Experimental Evaluation of Reverse Direction Transmissions in WLAN Using the WARP Platform Raul Palacios a , Francesco Franch a , Francisco Vazquez-Gallego b , Jesus Alonso-Zarate b and Fabrizio Granelli a a DISI, University of Trento, Italy, {palaciostrujillo@disi, f.franch@studenti, granelli@disi}.unitn.it b Centre Tecnològic de Telecomunicacions de Catalunya (CTTC), Spain, {fvazquez, jesus.alonso}@cttc.es Abstract—This paper describes an experimental implementa- tion of a variation of the Reverse Direction (RD) Medium Access Control (MAC) Protocol (RDP) defined in the IEEE 802.11n using the Wireless Open-Access Research Platform (WARP). The proposed approach, named Bidirectional MAC (BidMAC), allows the receiver of a valid data sequence to perform an RD transmission to the transmitter without contending for the channel. Whereas in RDP the RD transmission must be initiated by the transmitter, in BidMAC it can be dynamically initiated by the receiver according to its traffic requirements. Previous results based on mathematical analyses and computer-based simulations have shown that BidMAC can better balance down- link and uplink transmission opportunities in a Wireless Local Area Network (WLAN) where the Access Point (AP) handles bidirectional data flows for some of its wireless stations (STAs). This paper aims at going one step further and demonstrating that such superior performance can be attained in real environments. Towards this end, an implementation of BidMAC has been carried out in a reference design of WARP compatible with the IEEE 802.11a/g and tested in a proof-of-concept network formed by an AP and two STAs. Experimental results confirm the superior performance of BidMAC when compared to the legacy Distributed Coordination Function (DCF) of the IEEE 802.11 versus the traffic load, packet length, and data rate, yielding gains of up to 60%. * I. I NTRODUCTION Recently, the use of Reverse Direction (RD) transmissions has been proposed in the IEEE 802.11 Standard to improve the throughput and energy efficiency of Wireless Local Area Networks (WLAN) based on the IEEE 802.11 [1]. More specifically, the Reverse Direction Protocol (RDP) has been defined in the IEEE 802.11n as a Medium Access Control (MAC) layer enhancement of the legacy Distributed Coordina- tion Function (DCF) to increase channel utilization. The RDP breaks with the basic operation of the DCF where a wireless station (STA) gains a Transmission Opportunity (TXOP) by competing to get access to the wireless channel in order to transmit data to one arbitrary destination. In RDP, the holder of a TXOP, once it has seized the channel, can allocate the unused TXOP duration to one or more receivers in order to allow data transmissions in the reverse link. For scenarios with bidirectional traffic, this approach is very convenient as it reduces contention in the wireless channel. The concept of reverse direction (or bidirectional) trans- mission in WLAN was first introduced by [2], prior to the * This work has been funded by the GREENET research project (PITN-GA- 2010-264759) and by the Generalitat de Catalunya through 2014-SGR-1551. standardization of the RDP. Since then, several works have proposed similar approaches with different purposes. Existing RD-based protocols can be classified into two categories: (i) proactive, i.e. RD exchange sequence initiated by the transmit- ter, or (ii) reactive, i.e. RD exchange sequence initiated by the receiver. Proactive RD protocols [3], [4] allow the transmitter to grant the receiver the remaining time of its TXOP for reverse data transfer, in a way similar to RDP. On the other hand, reactive RD protocols [2], [5]–[9] allow the receiver to reserve the wireless channel for a backward transmission by extending the transmitter’s TXOP time, without needing to compete for the channel. This sort of RD protocols can achieve higher performance in some scenarios because they are more adaptive to the actual needs of a network. In particular, the work in [7] and our previous works [8], [9] investigate the feasibility of reactive RD exchange operation in infrastructure WLAN, wherein an Access Point (AP) is connected to a cable network infrastructure and provides wireless Internet access for a number of STAs in its coverage area. Results show that reactive RD approaches can effectively address the unbalanced operation of DCF between uplink and downlink traffic when traffic flows are highly bidirectional. Indeed, DCF provides equal channel access opportunities for all STAs, including the AP. Therefore, the AP only receives an equal share of the wireless channel to deliver downlink traffic to all the STAs, while it has data to transmit to all of them. Note that we consider the case when all STAs route all their traffic through the AP. Therefore, by allowing the AP to dynamically initiate RD exchange sequences when receiving data from the STAs, uplink and downlink transmission oppor- tunities can be better balanced, hence improving the overall WLAN performance. The results presented in the works discussed above are based on theoretical analyses and computer-based simulations. Whereas theoretical models typically adopt simplified assump- tions for mathematical tractability, computer-based simulations usually lack PHY-layer modeling accuracy, thus leading to inaccurate results and conclusions. In contrast, real-world implementation can help reveal unexpected challenges to the development of new MAC protocols and also provide new insights in the operation of communication protocols. This is the main motivation for the work presented in this paper where we describe an experimental implementation of a reactive RD MAC protocol, named Bidirectional MAC (BidMAC) [8], [9], for infrastructure WLAN in real hardware. This implemen-