Topological and MAI Constraints on the Performance of Wireless CDMA Sensor Networks Swades De, Chunming Qiao, Dimitris A. Pados, and Mainak Chatterjee Abstract— In this paper, we characterize analytically the mul- tiaccess interference (MAI) in wireless CDMA sensor networks with uniformly random distributed nodes and study the trade- off between interference and connectivity. To provide a guideline for improving system behavior, three competitive deterministic topologies are evaluated along with the random topology in terms of link-level and network-level (routing) performance. The impact of the signature code length and the receiver design on network performance for different topologies is also studied. Keywords – Wireless sensor networks, ad hoc networks, code-division-multiple-access (CDMA), network connectiv- ity, network topology, interference suppression, spreading signatures, throughput I. I NTRODUCTION AND MOTIVATION We consider a wireless sensor network which may contain thousands of tiny low-cost sensors scattered over a region of interest [1],[2]. The sensors (also called nodes) are battery operated (i.e., energy constrained), have limited memory and processing power, and form a randomly connected ad hoc network. Since the nodes are mostly stationary, their location information may be obtained via GPS or other means [3]-[6] to assist efficient information processing (e.g., data aggregation) and distributed routing. Since the sensors have limited energy, buffer space and other resources, contention-based protocols based, for ex- ample, on the 802.11 direct-sequence spread-spectrum (DS- SS) technique may not be a suitable option. Here, as an alternative, we suggest the use of several codes (signatures) that can be allocated to different nodes with possible code re- use between spatially separated nodes as in cellular CDMA systems. The use of multiple properly designed codes will reduce the channel access conflict but at the expense of multiaccess interference (MAI) which is absent in 802.11 DS-SS systems. It is well-known that MAI is a key factor in determining the performance (e.g., throughput) of a CDMA network [7],[8]. Even if each node transmits at the lowest possible power to its intended receiver, random distribution of active (i.e., transmit- ting/receiving) nodes, lack of coordination (e.g., reservation Swades De is with the Dept. of Electrical Eng., State University of New York at Buffalo; email: swadesd@eng.buffalo.edu. Chunming Qiao is with the Dept. of Computer Sc. and Eng., State University of New York at Buffalo; email: qiao@cse.buffalo.edu. Dimitris A. Pados is with the Dept. of Electrical Eng., State University of New York at Buffalo; email: pa- dos@eng.buffalo.edu. Mainak Chatterjee is with the Dept. of Electrical and Computer Eng., University of Central Florida; email: mainak@cs.ucf.edu. [9] or request/acknowledgment [10] based transmission), and decentralized control will result in a significant amount of interference power from neighboring nodes. The interference problem becomes more severe as the node density increases, although a higher node density might otherwise help improve connectivity and network performance. Liu and Asada [11] addressed the MAI issue in DS- CDMA based sensor networks by attempting to minimize MAI through minimum energy channel coding and on-off keying data transmission. In this approach, to deal with higher number of users, MAI is controlled by increasing the number of redundant bits (that is, lowering the channel code rate) at the cost of reduced information transmission rate. In [12], Dousse et al. assumed a TDMA-based channel access scheme on top of the CDMA codes to address the interference-related connectivity problem in a large ad hoc network. Muqattash and Krunz [13] proposed a controlled access CDMA protocol for wireless ad hoc networks where out-of-band RTS-CTS (request-to-send/clear-to-send) packets are used to determine MAI before a data packet transmis- sion and then adjust the transmission power accordingly. In terms of network connectivity, Bettstetter [14] studied the relationship between k-connectivity and node density for a uniformly random node distribution, where a graph is said to be k-connected (k ≥ 1) if for each node pair there exist at least k mutually independent paths connecting them. Yet, along with these studies, one also needs to consider the physical and medium-access-control (MAC) layer constraints so that for a given receiver structure and for an acceptable packet error performance, an optimum network topology can be determined (or for a given network topology an optimum receiver can be designed to achieve a target error performance). In this work, we first characterize theoretically the MAI in a network with a random topology and study the associated trade-off with k-connectivity. Instead of controlling MAI at the channel code design level, we propose to control MAI to a certain extent by proper placement/activation of nodes for a given signature code set and receiver structure, while maintaining the desired graph connectivity. To this end, we study the link-level (i.e., single-hop) bit error rate (BER) performance of different network topologies, such as random, hexagonal, square grid, and triangular, using conventional matched-filter (MF) as well as MAI suppressive minimum-mean-square-error (MMSE) receivers [16] under Gold [17] or other recently identified minimum total-squared- 0-7803-8356-7/04/$20.00 (C) 2004 IEEE IEEE INFOCOM 2004