Self-Synchronized Duty-Cycling for Mobile Sensor Networks with Energy Harvesting Capabilities: A Swarm Intelligence Study Hugo Hern´ andez 1 , Christian Blum 1 , Martin Middendorf 2 , Kai Ramsch 2 and Alexander Scheidler 2 1 ALBCOM Research Group, Universitat Polit` ecnica de Catalunya, Barcelona, Spain {hhernandez,cblum}@lsi.upc.edu 2 Department of Computer Science, University of Leipzig, Leipzig, Germany {middendorf,kairamsch,scheidler}@informatik.uni-leipzig.de Abstract— When asked if ants rest or if they work untiringly all day long, most people would probably respond that they had no idea. In fact, when watching the bustling life of an ant hill it is hard to imagine that ants take a rest from now and then. How- ever, biologists discovered that ants rest quite a large fraction of their time. Surprisingly, not only single ants show alternate phases of resting and being active, but whole ant colonies exhibit synchronized activity phases that result from self-organization. Inspired by this self-synchronization behaviour of ant colonies, we develop a mechanism for self-synchronized duty-cycling in mobile sensor networks. In addition, we equip sensor nodes with energy harvesting capabilities such as, for example, solar cells. We show that the self-synchronization mechanism can be made adaptive depending on the available energy. I. I NTRODUCTION In contrary to the general belief that ants are always busy, different species of ants have been observed to spend a large portion of their time resting; see, for example, [17], [9], [7], [4]. For example, ants of the species L. acervorum rest about 72% of their time. Moreover, not only individual ants present patterns of alternate activation, but also whole colonies show synchronized patterns of activity ([8], [14], [3]). In addition, activity phases are not just synchronized, but self-synchronized because no external signal has been found as a possible cause of colony synchronization. Delgado and Sol´ e [5] modelled this behaviour of ant colonies by means of fluid neural networks. With their model they were able to show that synchronized activity pattern enable the colony to accomplish tasks more efficiently. In this work we use the self-synchronization behaviour of ant colonies for the development of a self-synchronized duty-cycling mechanism for mobile sensor networks. Sensor networks ([19], [20]) aim to monitor large areas and to analyze complex phenomena for extended periods of time. Recent hardware advances produced sensors for a wide range of physical data such as light intensity, humidity, temperature and the oxygen level, as well as for the characteristics of objects such as direction and speed. This means that sensor networks can be used for many different tasks such as environmental monitoring, patient monitoring in health care, industrial machinery surveillance, etc. Some of these applications require the nodes of a sensor network to be distributed within wide areas without power sources, as, for example, forests and seas. Moreover, sensor nodes might be mobile. Therefore, they are often equipped with batteries, which makes energy a scarce resource. Several approaches can be found in the literature for extending the lifetime of a sensor network that is subject to energy limitations. A rather recent approach is referred to as energy harvesting (see, for example, [15], [12], [11]). The idea is to transform light or vibrations into energy that can be used to recharge the batteries. However, energy harvesting alone might not be enough for obtaining a sufficiently long network life time. An approach that aims at saving energy is duty-cycling (see, for example, [2], [6]). Hereby, sensor nodes periodically switch between energy intensive states and low energy states. Nodes in energy intensive states can perform all normal duties of a sensor node, whereas nodes in low energy states are restricted to certain functions in order to save energy. Recently, some researchers made an attempt at combining dynamic duty-cycling with energy harvesting capabilities. Most of these works (see, for example, [10], [13]) require an apriori known energy profile. Such techniques require the energy source to exhibit little variations, which is, in many occasions, not very realistic. A recent work from Vigorito et al. [18] considers adaptive control techniques for adjusting duty-cycling without any apriori given energy profile. The main disadvantage of this approach is that it still considers each node separately and, although better performance and network lifetime is obtained, the lack of synchronizity between the nodes may imply that the network is restricted to certain applications. This is were the approach that we present in this paper comes into play. As mentioned before, our approach offers adaptive, self-synchronized duty- cycling. Finally, let us remark that the study presented in this work is done entirely from the swarm intelligence point of view. It can be seen as a first feasibility study. We are perfectly aware of the fact that the proposed mechanism might be required to change when being adapted to real sensor networks. The outline of this work is as follows. In Section 2 we shortly outline the model of ants’ self-synchronization be-