Diistributed cognitive coexistence of 802.15.4 with 802.11 Sofie Pollin 13 Mustafa Ergen 2, Michael Timmers 13, Antoine Dejonghe 1, Liesbet Van der Perre , Francky Catthoor 1 I3 Ingrid Moerman L4 Ahmad Bahai 2 1 Interuniversity Micro-Electronics Center (IMEC); E-mail: {pollins}@imec.be 2 University of Califiornia Berkeley; 3 Katholieke Universiteit Leuven; 4 Universiteit Gent 802 11 2.412GHz 5MH 22MHz Abstract-Thanks to recent advances in wireless technology, a broad range of standards are currently emerging. Interoperabil- ity and coexistence between these heterogeneous networks are 24-2483GH ' " "' becoming key issues, which require new adaptation strategies ISM Band to avoid harmful interference. In this paper, we focus on the coexistence of 802.11 Wireless LAN and 802.15.4 sensor networks in the ISM band. Those networks have very different transmission characteristics that result in asymmetric interfer- 802.15.4 2.405GHz 5MHH 2T Hz ence patterns. We propose distributed adaptation strategies for 802.15.4 nodes, to minimize the impact of the 802.11 interference. Fig. 1. 802.11 and. 802.15.4 channels in the 2.4GHz ISM band. This interference varies in time, frequency and space and the sensor nodes adapt by changing their frequency channel selection common requirement: both 802.15.4 and 802.11 devices are over time. Different distributed techniques are proposed, based battery-powered so that energy consumption is a major design on scanning (with increasing power cost) on the one hand, and criterion. Any algorithm for those networks should take the based on increased cognition through learning on the other hand. energy cost into account including the non-negligible hard- These techniques are evaluated both for performance and energy .energ cos into account aincldingithe nonegibehard- cost. We show that it is possible to achieve distributed frequency ware power contribution associated with idle mode operation, allocation approaches that result only in an increase of 20% of scanning and receive processing. the delay performance compared to ideal frequency allocation. The purpose is to introduce distributed algorithms to op- Moreover, it is shown that a factor of two in energy consumption timize the 802.15.4 performance under varying 802.11 in- can be saved by adding learning to the system. terference patterns. The sensor network is indeed critically I. INTRODUCTION affected by the coexistence, since it's output power is much Interest in wireless technology has experienced an explosive lower than that of 802.11 networks. The proposed algorithms growth over the last decades. The finalization of diverse should be fully distributed to improve scalability (since sensor standards has eased the development of wireless applications. networks are large), robustness (which is an important require- As a result, the spectrum is getting used by a variety of ment for sensor network applications) and adaptability. More heterogeneous devices, standards and applications. This is specifically, our goal is to design distributed channel selection especially the case for the Industrial, Scientific and Medical algorithms that allow the sensor nodes to dynamically adapt (ISM) bands that are unlicensed and hence host the most their channel in response to the 802.11 interference. The heterogeneous range of networks. energy cost of the proposed algorithms will always be taken In this paper we focus on the coexistence between two into account. major wireless standards that operate in the 2.4GHz ISM band, Traditional approaches for coexistence of wireless devices namely 802.11g Wireless LAN [1] and 802.15.4 Sensor Net- focus on transmit power control. In [4], for instance, the works [2]. Their overlapping frequency channels are shown in allowable transmit power is determined in order to guarantee a Fig. 1. The characteristics of both networks are very different, protected radius to primary users that should not be interfered resulting in a problem that is asymmetric in nature. Indeed, the with. This is especially useful to enable spectrum sharing output power of 802.15.4 devices is typically as low as OdBm between systems with different levels of regulatory status, e.g., 13], whereas the output power of 802.1 Ig devices is 15dBm or primar and secondar users. This type of sharing is tpi- above. Also, 802.15.4 sensor networks are designed to monitor cally referred to as vertical sharing, while horizontal sharing the environment or buildings, and can be very large, while considers systems with equal regulatory status. Concerning 802.1 1 networks are mostly local hotspots organized around an the latter horizontal sharing between homogeneous devices, Access Point (AP). Finally, sensor network applications are not game-theoretic concepts are used in 15] to achieve distributed demanding in terms of throughput, but however require a high transmit power allocations. These can however not be used reliabilLity and robustness against attacks or unknown events, wheln the operatinag conditions of the considered networks are They should also be self-organizing since it is ilmpossible yer asymmetric. to lmaintain such lLarge networks efficientlLy. In comparison, Another categoryof solutions focuses on dynamic frequency 802.1L 1 lnetworks are tpically used by a lLimited number of selectioln to avoid interferelnce. In this colntext, the specialL case throughput-intenlsive applications. There is in fact only one of yertical coexistenlce between 802.1L1 and 802.15.4 networks 1-4244-0381 -21061$20 00 ©2006 IEEE Authorized licensed use limited to: Univ of Calif Berkeley. Downloaded on March 11, 2009 at 20:48 from IEEE Xplore. Restrictions apply.