Long Distance Wi-Fi using Rotating Antennas

1. INTRODUCTION

In this paper, the challenges faced in long distance Wi-Fi communication using rotating antennas are presented. Wi-Fi networks can provide connectivity at longer distances with considerably less cost compared to other approaches [1]. It is well understood that wireless networks are inherently different from wired networks in that they are probabilistic in nature. This implies that the protocols have to cater for the possibility that communication may not occur due to numerous factors; for example, the 802.11 MAC layer has to make sure that the packets are received at the receiver in the exact order to allow transport layers to actually communicate meaningfully [2]. These challenges in Wi-Fi communications are intensified over long distances. The factors that induce loss in Wi-Fi communication increase drastically over longer distances and so meaningful communication becomes all the more difficult. Further, if rotating antennae are used for Wi-Fi communication over long distances, the problem increases. This project aims to understand the basics of Wi-Fi and its limitations through various experiments. The limitations and overheads introduced by long distances and rotating antennas are discussed and initial experiments carried out to gauge the problem. The project also proposes solutions and suggests experiments that can test the benefit of proposed solutions. This is a work in progress and the final goal is the development of an efficient communication system by means of Wi-Fi over long distances using rotating antennas [3].

2. Problem Motivation

Rural and remote areas are generally isolated from the urban centres as far as internet connectivity is concerned. This creates a huge technological gap and communities living in those areas are not able to access all the resources available on the internet. Also, they are deprived of an economical and easy source of communication. Therefore, it is of paramount importance that rural and remote areas be provided with internet connectivity systems. However, the traditional approaches that cover large areas and provide connectivity, such as cellular services, WI-MAX and satellite etc. are too expensive. The initial costs of installing the infrastructure in rural and remote areas are very high and/or not viable due to sparse population distribution [4]. Moreover, because everyday service charges are extremely high, low income households are unable afford such services. For these reasons the obvious choice to be considered is Wi-Fi. The installation costs of Wi-Fi equipment are much lower compared to other technologies. Low-cost commodity hardware means everyone can afford their own communication stations such as computers and mobile phones, etc. Also Wi-Fi utilizes an unlicensed spectrum; therefore there are no service charges to be given to any cellular or other provider and internet service is extremely economical compared to other communication services.

Therefore, the most viable option is to provide connectivity to rural and remote areas in the form of Wi-Fi. However, there are fundamental shortcomings with the 802.11 MAC protocol − it is not designed for communication over long distances, which is the basic characteristic of remote areas. In order to further reduce costs, rotating antennas can be used at a base station; this can act as an ‘access point’ and can communicate with the target areas where the connectivity is provided, referred to as ‘clients’. This added feature of rotating antennas over long distances further illuminates the shortcomings of Wi-Fi protocol for such a setting. Furthermore, loss differentiation becomes extremely complex in this situation [4].

The aim of this project is to fully understand the workings and limitations of the 802.11 Wi-Fi protocol and how it is designed to accomplish certain specific goals. The project further adds long distances and rotating antennas and argues about the different set of requirements for such a setting and how the Wi-Fi MAC protocol is insufficient for this. The project proposes changes to 802.11 Wi-Fi MAC protocol and discusses general design requirements for this project.

3. Proposed Solutions and Discussion

3.1 IEEE 802.11 Wi-Fi MAC Protocol

The current Wi-Fi MAC protocol has some fundamental shortcomings. However, before the associated technicalities and shortcomings are discussed, it is important develop a complete understanding of the current Wi-Fi protocol. For this reason, experiments were carried out at a test bed on Emulab (www.emulab.net) where multiple nodes were used to carry out different runs. In each experiment, a certain parameter was changed and experiments were repeated 50 times.

Figure 1 shows a plot of bit rate versus the throughput experienced. The results show that as bit rate is increased, the throughput also increases until a certain point, when it drops drastically. This is because there is a possibility that the bit rate which gave very low throughput was already being used by some other nodes to communicate. This would mean that the interference would be high and throughput would decrease. It is also worth noting that if there are no external losses, the throughput would still only be 70% of the available bit rate due to the overheads introduced by the MAC layer. This is discussed in detail in [3].

Figure 1. Bit Rate v Throughput

Further experiments of throughput against transmission power also provide a thorough understanding of the dynamics of wireless communication. Figure 2 shows how increasing the transmission power increases the throughput. This is expected since larger transmission power means that even after attenuation, the received signals are fit for communication. As a side note, high transmission power also means that the transmission will interfere with other transmissions and may cause loss in them.

Figure 2. Transmission Power v Throughput

The same results can be observed if the plot of throughput against packet size is analysed. As packet size increases, the throughput increases as well because the effects of MAC overheads are minimized. However, after a certain point, throughput starts to decrease because excessively large packets introduce delays for subsequent packets, and if error occurs in the packet due to loss, the entire packet has to be resent. Since the packet size is large the retransmissions take more time so throughput decreases. Figure 3 shows these results.

Wi-Fi MAC protocol is not suited for long distance communication. It requires independent ACKs utilization for a stop-and-wait link recovery mechanism, which induces an underutilized channel when used over long distances. This can be changed by using bulk acknowledgments, as discussed in [2]. This is also now a part of the newer 802.11 protocols.

Figure 3. Packet Size v Throughput

The CSMA/CA is not suitable for such a scenario as large propagation delays suggest that the transmitter can have no real knowledge of the link and collisions can occur. This can be changed by a time multiplexed approach. This makes intuitive sense because a rotating antenna is inherently time multiplexed as it will be facing any one client at any particular time. The connection between the two will be short since the rotation will lead the antenna to face other way. This implies that the time multiplexed approach should be part of the new MAC protocol. Unnecessary collisions would be avoided as each client would only communicate in its own time slot. This would require an understanding between the server and the client as to when the slot started and when it ended.

The loss characteristics of Wi-Fi over long distances using rotating antennas are far more complex. External Wi-Fi interferences can lead to losses due to the hidden terminal problem. Multiple-link interferences can also degrade throughput due to the overlapping of channels or use of the same channel. Long distances lead to loss of connectivity so MAC layer throughout decreases. Rotating antennas introduce the problem of connection disassociation in every rotation. This forces the current MAC protocol to choose sub-optimal bit rates that further reduce throughput. This has to be changed in the MAC protocol; it should be designed to use the best possible bit rate so that throughout can be high in the already limited time slot available for communication.

The current Wi-Fi protocol does not take into account rapid re-establishment of the connection in every turn of the rotating antenna. This implies that if the connection is not gracefully suspended and resumed at every rotation, valuable time would be wasted in trying to set up a connection in every rotation. This would further reduce the throughput.

3.2 Rotation of the Antenna

An important aspect of the project is how the antenna will be rotated. There are a number of design questions that need to be considered in terms of their practical implications. First of all, the speed of rotation needs to be optimal to ensure meaningful communication at every rotation. Also the amount of throughput should be high in comparison with the connection re-establishment overhead; therefore the speed cannot be set too high. On the other hand, if the speed is too slow the clients would experience large delays because the antenna would be facing away for a large amount of time. In order to minimize latency, it is recommended that the rotation be on the higher side. However, to maximize throughout, the rotation should be adjusted on the lower side. This is a general understanding, and further experiments need to be carried out to establish this relationship. Therefore, a balance has to be reached where latency can be minimized and throughput can be increased / maintained.

As far as the antenna rotation is concerned, another important factor is the path of rotation. It must be decided whether the antenna will rotate in full circles irrespective of the position of the clients or whether it will only rotate in an arc as long as the clients are covered. The other design limitations to be considered are the deceleration and acceleration of the antenna as it moves in a certain arc. This may lead to excessive mechanical wear and tear in the antenna and related equipment because stopping and moving the other way can induce mechanical malfunction.

4. Related Work

Recent work on long-distance Wi-Fi include [1] comparing mesh networks to WiLD networks and the advantages of mesh networks over WiLD settings. The paper focused on these issues (such as hidden nodes and SNR) and the suggestions were very helpful in offering a basic understanding of the project. Another related work is that of [2] who propose several solutions for establishing multihop point-to-point links. The use of bulk acknowledgements with a sliding window helps to cater for large distance propagation delays and the subsequent high loss variability. They use TDMA coupled with an FEC approach to do channel optimization and error correction. Another prior work is of [3] where several 802.11n/ac features such as frame aggregation, channel bonding and MCS values were considered for high performance in real-time applications. Here, the researcher takes leverage of all the previous work completed with regards to long-distance WiFi and plans to use these suggestions once rotating-antenna connections are well understood and MAC-level implementations are focused upon.

5. Experiments and Evaluation

The researchers carried out experiments on the rooftops of the University Library building. The distance between the two points was more than 300m. Figure 4 shows the experimental setup.

Figure 4. Experimental setup

The results obtained through sniffing for packets on the client size had a lot of data. These results needed to be carefully analysed for the presence of Wi-Fi MAC protocol control frames, such as null frames, probe requests and responses, etc. These control frames allowed better understanding of how Wi-Fi can be operated at long distances by using rotating antennas and further provided insight into how to a better design the MAC protocol for this situation. The research undertaken during this study had limited resources and therefore optimal levels of experimental setups might not have been accomplished.

However, the researchers did manage to obtain useful initial information which can explain the behaviour of Wi-Fi over long distances using rotating antennas. The research used Wireshark to sniff packets and extracted useful information such as the RSSI from them. Figure 5 shows the plots of RSSI against time which reflects a form of periodic graph. This is as expected since the rotating antenna produces a periodic trend.

Figure 5. RSSI v Time

Compressive analyses of the above graph as presented in Figure 6 (zoomed version) clearly shows that RSSI values are improved when the antenna faces towards the client and are too low when the antenna faces away. The amount of time the RSSI values are high is comparatively small with respect to the amount of time which has low RSSI values. This makes sense since the antenna spends less time directly facing the client and more time while rotating away from the client.

Figure 6. RSSI v time – zoomed in

This understanding is further strengthened if the graph of a CDF of RSSI values of all packets is analysed, as it shows that the percentage of packets that have a comparatively high RSSI value, i.e. greater than -80 dBm. Figure 7 shows that this percentage is less than 20%, which is a direct indicator of the amount of time the antenna faces the client.

Fig. 7. CDF of RSSI values

These results are from initial experiments and further experiments will allow an even better understanding.

6. Conclusion

Research has been conducted to develop an understanding of the basic factors involved in determining loss-differentiation in Wi-Fi. Hence a detailed study of Wi-Fi and its characteristics was completed before initiating the experimental setup. The research included a review of several Wi-Fi characteristics by conducting several Emulab experiments by changing packet size, bit rate and txpower in order to observe changes in throughput. Experiments were then conducted for long-distances on campus by simulating large distances by using two distant buildings. The server was run on the rotating antenna and the client was run on a simple directional antenna and dumps were collected. From the results obtained, periodic increase in RSSI values characterized to rotating antennas were observed, and the observations were predictable. It was also observed that large loss / variability in packets were a result of MAC inefficiencies which the researcher plans to address in the future. The MAC protocol needs to be aware of the rotation of the antenna and needs to avoid sub-optimal bit-rate selection. Re-establishments must be timely predicted by the rotating antenna and should be characterized properly.

7. References

• Subramanian, S. Surana, R. Patra, S. Nedevschi, M. Ho, E. Brewer, and A. Sheth. Rethinking Wireless for the Developing World. In Fifth Workshop on Hot Topics in Networks (HotNets-V), Irvine, CA, USA, November 2006

• Patra, S. Nedevschi, S. Surana, A. Sheth, L. Subramanian, and E. Brewer, Wildnet: Design and implementation of high performancewifi based long distance networks, in NSDI 2007

• Nishat, R.Anwar, A.Mehfooz, B.Zaidi, H.Choudhary, I.Qazi, On the Effectiveness of High-Speed WLAN Standards for Long Distance Communication, in INFOCOM 2014

• Hassiani and D. Gan, Relevence of WI-FI networks developments’, vol. 1, pp. 28-38, 2011.

• Pfurtscheller and C. Neuper, ‘Imagery and direct brain-computer communication’, Proc. IEEE, vol. 89, no. 7, pp. 1123-1134, 2001.

• de Vries and T. Mulder, ‘Motor imagery and stroke rehabilitation: a critical discussion’, Journal of Rehabilitation Medicine, vol. 39, no. 1, pp. 5-13, 2007.

• Szafir, ‘Non-Invasive BCI through EEG An Exploration of the Utilization of Electroencephalography to Create Thought-Based Brain-Computer Interface’, Thesis, Boston College, 2009…

• [G. Townsend, B. Graimann and G. Pfurtscheller, ‘Continuous EEG Classification, 2004

• Choudhury, T. Charman, V. Bird and S. Blakemore, ‘Adolescent development of motor imagery in a visually guided pointing task’, Consciousness and Cognition, vol. 16, no. 4, pp. 886-896, 2007.

• Pires, M. Torres, N. Casaleiro, U. Nunes and M. Castelo-Branco, ‘Playing Tetris with non-invasive BCI’, 2011 IEEE 1st International Conference on Serious Games and Applications for Health (SeGAH), 2011.