1: \begin{abstract}

2: %%%%%

3: We consider a linear consensus system with~$n$ agents

4: that can switch between~$r$ different   connectivity

5: patterns.

6: A natural question is which switching law yields the best (or worst)

7: possible speed of convergence to consensus?

8: We formulate this question in a rigorous manner by relaxing

9: the switched system into a bilinear consensus control system,

10: with the control playing the role of the switching law.

11: A best (or worst) possible switching law then corresponds to

12: an optimal control.

13: We derive a necessary condition for optimality,

14: stated in the form of a maximum principle~(MP).

15: Our approach, combined with suitable algorithms for numerically

16: solving optimal control problems, may be used to obtain explicit

17: lower and upper bounds on the achievable rate  of convergence to consensus.

18:  We also show that the system will converge to consensus for any switching law

19:  if and only if a certain~$(n-1)$ dimensional linear switched system converges to the origin

20:  for any switching law. For the case~$n=3$ and~$r=2$, this yields

21:  a necessary and sufficient condition for convergence to consensus that

22:  admits a simple graph-theoretic interpretation.

23: %%%%%

24: \end{abstract}