| Removing humans from dangerous situation by shifting them to a supervisory role has existed for decades. Oil and gas refineries are beginning to shift to this line of thought, as equipment is monitored electronically; however, accidents still occur when operators must physically verify alerts before actions can be taken. A mobile robotic system is a suitable analog for this process. The operator can remotely perform inspection tasks from an operator control station through a mobile robotic system; however, communication between the operator and the robot is paramount. If communication is ever lost, the human operator will be exposed to the dangers of the environment. In an autonomous system, localization of the mobile robot is key - if the operator tells the robot to move from location A to location B, the robot needs to know exactly where it is at all times to avoid causing damage to the environment or itself. The work in this thesis focuses on the WiFi communication and localization of a mobile robot. First, extensive experiments are conducted to understand the relationship of received signal strength, bandwidth, link quality, and distance for both indoor and outdoor environments in a 2.4 GHz WiFi network. Findings from these empirical studies are then used to determine both single and k-coverage of a given area. Single-coverage is required to ensure that at every point in the region of interest, communication can occur between the mobile robotic system and the operator control station. Coverage is then expanded to k-coverage to provide a more robust network for localization. Algorithms are implemented to determine a minimal 3-coverage deployment that ensures a minimum threshold distance between neighboring access points. Channel allocation is determined through a graph coloring approach where two heuristics are implemented and their results are compared. WiFi localization is implemented through RSSI fingerprinting, a matching heuristic, where a new approach is considered for determining the k-closest neighbors. The results from WiFi localization are then fused with dead reckoning, and a fiducial marker system using an extended Kalman filter and a validation gate. An accuracy of 0.43 m is achieved with the hybrid localization technique. |
