File pdf 3.0MB

Introduction to Multiagent Control and Open Problems Related to a Higher-Order Cascaded Integrator Consensus Algorithm

The control of several devices and systems to perform a group objective is considered multiagent control. It has many applications in unmanned aerial vehicles (UAVs), spacecraft, robotics, and vehicular systems. These strategies enable tasks such as flocking of swarms, group formation, synchronization, rendezvous, and consensus of systems. Distributed control is a strong way to scale these strategies to large systems in an adaptable and reconfigurable manner. Instead of having a “centralized brain” of the system, information locally available to each agent is used through cooperative control algorithms so that each agent works independently to achieve the group objective.

This paper will include a brief introduction to multiagent control and a brief evaluation of a control algorithm proposed by Ren et al. utilizing the state information of agents that may be modeled as a chain of integrators1. This agent model and control algorithm have several potential applications including the planar vertical takeoff and landing (PVTOL) of aircraft and Segway vehicles. Specifically, the control algorithm has the potential to control the takeoffs and landings of swarms of fixed-wing aircrafts and the motion of swarms of Segways.

We present an open problem in the resilience of the consensus algorithms proposed for higher-order integrator agents and demonstrate that they are susceptible to attacks from adversaries. One class of attacks on multiagent systems consists of hijacked agents that attempt to subvert the group objective by acting abnormally or communicating false and even harmful information. Proper resilience to hijacked agents requires that the other agents of the system are still able to achieve the desired task. The addition of resilience would improve the safety and security of the control protocol in real-world applications.