Random Access in IRS-assisted 802.11 Networks

Most works on intelligent reconfigurable surface (IRS)-assisted networks consider link-level simulations and do not address the upper protocol layers. This limitation becomes particularly relevant in practical systems like 802.11 Wi-Fi, where channel access is based on listen-before-talk (LBT) scheme, making conventional IRS scheduling strategies (time-division or centralized coordination) difficult to apply. Given these reasons, we shift the focus to random channel access based on carrier sense multiple access with collision avoidance (CSMA/CA) to explore how an IRS can be beneficial in such scenarios. We first discover that improving the SNR of links from stations (STAs) to an access point (AP) introduces the potential risk of hidden terminals among the STAs in the uplink. To analyze this, we introduce ns3IRS, a framework that integrates a model of an IRS into the ns-3 network simulator, which allows us to run the full WiFi stack within an IRS-assisted wireless network. We propose two solutions to mitigate the hidden terminal problem caused by the IRS also by assuring mutual carrier sensing: (i) splitting a single centralized IRS, or (ii) using additional small IRSs nearby the STAs – to create direct links among them, with the purpose of enabling mutual carrier sensing. Results show that these two solutions can improve the throughput compared to the traditional approach where virtual carrier sensing, i.e., CSMA/CA with Request-to-Send/Clear-to-Send (RTS/CTS), is used. For instance, for a system with four users, the proposed approaches improve the throughput from 25Mbit/s to around 35Mbit/s compared to the baseline scenario with RTS/CTS. Additionally, latency and jitter are decreased. However, as the number of users increases, achieving performance gains requires a larger IRS, since the gain scales with the number of reflecting elements, especially for the centralized IRS. In scenarios where scaling the IRS is impractical, CSMA/CA with RTS/CTS may become a more effective alternative.