Denso and Toshiba to develop vehicle AI modelled on human brain

The technology will be used to improve safety systems in autonomous vehicles

Denso and Toshiba are to develop artificial intelligence for vehicles modelled on the human brain.

The AI technology, called Deep Neural Network (DNN), will be used in image recognition systems independently developed by the companies to support the steady introduction of advanced driver assistance and autonomous driving technologies.

The area of AI neural networks is a growing field, where algorithms or actual hardware are loosely modelled after the neuron structure of the mammalian cerebral cortex but on much smaller scales. These systems are able to respond and learn from external input, just like the human brain.

DNN is an algorithm modelled after the neural networks of the human brain, yet the joint-venture between the companies is hoping the technology will be able to perform image recognition processing to an even higher level.

The shift towards automated driving means automotive computers need to be able to comprehensively identify different road traffic situations, obstacles and road markings. However, in conventional image recognition using patterns and machine learning, objects must be individually characterised and extracted.

In DNN-based image recognition, a computer can extract and learn the characteristics of objects on its own, significantly improving the accuracy of detection and identification of a wide range of objects.

DNN technology has progressed rapidly over recent years, and the technology will be refined and implemented into smaller in-vehicle processors to provide additional support for automated driving systems.