Control Policy Optimization for Data Harvesting in a Wireless Sensor Network

We consider the problem of finding an optimal trajectory of a mobile agent tasked with harvesting data from multiple sensors nodes in a wireless sensor network. We describe the data transmission model using a free-space broadcast communication scheme and formulate an optimal control problem to extract the data from all the nodes in minimal time. In a one-dimensional mission space, we show that the optimal motion policy can be described in parametric form. Under such a policy, the agent motion is defined by a sequence of positions where the agent will dwell for some amount of time before proceeding at full speed to the next location in the sequence. Using Infinitesimal Perturbation Analysis to estimate the gradients of the resulting hybrid system, we apply a gradient-descent based scheme to find locally optimal parameters. Additionally, we develop globally optimal policies under certain special conditions. The results are demonstrated through several simulations.