Background:
We address the problem of localizing non-collaborative WiFi devices in a large region. Our main motivation is to localize humans by localizing their WiFi devices, e.g., in search and rescue operations. We use an active sensing approach that relies on Unmanned Aerial Vehicles (UAVs) to collect signal-strength measurements at informative locations. The problem is challenging since the WiFi signals are transmitted at random times, they are very noisy and they are received only when the UAV is in close proximity to the device. For these reasons, it is extremely important to use a technique which can make robust and efficient estimates with sparse and noisy data. Bayesian Optimization based on Gaussian process (GP) regression is well suited for such problems where GPs give reliable predictions with very few measurements while Bayesian Optimization makes a judicious exploration versus exploitation trade-off. (Our previous work on this problem has been published and can be found here (http://arxiv.org/abs/1510.03592) for further details.)
Scope and Objectives:
– Study the state-of-the-art in radio based device localization.
– We have already developed a scheme where we assume a zero mean GP prior, and now we would like to explore whether or not there is much to be gained with a non-zero mean GP prior.
– The project would involve simulating the non-zero mean GP prior followed by its real-time implementation on the UAV hardware, resulting in a live demonstration.
Prerequisites:
Knowledge of probability theory, programming skills.
Knowledge of machine learning techniques is a plus.
Laboratory: LCM, IC faculty
Project available for: Master in Communication Systems / Computer Science
Number of students: 1
Supervisor:
Akshat Dewan,
Professor:
Bixio Rimoldi, tel: 32679, office INR 111, [email protected]