A Neighbor Discovery Protocol for Directional Antenna ... - IEEE Xplore

3 downloads 2 Views 3MB Size Report
Guangyu Pei , Marcelo M. Albuquerque, Jae H. Kim, Douglas P. Nast and Paul R. Norris. Boeing Phantom Works. P.O. Box 3707, MC 7L-49, Seattle, WA 98124- ...

A Neighbor Discovery Protocol for Directional Antenna Networks Guangyu Pei , Marcelo M. Albuquerque, Jae H. Kim, Douglas P. Nast and Paul R. Norris

Boeing Phantom Works P.O. Box 3707, MC 7L-49, Seattle, WA 98124-2207 {guangyu.pei; marcelo.m.albuquerque; jae.h.kim; douglas.p.nast,paul.r.norris} boeing.com ABSTRACT

We present a novel neighbor discovery protocol based on synchronized search that uses in-band signal. The transmission power is carefully controlled during the neighbor discovery process to minimize the unnecessary high transmission power exposure time, allowing for a better low probability of detection (LPD) performance. The underlying MAC protocol is TDMA and configurable control slots are used for the neighbor discovery process. The protocol allows us to easily control the trade-off between network entry time and channel overhead. When compared with existing directional protocols, our protocol exhibits many unique features such as LPD, low channel overhead and convergence time gradually increasing with distance. We then analyze the protocol via a simplified theoretical model. Finally, we implement the protocol in simulation and evaluate its performance against the theoretical results. Our analysis shows that the proposed approach is an effective neighbor discovery protocol for closer neighbors, while providing reasonable network entry times for distant nodes. We also present various solutions to reduce the network entry time for these distant nodes.

I. INTRODUCTION Global network centric operations are the key elements of providing battle space awareness to support information superiority in future military missions. Directional antennas such as phased array antennas (PAA) are an enabling technology of

future combat systems. These antennas can focus radiation energy in a narrow angle to form pointto-point wireless links, which have many attractive properties such as high data rate, long communica-

tion range, anti-jamming (AJ), low probability of exploitation (LPE), low probability of interception (LPI), etc. These directional links are an ideal technology for tier 2 backbone networks. However, directionality of the physical layer and other security requirements such as LPD make the neighbor discovery process very challenging. Unlike omni-directional radio networks, where the neighbor discovery is trivial since every node within the range is considered a potential neighbor, the directional antenna network requires the neighbor discovery process to establish a network. This is because the transceivers must align their antennas in a synchronized fashion in order to discover a neighbor at the maximum achievable distance. Even if two nodes are within range, they cannot discover each other if their antennas do not align. At the same time, LPD requires the transmission power to be kept to a minimum. The early work on neighbor discovery can be viewed as part of MAC protocols and be categorized in the following two broad categories: (1) random access based approach; and (2) synchronized search based schemes. Since the IEEE802.11 MAC protocol is very popular in the MANET research community, several approaches have been proposed for directional wireless links by modifying the IEEE802.11 MAC protocol [1], [2]. These protocols are examples of random access based approaches. They use the Directional Virtual Carrier Sensing (DVCS) concept that extends the IEEE 802.11 Distributed Coordinated Function (DCF) to directional wireless networks. However, these protocols require the receiver in omni-mode to receive RTS control packets. Thus, the maximum range within which a node can discover a neighbor is much less than the range with an approach where both transmitter and receiver

t Corresponding author

1 of 6

use the maximum gain. Furthermore, these protocols assume short transmission ranges (in the order of hundreds to thousands of meters) and thus the prorogation delay is small. As the ranges scale up to tens and hundreds of nautical miles, the propagation delay becomes large and these protocols yield inefficiency. Finally, these protocols continuously use maximum transmission power to reach all possible neighbors, which is not desirable for LPD. [3] proposed probabilistic neighbor discovery algorithms which are examples of synchronous search (these algorithms can also be extended to asynchronous networks). However, these algorithms have several performance issues. First, they assume the nodes are static, which is not a realistic assumption. The nodes with directional links are usually tier-2 backbone nodes such as an aerial node. The reason for this assumption is that each node randomly picks a direction for transmission in a slot and it assumes that it can cover a node with fixed probability, which is proportional to the beamwidth. Second, the analysis assumes 2-D rather than 3-D (i.e., 42t sphere). It will take much longer time to converge due to random selection of the directions in a 3-D environment. Moreover, for directional transmitters and directional receivers, it takes even longer, which hinders its ability to scale up in discovery range. Finally, the optimal transmission probability is inversely proportional to the beamwidth. In other words, it requires a lot more transmissions with narrower beams. This is not a desired property. In order to increase the coverage area, narrower beams are needed to achieve high antenna gains. Thus, in this case, these algorithms require more transmissions and thus increase the probability of detection. In this paper, we propose a new synchronized neighbor discovery protocol that overcomes the limitations of existing protocols. It is designed to take full advantage of the PAA technology. The goal is to provide bounded discovery time for nodes that are very far apart (on the order of tens to a couple hundred nautical miles) while favoring fast discovery for nearest nodes. The rest of the paper is organized as follows. First we present our protocol in section 11. We then present the analysis in section III followed by the performance evaluation in section IV. Section V concludes this paper.

II. PROTOCOL DESCRIPTION Our protocol is based on synchronized scanning search. It has the following properties: (1) Nodes discover their nearest neighbor first; (2) Nodes should be able to discover neighbors at distances that can be reached via highest antenna gains by both transmitter and receiver; (3) Low probability of detection. The RF energy of a neighbor discovery packet received at the maximum desired range should not be more detectable than the side-lobe energy in a directional data communication burst. To achieve this, the transmitter should not use narrow beam with high gains in order to limit the EIRP. Since the EIRP of the transmitting antenna operating in wide beam mode with a low antenna gain, we use high processing gain for the neighbor discovery announcement packets to compensate for the antenna gain loss with wide beam. At the same time, the receiver operates with high gain narrow reception beams. We assume each node is equipped with GPS and INS. The network is synchronized. Each TDMA frame has several dedicated control slot pairs for neighbor discovery purpose. The first slot in each pair is used by an announcer for transmitting an announcement packet and the corresponding second slot is used by the responder to send a response directionally. Each pair represents one scanning direction in the 3-D space. The two slots in a slot pair are not required to be consecutive. Figure 1 illustrates the main procedure of the neighbor discovery protocol. Essentially, the scan search process is organized into multiple rounds with two loops (line 4 and 7). The outer loop increases the transmission power levels and the inner loop scans through the 421 sphere. Unlike any existing approach in which each node uses maximal power at all times, we gradually increase the power levels. This provides several advantages: (1) it reduces the RF exposure and thus lowers the probability of detection; (2) it favors the nearest neighbor and provides robust neighbor discovery; (3) it is highly adaptive for LPD. The number of power levels can be adjusted based on the characteristics of the operational environment. Another important feature that sets apart our approach from existing approaches is the combination of high processing gain, wide beam transmission and

2 of 6

highly directional reception. This allows for discovery ranges equivalent to those observed when using directional transmitter and receiver. For a slot pair (x,y), each node randomly and independently decides whether it transmits an announcement packet (AP) in slot x with wide beam, or receives directionally with narrow beam on a lock-step scanning direction (line 9, 10, 19). If a node transmits an AP in slot x, it will always attempt to receive a response in slot y. If a node receives an AP in slot x, it will then transmit a response directionally in slot y. This completes the two-way handshake process and two nodes can then start the link establishment and maintenance process. Note that we only present the baseline protocol here. Gossip techniques [3] can certainly improve the performance by including multiple known neighbors' location information in the announcement packet. The receiver can cache this information and respond accordingly with a directional transmission of the response packet. The detailed discussions of these extensions are outside the scope of this paper.

III. ANALYSIS OF THE PROTOCOL In this section, we present the analysis of the protocol. The goal of the analysis is first to obtain insights of the performance and second to provide comparison basis for the simulation validation. Let us define the following:

1: while (true) 2: /* start a new cycle */ Tx Power = Minimum power; 3: 4: while (Tx Power

Suggest Documents