Triggered Gradient Tracking for asynchronous distributed optimization

This paper proposes Asynchronous Triggered Gradient Tracking , i.e., a distributed optimization algorithm to solve consensus over networks with asynchronous communication. As building block, we devise the continuous-time counterpart of recently proposed (discrete-time) gradient tracking called Continuous . By using Lyapunov approach, prove exponential stability equilibrium corresponding agents’ estimates being consensual optimal solution, arbitrary initialization local estimates. Then, propose two triggered versions algorithm. In first one, agents continuously integrate their dynamics and exchange neighbors current variables in synchronous way. totally scheme which each agent sends its based on triggering condition that depends locally verifiable condition. The protocol preserves linear convergence avoids Zeno behavior, an infinite number events finite interval time is excluded. analysis as preparatory result, show point holds for both algorithms any estimate initialization. Finally, simulations validate effectiveness methods data analytics problem, showing also improved performance terms inter-agent