This thesis claims the following major contributions to the field of distributed con- trol of networked multi-agent Euler-Lagrange systems:
1. The first main contribution of this thesis is systematic distributed controller synthesis for networked multi-agent EL systems. Unlike most of the con- troller design approaches in the literature for networked-multi-agent EL sys- tems (cf. [62, 63, 64, 71, 156] and references therein), which only rely on stability analysis, in this thesis we provide a performance index for each EL system. This performance index is a suitable measure for expected perfor- mance of each EL system in the network. By minimizing this performance index we obtain a distributed controller for each agent that guarantees global
state synchronization and set-point tracking for EL systems in the network. 2. The second main contribution of this thesis is considering parameter uncer-
tainty in networked EL systems and switchings in the communication net- work topology. Both of these problems have been considered separately in the literature, however, since in practice these two challenges may occur con- currently, new stability analysis based on switched systems theory is neces- sary to demonstrate stability of the entire EL systems network.
3. The third main contribution of this thesis is in the design of controller recon- figuration strategy for multi-agent EL systems subject to both intermittent and permanent additive actuator faults. Reconfigurable controller is a chal- lenging problem which is being only considered for single EL systems in the literature. This problem for multi-agent EL systems with switchings in the communication network topology is considered for the first time.
4. The fourth contribution of this thesis is in the development of distributed control of networked EL systems in presence of parameter uncertainties and external disturbances based on H∞optimal control approach. Specifically, in this thesis for the first time in the literature we employ H∞synthesis to design the state synchronization and set-point tracking for networked EL systems. We also discuss input-to-state stability (ISS) of the networked EL systems for the first time in the literature.
5. The fifth contribution of this thesis is in the analysis and prediction of the performance of the networked EL systems in presence of actuator faults when the developed H∞controller is utilized. The following realistic scenarios are considered in our analysis: (1) fault detection imperfection, that is when fault
is not detected by the FDI algorithm, (2) fault isolation imperfection, that is when the fault is detected in the wrong channel or in the wrong agent, and (3)
fault identification imperfection, that is when the fault estimation is not exact.
6. The sixth contribution of this thesis is in the development of controller recon- figuration strategy to improve the performance of our proposed H∞controller in presence of fault identification imperfection. To the best of our knowledge this realistic problem has not been considered before for single and multi- agent EL systems in the literature.
7. The seventh contribution of this thesis is in the development of distributed control algorithms for multi-agent EL systems in presence of both actuator saturation constraints and in absence of velocity measurements. We have shown in our simulations that our proposed controllers outperform the exist- ing distributed constrained controllers for networked EL systems.
8. The eighth contribution of this thesis is in the development of controller re- configuration strategy to compensate for the effects of actuator faults which results in a change in the available maximum control effort.
9. The ninth contribution of this thesis is in the extension and modification of our proposed constrained velocity-free controllers for EL systems to attitude control of SC formation flying missions. It is shown in the simulations that our proposed controller outperforms similar controllers in the literature.
Chapter 2
Background, Preliminaries and
Definitions
2.1
Multi-Agent Systems
Multi-agent systems, in this thesis, refers to a network of multiple dynamical sys- tems. The dynamics of these systems is not coupled individually, however, the system’s dynamic states are coupled through a common (shared) control law. In this thesis we consider a class of nonlinear systems, which are known as Euler- Lagrange (EL) systems. The coupling among the agents of the network can be considered for all the states or only for parts of the states, e.g. outputs of the agents. The control law, which constitutes the connection among the agents in the network can be classified into three general types. The first type of the controllers, which are known as centralized controllers (e.g. Multi-Input Multi-Output controllers in [18]), receive the information from all of the agents in the network. The control command, in this case, is generated by a central controller and transmitted to all the agents in the network. The second type of controllers, known as decentralized
controllers [157], receive information only from the agent. In this type the commu- nication among the controllers is not preplanned (or predefined) and the commu- nication among the agents can be considered as an stochastic process. Therefore, the controller must make decisions based on the agent’s own states. The third type of controllers, known as distributed controllers [12], receive information from the associated agent and the neighboring agents. The controller command is then gen- erated based on the information the controller receives. Therefore, the controller needs to know the neighboring agent’s states and must have a general knowledge of the network structure.