Research On Key Technologies Of Energy Harvesting Wireless Sensor Networks

Energy harvesting wireless sensor networks has the ability of perceiving, acquiring and cooperatively transmitting data of monitored objects using sensor nodes with energy harvesting capabilities. It can guarantee the nodes’energy is not depleted, effectively solving the energy-constrained problems in traditional wireless sensor networks. However, the process of energy harvesting is random, which makes existing advanced technology in traditional can not directly applied to energy harvesting wireless sensor networks. In this context, the key technologies of energy harvesting wireless sensor network are studied, focusing mainly on energy management technology and medium access control for energy harvesting wireless sensor networks.Firstly, the optimal energy management policy of limited battery capacity sensor node based on independent and identically distributed random energy harvesting model is studied. The problem is modeled as maximizing the amount of data correctly received during the transmitter’s life. Under the assumption of geometric distribution for the life of the node, the problem can be formulated as an infinite discounted Markov Decision Problems. Then based on the Bellman optimal criteria, we prove the optimal energy management policy is threshold policy depending on the length of the energy queue. Finally, through simulation on different channel conditions, we analyze the impact of varying battery capacity and probability of energy harvesting on the performance of the optimal strategy. Simulation results show that when the battery capacity is greater than a certain value, the optimal reward does not increase. In addition, the optimal reward improves with the increase of the probability of energy harvesting.Secondly, the problem of energy management strategies on Markov energy model is further studied. An adaptive energy management strategy based on probability of successful transmission with low storage capacity is proposed. Only if the probability of successful transmission on the current slot is greater than threshold of energy harvesting state of previous slot, the data can be transmitted; otherwise it discards data. Thresholds can be adopted in different energy harvesting status adaptively. Furthermore, we narrow feasible region of threshold from the real domain to discrete set with limited number, which makes the way of combining optimal equations and enumeration algorithm to obtain optimal threshold workable. Simulation results show that in different energy harvesting situations, compared to the performance of optimal strategy in policy iteration algorithm, adaptive energy management strategy based on probability of successful transmission is almost as better as performance of optimal strategy; at the same time, comparing to policy iteration algorithm which stores all states and corresponding actions, our proposed energy management policy just stores threshold which reduces the amount of storage. In addition, for the secene in which the statistical information of energy harvesting is difficult to predict, the performence of Q-learning with different learning rate is verified. Simulation results show that the Q-learning having same learning rate for all states and actions can learn the optimal policy quickly.Finally, fairness among sensor nodes on RF energy harvesting wireless sensor networks is studied. An adaptive energy harvesting medium access control is proposed based on the time slotted CSMA/CA. In the proposed algorithm, contention time of each sensor node is controlled only by its own energy harvesting rate. The contention time of nodes with lower energy harvesting rate is smaller to improve the throughput. Under the assumption that the packets arrival process is Poisson, simulation results shows that when the RF transmit power is below a certain threshold, fairness in our proposed medium access control is higher than time slotted CSMA/CA.