In this dissertation, we investigate the problem of estimation and control in multi-agent systems under limited sensing capabilities, where agents suffer from the lack of (relative) position information to their common target or neighboring agents. Du...
In this dissertation, we investigate the problem of estimation and control in multi-agent systems under limited sensing capabilities, where agents suffer from the lack of (relative) position information to their common target or neighboring agents. Due to the ambiguity caused by sensing limitations, it becomes essential to compensate for the missing information through data accumulation across the network. Specifically, we discuss how to recover the key information required to solve the target position estimation problem or the formation control problem, and under what conditions such information is retrievable. In particular, we address the following three problems: (i) estimating a target’s position in a global reference frame, (ii) estimating a target’s position in local reference frames, and (iii) displacement-based formation control.
The bearing-only target localization problem is addressed in both global and local reference frames. We first consider the scenario in which each agent knows its own position within a common reference frame but can only measure the bearing to a target. We then extend the analysis to the case where agents operate in local reference frames, with no access to a shared global frame. In both settings, we investigate localizability—that is, the conditions under which the target’s position can be uniquely determined—by analyzing the collective behavior of the proposed estimator when the diffusive coupling among agents is sufficiently strong. Our results show that localizability is governed by the stability of the blended dynamics, a reduced-order system that captures the overall behavior of the network. This characterization provides insight into the geometric configurations of agents that are necessary for successful target localization.
We further examine the problem of formation control using range-only measurements. In this scenario, agents aim to maintain a prescribed formation without any information about relative positions or orientations, where the desired shape is defined by inter-agent displacements. Instead of accessing relative positions, each agent measures only the range (distance) to its neighbors. The proposed solution involves a data-driven estimation algorithm that uses persistently exciting data between agent pairs to infer relative information. A discrete-time averaging consensus protocol is then employed to achieve distributed orientation estimation and formation control.