| Popis: |
The size of a system is fundamental information for effective and proper operation of many distributed algorithms. This paper addresses the average consensus algorithm for distributed system estimation, or more specifically, a comparative study of its five frequently applied weights. We use four different methodologies in order to find the most performing weights in terms of the asymptotic convergence factor, the per-step convergence factor, their associated convergence times, the mean square error and the convergence rate over 30 random geometric graphs with varying leaders and either with or without bounded execution. |
