Robust and efficient swarm communication topologies for hostile environments

Vipul Mann,Abhishek Sivaram,Laya Das,Venkat Venkatasubramanian
DOI: https://doi.org/10.1016/j.swevo.2021.100848
IF: 10.267
2021-04-01
Swarm and Evolutionary Computation
Abstract:<p>Swarm Intelligence-based optimization techniques rely strongly on information communicated among agents during the search. These algorithms are generally used for solving challenging problems where the search function landscape, typically highly nonlinear, is not adequately known a priori. There are multiple local optima that could result in premature convergence for many algorithms. Furthermore, in certain applications, the search environment is hostile in the sense that the agents, and/or their communication channels, can die during the search, thereby adversely impacting the search performance. Such a loss changes the communication topology of the agents and hence the information available to them, ultimately influencing the performance of the algorithm. Here, we present a study of the impact of loss of agents on performance as a function of the initial network configuration. We use particle swarm optimization for an objective function with multiple sub-optimal regions in a hostile environment and study its performance for a range of network topologies with loss of agents. The results reveal interesting trade-offs between efficiency, robustness, and overall performance for different topologies. These can be analyzed to discover general features of networks that maximize performance. We observe that small-world networks seem to perform well under hostile conditions.</p>
computer science, artificial intelligence, theory & methods
What problem does this paper attempt to address?