Event-triggered Synchronous Distributed Model Predictive Control for Multi-agent Systems
- PDF / 444,933 Bytes
- 10 Pages / 594.77 x 793.026 pts Page_size
- 4 Downloads / 173 Views
ISSN:1598-6446 eISSN:2005-4092 http://www.springer.com/12555
Event-triggered Synchronous Distributed Model Predictive Control for Multi-agent Systems Xiaoming Tang*, Mengyue Li, Shanbi Wei, and Baocang Ding Abstract: An event-triggered distributed model predictive control (DMPC) approach for a type of dynamically decoupled, independently constrained systems with a coupled performance objective, is presented. The approach employs, for each agent, a compatibility constraint (in the spirit of Dunbar and Murray) in the optimization problem. An event-triggering condition, based-on the overall stability condition of the system, is developed. If the triggering condition for an agent is satisfied, then the agent solves its optimization problem; otherwise, then the agent retain feasibility and stability by simply adopting the tail of its previous solutions. A simulation example is provided to illustrate the effectiveness of the provided approach. Keywords: Compatibility constraint, event-triggered control, feasibility, model predictive control, multi-agent system.
1.
INTRODUCTION
For a multi-agent system, the distributed control procedure, due to its flexibility and computational efficiency, outweighs the centralized one [1–7]. In the distributed control technique, the control problem of the global system for interactive agents is decomposed into a number of local control problems; the control input for each agent is computed by using both local measurements and information received from its neighboring agents. Model predictive control (MPC), as an efficient approach for handling the system with physical constraints, has gained notable attentions for the past few decades [8–13]. The distributed MPC (DMPC), which handles the multi-variable control problem for multi-agent system satisfying the physical constraints, has been widely studied in the literature [14–17]. The agents may work sequentially or synchronously, and exchange information iteratively or non-iteratively. In [18], a DMPC approach is proposed, where the agents solve, in a synchronous order, their optimization problems. In [19], the agents solve, in asynchronous and specified order, their optimization problems. In [20], the communications among agents are considered, where each agent communicates, iteratively in one sampling period, with its neighbours, so as to yield a convergent solution. In [21], the information is transmitted in a non-iterative way. Usually, as compared with non-iterative
solutions, iterative ones involve much heavier computational burden. In the real applications, there are only limited communication and computation resources, so the present paper chooses the synchronous DMPC framework. The aforementioned literatures are, mainly, based-on the time-triggered mechanism, i.e., the local DMPC optimization problem for each agent is solved at each sampling time; this may bring heavy communication and computational burdens. In order to alleviate the burdens, the event-triggered mechanism has been invented in time [22–26]. The so-called event-
Data Loading...
