with peripherals such as cameras, LIDAR sensors and HD-3D mapping, NVIDIA invented a new neural network called “DaveNet”. The result, a complete end-to-end computing platform with very deep-learning capabilities and the ability to analyze a minimum of 1.8 million points per second.

The PX-2 and the RoboRace series sits firmly in NVIDIA’s grand scheme of things. Enough so, Jen Hsun Huang, NVIDIA’s CEO closed his keynote speech with it at their 2016 GPU convention in San Jose with a huge applause. If NVIDIA is ready to prove their Drive PX 2 is more than ready to face one of the toughest challenges an automobile can take, then what better way to prove its worth than a high-tech stakes race series.

NVIDIA has been tight lipped in revealing any further details of the NVIDIA Drive PX 2 system other than they use next generation Tegra processors and a couple of Pascal GPUs. As we get closer to RoboRace becoming a reality, more details of the systems and also the mechanics will be revealed. The series will absolutely push the limits of self driving automobiles.


This is the new NVIDIA Drive PX 2 you can hold in your hand. The front face reveals two next-gen Tegra processors and a PLX switching chip to interface with the twin Pascal GPUs.


On the backside you catch a glimpse of the two Pascal GPUs used to augment the Tegras on the front. Prior to this the super computer would take up a good portion of a car’s trunk.

Author: Bob Koveleski, San Diego. Automorrow


Leave a Reply

Your email address will not be published. Required fields are marked *