I. Introduction
The Internet-of-Things (IoT) consists of interconnected edge devices, such as sensors, wearables, etc., that gather and process data and that communicate with each other. There are many application areas where IoT technology is used, such as smart homes, e-health, smart logistics, smart mobility, etc. Energy consumption is one of the most important constraints for these devices. Since wireless sensor networks (WSN) are typically deployed in areas where electrical energy is not available, they are usually powered by means of batteries. However, the battery lifetime limitations pose an important problem, e.g., in hard-to-reach or hostile areas or in massive-scale deployments. Moreover, these batteries are often bulky and contain hazardous chemicals. To cope with these problems, battery-less devices using energy harvesting seem to be a possible alternative. These devices capture energy from the environment, accumulate it and convert it into electrical energy. For an overview of energy harvesting in wireless sensor networks (EH-WSN), we refer to e.g., [1], 2], [3]. In such a device, the energy is harvested from the environment (solar, thermal, wind, hydro, etc.) and stored in a capacitor. During the transmission and the reception of packets or other communication related functions, the radio uses the stored energy. Packets can only be transmitted or received if the device has harvested enough energy and reaches a certain threshold. Below this threshold, the radio is put asleep and energy is harvested. The aim of this paper is to propose a generic model for a class of MAC protocols that satisfy a number of characteristics with respect to (i) the origin of the packets that need to be transmitted, (ii) the instant at which transmission of a packet starts, (iii) the receiver-initiated MAC protocol that is used. The impact of important system parameters on the occupancy and delay are investigated, in particular the balance between receiving and transmitting packets. As the aim is not to evaluate a specific MAC protocol, but rather to present a generic queueing model that allows to compute the occupancy and delay in a node, no comparison with simulation or measurements is given. In Section III, a more detailed description of the system is given. Section IV presents a Markov chain model for this system. The system occupancy at an arbitrary time instant and the average response time are determined. Numerical results illustrating the possible application of the model are given in Section V. Finally, conclusions are drawn in Section VI.