Sensing Assisted Communication for the Next Generation mmWave WLAN: System Simulation Perspective

Fig. 1 shows the scenarios of SAC in the next generation Wi-Fi including the living room and the street canyon, which are also the target simulation scenarios in the following sections. We illustrate advantages brought from SAC according to the example scenarios shown in Fig. 1. The living room represents the indoor scenario, while the street canyon represents the outdoor scenario. There are one AP and one target STA in both scenarios, and the STA keeps moving in the scenarios. In the living room scenario, the beam will be reflected by the ceiling, floor, wall and furniture. The direct path will be blocked by some furniture and human body. In the street canyon scenario, the beam will be reflected or blocked by the ground, trees, buildings on the street and pedestrians on the street. The AP and the target STA communicate by using mmWave beam link. The moving speed and the beam width can be changed. The detailed parameter settings will be described in the next subsection.

To deeply analyze the performance of SAC enabled next generation Wi-Fi, we deploy a series of simulations. We focus on two scenarios: living room and street canyon as shown in Fig. 1 suggested and described in the IEEE 802.11ay document. We established an integrated system level (protocol and signaling of the MAC and the upper layers) and link level (channel model and PHY abstraction) simulation platform towards millimeter wave Wi-Fi based on network simulator 3 (NS3), and integrate sensing into the simulation platform to evaluate the performance gain of the SAC. We use the ray tracing method to model the channel based on the layout of each scenario and the antenna pattern.

i) Simulation Results of the Living Room Scenario.

There are one AP and one STA in the living room, and the STA randomly moves in the room. It is worth noting that the change of relative position between nodes needs to be depicted in system simulation, whether the moving node is the transmitter, receiver or the blocker, the final effect is equivalent. There are $18$ beam sectors in the ${360^{o}}$ range. Both uplink and downlink possess traffic to transmit and have the same traffic rate. Unless otherwise specified, the frame aggregation size is 1. The modulation and coding scheme (MCS) index is 14 of IEEE 802.11ay. There are three schemes:

• •

  802.11ay with beamtracking indicates that the system strictly works according to the 802.11ay standard with the beamtracking function.

• •

  The ISAC scheme indicates the AP and the STA dynamically switch their $\left\langle{tx,rx}\right\rangle$ beam pair according to the location sensing results. The sensing accuracy error varies, which means the sensed locations are randomly distributed around the real location and the radius of the sensing accuracy error.

• •

  Theoretical optimum means the AP and the STA can always select the optimal $\left\langle{tx,rx}\right\rangle$ beam pair based on the real location without any overhead.

Fig. 3 shows the throughput performance comparison with traffic rate varying in the living room scenario. The value in the X-axis indicates the traffic rate for both uplink and downlink. It can be observed that the throughput achieved by all schemes increases first and then stabilizes with the increase of traffic rate. This is because the network has reached saturation from unsaturated state. It is worth noting that even if the sensing error reaches $40cm$ that is not a small value, the throughput of the ISAC scheme is still very close to the theoretical performance. The theoretical performance means the sender and the receiver can always use the theoretically optimal beam pair and, consequently, can obtain the optimal throughput. The throughput of the 802.11ay with beamtracking scheme is lower. This is because frequently beamtracking leads to considerable signaling overhead. Moreover, when the sensing error is very large, for example $80cm$ in Fig. 3, the performance of the ISAC scheme deteriorates since large sensing error leads to serious mismatch beam pair. Thus, on the one hand, the source of the gain of this simulation is the signaling overhead. On the other hand, the sensing accuracy is a parameter that significantly affects system performance and should therefore be as accurate as possible.

Fig. 4 shows the throughput performance comparison with frame aggregation size varying in the living room scenario. In this scenario, the traffic rate is $4,000Mbps$ for both uplink and downlink. It can be observed that the throughput achieved by all schemes increases first and then stabilizes with the increase of frame aggregation size. This is because frame aggregation can enhance the MAC efficiency and the network has reached saturation. It is worth noting that the throughput of the ISAC scheme is close to the theoretical performance. Even if the sensing error reaches $60cm$ that is a relatively large value in the living room scenario, the throughput of the ISAC scheme still outperforms the 802.11ay with beamtracking scheme because of significantly low protocol overhead.

Fig. 5 shows the throughput performance comparison with moving speed varying in the living room scenario. In this simulation, the traffic rate is unsaturated, $10Mbps$ for both uplink and downlink. It can be seen that with the increase of the moving speed, the throughput of the ISAC scheme is stable and close to $20Mbps$. This is because the ISAC scheme can serve almost all the packets in time. Considering that the traffic rate is $10Mbps$, the total traffic rate of the uplink and the downlink is $20Mbps$. It is interesting although the traffic rate is unsaturated, the throughput of the 802.11ay with beamtracking scheme decreases with the increase of the moving speed. The reason is that unsaturated traffic leads to large packet arrival interval and correspondingly large move distance, which causes beam link failure. When beam link failure occurs, the beamtracking becomes ineffective because preamble cannot be successfully received. Then, the AP and the STA have to wait for the next beacon interval to re-train the beamform link. This process takes a long time. However, the sender continuously tries to retransmit data. When the maximum retransmission time is reached, the packet (MSDU) will be dropped from the MAC queue.

Fig. 6 shows the per packet average delay performance comparison with moving speed varying in the living room scenario. It can be seen that with the increase of the moving speed. The delay of the ISAC scheme is quite stable. In the unsaturated scenario, almost every packet can be transmitted in time because the sender and the receiver always utilize the approximating optimal beam pair. However, the per packet average delay of the 802.11ay with beamtracking scheme keeps increasing with the increase of the moving speed. The beam link failure problem occurs results longer waiting time to re-train the beamform link. Low latency traffic or named real-time application is one important requirement in the future, the ISAC scheme can significantly guarantee the delay performance.

Fig. 7 shows the packet dropping rate performance comparison with moving speed varying in the living room scenario. The per packet dropping rate of the ISAC scheme increases slowly with the increase of the moving speed because of the sensing error. In this scenario, the traffic rate is small, almost every packet can be transmitted in time. However, the per packet dropping rate of the 802.11ay with beamtracking scheme is significantly larger than that of the ISAC scheme. Because this unsaturated scenario leads to beam link failure problem, which causes longer waiting time to re-train the beamform link. During the waiting time, the 802.11ay with beamtracking scheme still retransmits the packets and, finally, the maximum retransmission time is reached and the packets are dropped from the queue.

ii) Simulation Results of Street Canyon Scenario.

There are one AP and one STA in the street canyon, and the STA walks along the street. It is worth noting that the change of relative position between nodes needs to be depicted in system simulation, whether the moving node is the transmitter, receiver or the blocker, the final effect is equivalent. There are $18$ beam sectors in the ${360^{o}}$ range. Both uplink and downlink possess traffic to transmit and have the same traffic rate. Unless otherwise specified, the frame aggregation size is 1. The modulation and coding scheme (MCS) index is 14 of IEEE 802.11ay. There are three schemes:

• •

  802.11ay with beamtracking indicates that the system strictly works according to the 802.11ay standard with the beamtracking function.

• •

  The ISAC scheme indicates the AP and the STA dynamically switch their $\left\langle{tx,rx}\right\rangle$ beam pair according to the location sensing results. The sensing accuracy error varies, which means the sensed locations are randomly distributed around the real location and the radius of the sensing accuracy error.

• •

  Theoretical optimum means the AP and the STA can always select the optimal $\left\langle{tx,rx}\right\rangle$ beam pair based on the real location without any overhead.

Fig. 8 shows the throughput performance comparison with traffic rate varying in the street canyon scenario. The value in the X-axis indicates the traffic rate for both uplink and downlink. It can be observed that the throughput achieved by all schemes increases first and then stabilizes with the increase of traffic rate. This is because the network has reached saturation from unsaturated state. The throughput of the ISAC scheme outperforms the 802.11ay with beamtracking scheme. This is because frequently beamtracking leads to considerable signaling overhead. When the sensing error becomes larger, the performance gain will be reduced.

Fig. 9 shows the throughput performance comparison with moving speed varying in the street canyon scenario. In this simulation, the traffic rate is unsaturated, $10Mbps$. It can be seen that with the increase of the moving speed, the throughput of the ISAC scheme is stable and close to $20Mbps$. This is because the ISAC scheme can serve almost all the packets in time. Considering that the traffic rate is $10Mbps$, the total traffic rate of the uplink and the downlink is $20Mbps$. It is interesting although the traffic rate is unsaturated, the throughput of the 802.11ay with beamtracking scheme decreases with the increase of the moving speed. The reason is that unsaturated traffic leads to large packet arrival interval and correspondingly large move distance, which causes beam link failure. When beam link failure occurs, the beamtracking becomes ineffective because preamble cannot be successfully received. Then, the AP and the STA have to wait for the next beacon interval to re-train the beamform link. This process takes a long time. However, the sender continuously tries to retransmit data. When the maximum retransmission time is reached, the packet (MSDU) will be dropped from the MAC queue.

Fig. 10 shows the per packet average delay performance comparison with moving speed varying in the street canyon scenario. It can be seen that with the increase of the moving speed. The delay of the ISAC scheme is quite stable and very small. In the unsaturated scenario, almost every packet can be transmitted in time because the sender and the receiver always utilize the approximating optimal beam pair. However, the per packet average delay of the 802.11ay with beamtracking scheme sharply increases with the increase of the moving speed. The beam link failure problem occurs results longer waiting time to re-train the beamform link.

Fig. 11 shows the packet dropping rate performance comparison with moving speed varying in the street canyon scenario. The per packet dropping rate of the ISAC scheme increases slowly with the increase of the moving speed because of the sensing error. In this scenario, the traffic rate is small, almost every packet can be transmitted in time. However, the per packet dropping rate of the 802.11ay with beamtracking scheme is significantly larger than that of the ISAC scheme. In this figure, the packet dropping rate of the 802.11ay with beamtracking scheme even exceeds 40 percent. Because this unsaturated scenario leads to beam link failure problem, which causes longer waiting time to re-train the beamform link. During the waiting time, the 802.11ay with beamtracking scheme still retransmits the packets and, finally, the maximum retransmission time is reached and the packets are dropped from the queue.

Fig. 12 shows the gain of the per packet average delay performance comparison with the beam width varying in the street canyon scenario. This scenario is still unsaturated. When the beam width is wide, the gain is small. The reason can be explained as follows. In the outdoor scenario, wider beam width covers larger area, which makes the beamtracking easier to track the varying location. On the other hand, when the beam width is getting narrow, the gain is significantly getting larger. More narrow beam width makes the beamtracking more difficult to track the varying location. Considering that mmWave wireless network usually has narrow beam, the SAC is more applicable to mmWave wireless network rather than the sub-7GHz band.

The simulation results confirm the performance gain and influencing factors of sensing-assisted communication analyzed above. Tab. I summarize the performance evaluation insights from system simulation.