TY - JOUR
T1 - Asymptotic analysis for greedy initialization of threshold-based distributed optimization of persistent monitoring on graphs
AU - Welikala, Shirantha
AU - Cassandras, Christos G.
N1 - Publisher Copyright:
Copyright © 2020 The Authors. This is an open access article under the CC BY-NC-ND license
PY - 2020
Y1 - 2020
N2 - We consider the optimal multi-agent persistent monitoring problem defined for a team of agents on a set of nodes (targets) interconnected according to a fixed graph topology. The objective is to minimize a measure of mean overall node state uncertainty evaluated over a finite time interval. In prior work, a class of distributed threshold-based parametric controllers has been proposed where agent dwell times at nodes and transitions from one node to the next are controlled by enforcing thresholds on the respective node uncertainties. Under such a threshold policy, on-line gradient-based techniques are then used to determine optimal threshold values. However, due to the non-convexity of the problem, this approach leads to often poor local optima highly dependent on the initial thresholds used. To overcome this initialization challenge, in this paper, the asymptotic steady-state behavior of the agent-target system is extensively analyzed for a single-agent system and dense graphs. Based on the obtained theoretical results, a computationally efficient off-line greedy technique is developed to systematically generate initial thresholds. Extensive numerical results show that the initial thresholds obtained lead to significantly better results than the locally optimal solutions known to date.
AB - We consider the optimal multi-agent persistent monitoring problem defined for a team of agents on a set of nodes (targets) interconnected according to a fixed graph topology. The objective is to minimize a measure of mean overall node state uncertainty evaluated over a finite time interval. In prior work, a class of distributed threshold-based parametric controllers has been proposed where agent dwell times at nodes and transitions from one node to the next are controlled by enforcing thresholds on the respective node uncertainties. Under such a threshold policy, on-line gradient-based techniques are then used to determine optimal threshold values. However, due to the non-convexity of the problem, this approach leads to often poor local optima highly dependent on the initial thresholds used. To overcome this initialization challenge, in this paper, the asymptotic steady-state behavior of the agent-target system is extensively analyzed for a single-agent system and dense graphs. Based on the obtained theoretical results, a computationally efficient off-line greedy technique is developed to systematically generate initial thresholds. Extensive numerical results show that the initial thresholds obtained lead to significantly better results than the locally optimal solutions known to date.
UR - http://www.scopus.com/inward/record.url?scp=85093756390&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85093756390&partnerID=8YFLogxK
U2 - 10.1016/j.ifacol.2020.12.1670
DO - 10.1016/j.ifacol.2020.12.1670
M3 - Conference article
AN - SCOPUS:85093756390
VL - 53
SP - 3433
EP - 3438
JO - IFAC-PapersOnLine
JF - IFAC-PapersOnLine
IS - 2
T2 - 21st IFAC World Congress 2020
Y2 - 12 July 2020 through 17 July 2020
ER -