Coordination Control of Multi-Agent Systems


Recent years have witnessed an increasing interest in decentralized control of multi-agent systems (MASs). Decentralized control is characterized by local interactions among agents where each agent exchanges information only with its neighbors. Based on these local interactions, a desired collective behavior of MASs is achieved. Examples are formation control, flocking, consensus control, etc. When compared with centralized control, decentralized control avoids a single point of failure which in turn increases robustness of MASs, allows for inexpensive and simple agents, and lowers the implementation cost. In addition, decentralized control scales better as the number of agents increases and is sometimes an intrinsic property of MASs. For instance, in industrial applications, we are facing coordination and cooperation of a number of small, inexpensive autonomous systems which replace complex large-scale integration devices. In natural exploration, disaster prevention and handling (firefighting, earthquake, mine clearance etc.) as well as in the service sector (healthcare of senior citizens), autonomous robots, as parts of distributed networks, can help in many ways, and even replace a human in complicated, dangerous and repetitive tasks.


Joint grant – PR China and Republic of Croatia



Principal investigators

Stjepan Bogdan and Ji Feng Zhang

More info

The problem of MAS formability is concerned with the existence and design of decentralized control laws that drive the MAS to a desired formation. Formability of a MAS depends on several key factors: agents dynamic structures, connectivity topology, properties of the desired formation and the admissible control set.
Our previous work is focused on formability of homogeneous (i.e.,identical) and heterogeneous linear time invariant (LTI) agents. Other important factors that impact MAS formability, but are often swept under the rug, involve realistic sensing and communication. Low-cost agents are equipped with off-the-shelf processing units, sensors and communication devices. Consequently, the influence of measurement noise, quantization, lossy communication channels, network throughput and limited sampling periods on MAS formability cannot be neglected. Detrimental aspects of realistic sensor and communication artifacts are usually addressed one (or several of them) at a time. The collaboration proposed herein aims at devising a comprehensive framework able to analyze all of these aspects simultaneously. To the best of our knowledge, the concurrent effects, that these detrimental aspects have on MAS formability, are not adequately studied in the literature. We plan to focus first on LTI-MASs with homogenous agents. Afterwards, heterogeneous LTI agents will be considered. If time permits, time-varying and nonlinear agents will be taken into account.