Robust perception, decision making and path prediction for autonomous vehicles

Development of autonomous vehicles are seeing a grown in many different applications. As we increase the levels of automation and move into Self-driving cars, it is expected that these systems will combine a variety of sensors to perceive their surroundings in a robust manner and avoid human errors. They also need to adapt quick to changes in their environments and make decisions based on a number on inputs that might be contradictory. Moreover, reports have found that autonomous vehicles can be vulnerable to a wide range of attacks such as physical perturbations or back-end malicious activity. This project looks at AI robust perception methods that combine a range of sensory inputs including computer vision to identify changes and suspicious sensor responses in order to make robust decisions on motion and planning. The project will also look at terrain models and vehicle- terrain interaction using specialised sensors such as hyperspectral cameras. Terrain map requires numerous on-the-ground observations and sample collection. The use of intelligent models obtained via hyperspectral sensors combined with artificial intelligence provides an alternative technique to traditional mapping that can derive not only quantified terrain information but also texture and manoeuvre decisions. The project combines, sensors, hyperspectral imaging, signal and image processing, artificial intelligence and control.