Research
Mohammad is a PhD student at Wolfson School, where his studies are focused on probabilistic deep learning for safe and reliable tyre-road friction estimation. His studies are a joint program between Loughborough University and Volvo Trucks.
In the years 2010 −2013, approximately 120,000 people died from car accidents where, 24% of which is weather-related accidents occurred on icy, snowy, or wet pavement or in the presence of rain, sleet, fog and snow.
It is reported that there is a correlation between road surface friction and accident risk. Active safety systems such as emergency collision avoidance (ECA), active front steering (AFS), anti-lock braking system (ABS), direct yaw moment control (DYC), and traction control system (TCS) all rely on the knowledge of the road friction coefficient.
These modules mostly rely on dynamic models for state estimation or measuring the system’s response in real time. While the car manufacturers utilises dynamic models to measure the friction in real-time, a predictive approach can enhance the adjustability of vehicles by estimating the surface friction ahead.
Therefore, the safety-related modules can be improved significantly if the environmental factors affecting them can be estimated in advance. Knowing friction coefficient ahead brings extra advantage for the components like ADAS, ASS, and autonomous navigation systems (ANS). This project aims to formulate the problem of friction estimation as a visual perceptual learning task.
The project aims to estimate the friction between tyre and road surface ahead using a monocular dashcam. Mohammad is studying neural networks, variational inference, semantic segmentation, latent variable models, and synthetic data.