Adaptive Diffusion for Distributed Optimization over Multi-Agent Network with Random Delays (Poster)

Optimizing an aggregate function over a multi-agent network based on diffusion strategies calls for node collaborations, where each sensor exchanges information with their predefined neighbors within its transmission/reception range, and then combines them linearly with fixed and non-adaptive scalar weights to obtain a consensus solution. However, the transmission/reception process may be corrupted by random delays due to imperfect network environment. This paper proposes a distributed method to obtain the adaptive weights considering the occurrence of random one-step delays, which are depicted by Bernoulli random variables with known probability distributions. Then, an adaptive diffusion algorithm for optimizing the aggregate function is presented. Finally, numerical simulations are provided for validating the proposed method.