Uav swarms: Models and effective interfaces

Abstract

This chapter examines modeling and the design of effective control interfaces, for human operators, of unmanned aerial vehicle (UAV) swarms. The swarm is modeled as a two-component hierarchical system consisting of the interaction dynamics among the UAVs in the swarm, referred to as the network dynamics, and the UAV dynamics itself. Human operators are assumed to be able to interface with the swarm via the network dynamics, which in turn has adopted a leader-follower consensus model. The system-theoretic and topological features of the network dynamics are then examined in order to design effective mechanisms for interfacing with the swarm. Specifically, the open loop H2 norm of the network is selected as a performance metric for reasoning about effective human-swarm interaction. The role of topological features of the network is highlighted in the context of this metric and is related through the effective resistance of the corresponding electrical network. This is then followed by exploiting such topological features for designing a network rewiring protocol to maximize the H2 norm. These topology design tools are applied to wind-gust rejection in disturbed swarming scenarios, demonstrating the viability of topology-assisted design for improved swarm performance. A network-based model reduction is also proposed to form a lower-order model of the network which is easier for the human operators to conceptualize and manage. The reduction process involves a novel partitioning scheme, dubbed leader partition, in order to fuse “similar” states in the UAV network and to form a graph-theoretic method for model reduction. This model reduction technique is then applied to derive improved swarming performance in the presence of wind gusts.