There is an ongoing revolution in developing autonomous intelligence machines to greatly improve effectiveness and efficiency of sensory perception, decision-making, and active control. However, there are a number of challenges raised in algorithm design, particularly regarding the core components of brain-like intelligence of see-think-act, beyond hardware.
This project will create a high-risk high-reward brain-inspired real-time closed-loop platform, leveraging neural spikes, to develop efficient and effective autonomous driving methodologies and techniques. It aims to address key challenges in sensory scene understanding (‘see’), multisensory information fusion (‘think’), and decision-marking and active control via cloud computing (‘act’). The aim is to develop a novel framework utilizing state-of-the-art machine-learning methodologies bolstered by computational principles of neuronal systems in the brain. Then we will create a virtual environment of simulated autonomous agents to verify the framework. In the end, we will integrate it with well-equipped autonomous robots to advance the next generation of intelligent machines. The outcome of this project will significantly change the paradigm of research and applications on the development of efficient machine vision and decision making systems in autonomous systems. It will have a wide and far-reaching impact on intelligence machines in the applications of neuroprosthesis for human wellbeing, neurorobotics for industry 4.0, and autonomous vehicles for society.
The supervision team has expertise in machine learning, computational neuroscience, computer vision, virtual reality, robotics, and autonomous vehicles, with regularly published research papers at the very top venues in related fields.