Autonomous vehicles require computationally-demanding processing that needs to be performed in a reliable and swift fashion. A local system performing the processing would satisfy both of these criteria; however, currently available processors are not well-suited for the computationally intensive nature of autonomous navigation. This has resulted in heavy reliance on cloud infrastructure for processing, which presents performance issues such as bandwidth and communication bottlenecks, as well as security issues such as susceptibility to man-in-the-middle type attacks. However, the researchers' team led by Professor Eren Kursun at the Department of Computer Science, Columbia University, has developed a neuromorphic processing system with the architecture optimized for computationally-intensive autonomous navigation applications.

This innovative neuromorphic processing system provides a powerful, local, real-time processing for autonomous guidance systems. Using this system, scientists managed to demonstrate a promise in efficiently computing highly unstructured streaming data, such as the visual and audio data a self-driving car needs to process. This technology is a neuromorphic processing system that is optimized for computationally-intensive autonomous robotics applications and describes a custom-designed architecture that is optimized to process data collected by an autonomous guidance system. Using a cognitive processing approach, this system allows for local, reliable, real-time processing for autonomic guidance.

The system can be applied to self-driving car navigation systems, flight-assist systems, autonomous warehouse robots, autonomous military robots, autonomous cleaning robots and autonomous drone navigation. Furthermore, this invention was optimized for the processing of autonomous navigational data and its compact design allows for local processing. In addition, the newest system can operate without continuous access to a network and offers greater security compared to cloud-based processing.