Examination Committee

Prof Roger CHENG, ECE/HKUST (Chairperson)
Prof Vincent LAU, ECE/HKUST (Thesis Supervisor)
Prof Ling SHI, ECE/HKUST

Abstract

Wi-Fi coverage characterization or surveying is curial at access points (APs) deployment stage. Technician has to deploy APs according to the building’s floor plan, as the distribution of APs and location of obstacles determine the wireless radio (RF) coverage. Once the APs have deployed, technician requires to analysis the coverage of existing wireless network and adjust the location of APs accordingly. 

Site surveying is a task of measuring Received signal strength (RSS) from APs at different locations, by using these measurements to predict the coverage of the floor plan. This is time consuming and labor intensive; hence motivate us to propose using robot to automate the surveying task. Intuitively, the locations of RSS measurements scatter spread widely over the site should benefit for the prediction accuracy compare to very concentrated, conversely robot consumes less energy to travel for the latter case. Therefore the robot should have a strategic way to measure RSS at different locations, which balance between accuracy and energy. In this work, we formulate the problem as MDP which use fisher information and action cost as optimisation quantity then solve by approximation technique to tackle the dimensionality issues. Results show that there is a positive performance gain (accuracy and energy aspect) compares to random action or random walk strategy (which usually humans do) to measure RSS.