Consensus-based cooperative control with applications to robotic and communication systems

In networked multi-agent systems, consensus refers to steering a specific variable of all agents to a common value across the network using local information exchange among neighboring agents. The consensus algorithm is one of the most promising enablers for the distributed group cooperation techniques widely studied in robotic and communication systems. How to design robust consensus algorithms for the networked uncertain multi-agent systems under practical communication channels is a fundamental problem for consensus applications. This dissertation pursues the stability and convergence properties of the robust consensus algorithm and studies three sub-problems: uncertain robotic network synchronization under directed communication topologies, distributed cooperative spectrum sensing under fading channels, and distributed radio environment mapping (REM) with unknown system dynamics.;For distributed robotic networks, this dissertation proposes a backstepping-based synchronization design for uncertain multi-agent systems under directed communication topologies. With the aid of the graph Laplacian matrix spectral analysis and the adaptive redesign technique, the proposed backstepping design achieves distributed synchronization of networked uncertain Lagrangian systems under general directed graphs, which encompasses existing double integrator consensus as a special case. Using distributed consensus estimator, the proposed design further extends to robust output consensus tracking of networked higher-order strictly feedback systems with mismatched model uncertainties.;For communication systems, this dissertation studies cooperative spectrum sensing and REM tracking for distributed cognitive radio networks. The proposed spectrum sensing scheme adopts a weighted consensus algorithm to achieve weighted combining of the network measurements using local communications. The main benefit of the proposed design is distributed weight setting according to measurement qualities of local measurements leads to global weighted soft measurement combining, which outweighs existing average consensus-based schemes. The convergence of the proposed algorithm is proved under communication channels with temporary link failures, and the detection performance with fading effects is comparable to the equivalent centralized methods. For REM tracking problem, a consensus estimator-based Kalman filter is developed to track the unknown temporal dynamics of the REM using distributed Expectation Maximization algorithm. The dynamic RF propagation heat map is recovered and visualized in simulation, which validates the effectiveness of the proposed distributed REM tracking scheme.