German engineering and electronics company, Bosch, has teamed up with US semiconductor company Nvidia to develop an artificial intelligence (AI) system for autonomous vehicles.
Speaking at Bosch Connected World in Berlin – the company’s annual Internet of Things (IoT) event – Bosch CEO Dr Volkmar Denner and Nvidia CEO Jen-Hsun Huang announced the product, which they say will be available to the mass market.
The agreement will see the companies develop an AI self-driving car computer built on Nvidia’s deep learning software and hardware, meaning vehicles can be trained remotely, operated autonomously and updated via the cloud.
The power behind the machine
The Bosch AI car computer system will be based on Nvidia’s Drive PX technology, an open AI car computing platform that should enable automakers and suppliers to accelerate the production of autonomous vehicles.
The platform will come with the recently announced Xavier AI supercomputer, the world’s first single-chip processor designed to achieve Level 4 autonomous driving – the level at which a car can drive itself without any human intervention. Nvidia claims the chip can process up to 30 trillion deep learning operations a second while drawing just 30 watts of power.
This technology combines deep learning, sensor fusion and surround vision to enable to vehicle to understand its surroundings, locate itself precisely on an HD map, and plan a safe route forward all in real time.
Training a car to drive
Nvidia’s Huang said his company will deliver technology enabling Level 3 autonomous capabilities (the level at which a car can drive itself but still requires driver intervention in some situations) by the end of this year, and Level 4 capabilities by the end of 2018.
Huang noted that while many brands (Ford, Audi, Tesla) are working on autonomous vehicle technology, such vehicles will require unprecedented levels of computer power that advanced driver assistance systems (ADAS) cannot provide.
Huang continued by saying that coded software can’t possibly be written that would anticipate the almost infinite number of things that can happen along a road, such as cars straying from their lanes, shifts in weather conditions, or animals that stray into the road.
“Self-driving cars is a challenge that can finally be solved with recent breakthroughs in deep learning and artificial intelligence,” Huang noted.
Speaking to Internet of Business recently, academy director at the Transport Research Laboratory, Nick Reed, clearly shares this view.
“The key aspect to consider with AI is that driving is infinitely variable – you never know what combination of pedestrians, traffic and weather you might encounter,” Reed said.
“You need the vehicle to have sufficient learned capabilities to know how to handle unexpected situations.
“It is possible to achieve this with the use of AI. However, it needs to be trained on masses of data to enable the AI to learn how it should react to ensure the vehicle makes safe progress in the widest range of potential situations that it might encounter. This is a key challenge for the industry at the moment.”
With Nvidia providing the computing know-how and Bosch opening the right doors through its extensive automotive network, the two companies are well-placed to tackle this challenge head-on.