| dc.description.abstract | Missions such as forest fire monitoring, military surveillance and infrastructure monitoring are referred to as persistent monitoring missions. These missions rely heavily on continual data collection from various locations, referred to as targets. In this dissertation, we consider a framework in which data is collected from the targets with the aid of unmanned aerial vehicles (UAVs). A UAV makes a physical visit to the targets for data collection, and immediately transmits the collected data to a base station for further analysis. Typically, the duration of these monitoring missions is long, and the monitoring vehicles are required to stay in flight for extended periods of time. Therefore, the batteries powering the UAVs must be recharged regularly at a recharging station/depot. From utilitarian and economic points of view, an efficient execution of these missions calls for two requisites: 1) minimizing the time delay between successive data collections at targets; 2) maximizing the total charge/energy drawn from batteries. The maximum time delay between successive data collections at any target is characterized by a function referred to as the walk revisit time, or simply the revisit time. Given a set of targets and a UAV tasked with monitoring the targets, the charge capacity of the battery powering the UAV can be surrogated by the number of visits the UAV can make to the targets without requiring a recharge. To minimize the wastage of energy resources, a charge penalty is imposed on the visits that are unutilized before each recharge. The aim of this work is to find optimal routes for the UAV(s) to visit the targets such that the sum of the revisit time and the charge penalty is minimized. The optimal route planning problem is determined by a number of factors such as the number of UAVs used for monitoring, the aerial platform on which the monitoring UAVs are built, the location of their depots, relative importance of the targets being monitored, etc. In this dissertation, we focus on equally weighted targets and address four different variants of the problem, all of which are computationally extremely challenging.
The variants considered are the following: 1) single UAV with no motion constraints and the depot located at one of the targets; 2) single UAV with curvature constraints on its path and the depot located at one of the targets; 3) single UAV with no motion constraints and its depot stationed at a location different from that of the targets; 4) multiple UAVs with no motion constraints with their depots located at the targets. This dissertation builds on the results of Variant 1; specifically, the characterization of the optimal solutions proved in this dissertation is the main contribution of this dissertation; it lends itself to a new formulation of the same problem that results in significant computational savings. The structural characterization also holds for Variant 2. Inspired by this result, conjectures are provided for the structure of optimal solution for variant 3 and is backed up by extensive numerical simulations. Variant 3 can also be perceived as a special case of targets with different weights/priorities, and therefore, the results developed in this dissertation can potentially be extended to solve a few special cases of the general problem involving arbitrarily weighted targets. | en |
