Optimization of multiple access in the energy harvesting wireless sensor network with delivery deadline constraint

With the wide application of the energy harvesting wireless sensor network (WSN) in many real-time communication scenarios, such as environmental monitoring, industrial automation and battlefield surveillance, the multiple access of such WSN needs to take into account both the delivery deadline constraint of data packets and the energy harvesting dynamics of sensor nodes.Due to the inherent decoupling of interference, delivery urgency and remaining energy, the design and optimization of such multiple access are more challenging than that of traditional multiple access that only needs to take into account the packet traffic pattern.A centralized access scheme was designed with the access actions relying on the global knowledge of current delivery urgency and remaining energy.And then, to avoid the costly overhead in the centralized access, a decentralized access scheme was designed with the access probabilities merely relying on the local knowledge of delivery urgency and remaining energy.Under the objective of maximizing the network throughput, the centralized access schemes were formulated with complete and simplified knowledge as two Markov decision processes (MDPs), respectively, and the backward induction algorithm was used to obtain optimal centralized policies for these MDPs.Furthermore, the decentralized access was formulated with simplified knowledge as a decentralized MDP, and the Markov policy search was used to propose an ε-optimal decentralized policy.Simulations under a wide range of network configurations were provided to verify the effectiveness of the simplified modeling and demonstrate the performance advantage of the proposed polices.