Distributed Cooperative Control of Multi-Agent Systems Under Detectability and Communication Constraints
Date of Award
Program or Major
Electrical and Computer Engineering
Doctor of Philosophy
Se Young Yoon
Cooperative control of multi-agent systems has recently gained widespread attention from the scientific communities due to numerous applications in areas such as the formation control in unmanned vehicles, cooperative attitude control of spacecrafts, clustering of micro-satellites, environmental monitoring and exploration by mobile sensor networks, etc. The primary goal of a cooperative control problem for multi-agent systems is to design a decentralized control algorithm for each agent, relying on the local coordination of their actions to exhibit a collective behavior. Common challenges encountered in the study of cooperative control problems are unavailable group-level information, and limited bandwidth of the shared communication. In this dissertation, we investigate one of such cooperative control problems, namely cooperative output regulation, under various local and global level constraints coming from physical and communication limitations.
The objective of the cooperative output regulation problem (CORP) for multi-agent systems is to design a distributed control strategy for the agents to synchronize their state with an external system, called the leader, in the presence of disturbance inputs. For the problem at hand, we additionally consider the scenario in which none of the agents can independently access the synchronization signal from their view of the leader, and therefore it is not possible for the agents to achieve the group objective by themselves unless they cooperate among members. To this end, we devise a novel distributed estimation algorithm to collectively gather the leader states under the discussed detectability constraint, and then use this estimation to synthesize a distributed control solution to the problem.
Next, we extend our results in CORP to the case with uncertain agent dynamics arising from modeling errors. In addition to the detectability constraint, we also assumed that the local regulated error signals are not available to the agents for feedback, and thus none of the agents have all the required measurements to independently synthesize a control solution. By combining the distributed observer and a control law based on the internal model principle for the agents, we offer a solution to the robust CORP under these added constraints.
In practical applications of multi-agent systems, it is difficult to consistently maintain a reliable communication between the agents. By considering such challenge in the communication, we study the CORP for the case when agents are connected through a time-varying communication topology. Due to the presence of the detectability constraint that none of the agents can independently access all the leader states at any switching instant, we devise a distributed estimation algorithm for the agents to collectively reconstruct the leader states. Then by using this estimation, a distributed dynamic control solution is offered to solve the CORP under the added communication constraint. Since the fixed communication network is a special case of this time-varying counterpart, the offered control solution can be viewed as a generalization of the former results.
For effective validation of previous theoretical results, we apply the control algorithms to a practical case study problem on synchronizing the position of networked motors under time-varying communication. Based on our experimental results, we also demonstrate the uniqueness of derived control solutions.
Another communication constraint affecting the cooperative control performance is the presence of network delays. To this regard, first we study the distributed state estimation problem of an autonomous plant by a network of observers under heterogeneous time-invariant delays and then extend to the time-varying counterpart. With the use of a low gain based estimation technique, we derive a sufficient stability condition in terms of the upper bound of the low gain parameter or the time delay to guarantee the convergence of estimation errors. Additionally, when the plant measurements are subject to bounded disturbances, we find that that the local estimation errors also remain bounded. Lastly, by using this estimation, we present a distributed control solution for a leader-follower synchronization problem of a multi-agent system.
Next, we present another case study concerning a synchronization control problem of a group of distributed generators in an islanded microgrid under unknown time-varying latency. Similar to the case of delayed communication in aforementioned works, we offer a low gain based distributed control protocol to synchronize the terminal voltage and inverter operating frequency.
Basu, Himadri, "Distributed Cooperative Control of Multi-Agent Systems Under Detectability and Communication Constraints" (2020). Doctoral Dissertations. 2527.