Mutual Aid Among Sensors: An Emergency Function for Sensor Networks

A large number of wireless sensor nodes in a certain area results in high contention. Inevitably, the transmissions of any possible critical data packets may fail due to collisions. In this letter, we introduce an aspect of human intelligence in wireless sensor networks, influenced by cooperative networking, which enhances the timely delivery of critical data. Mutual aid among sensors is an emergency out-of-the-box medium access control function for IEEE 802.15.4-2020. Specifically, the network coordinator detects critical data packets and sets an emergency flag to its next beacon to inform the nodes that they may overhear data packets. When a node overhears a critical data packet from a neighboring node, it switches to sleep mode and stays idle until the end of the superframe. Thus, interference is mitigated locally and temporarily. Simulation results, using the CC2650 radio parameters in OMNeT++, show that interference is reduced significantly, in favor of the timely delivery of critical data packets.


I. INTRODUCTION
Amy transmits critical data, and Jon stays silent instead of interrupting her. This may not be science fiction. Recently, the devastating fires in Australia and the COVID-19 outbreak have shifted our focus towards sensing [1]. On this occasion, we revisit interference mitigation in sensor networks with a new perspective: We argue that mutual aid deserves to be a built-in capability of sensors. Therefore, we propose a medium access control (MAC) function for IEEE 802. 15.4-2020 [2], based on the principle that each sensor node is willing to assist its neighbors.
The monitoring of an area is a common application in sensor networks. We place the sensors and expect them to transmit their data packets to the coordinator using random access, without any further configuration. However, interference increases as the number of nodes increases and the delivery of any critical data becomes uncertain. In the tradition of cooperative networking [3], we propose mutual aid among sensors (MAAS). First, the coordinator detects critical data packets and notifies the nodes that there is an emergency. Next, a node may overhear a critical data packet and switch into sleep mode to lower the level of interference. This moderate use of overhearing is feasible thanks to ultralow power consumption transceivers, such as CC2650 [4]. IEEE 802.15.4-2020 includes a modern MAC with several capabilities for multiple PHY. Apart from star topology, it supports peer to peer formations such as clusters. There is a distinction between full-function devices (FFDs) and reduced-function devices. The time is divided in superframes. A coordinator, essentially an FFD, transmits beacons periodically to inform the rest of the nodes about the upcoming superframe. A superframe may contain a contention access period (CAP), which supports slotted carrier-sense multiple access with collision avoidance (CSMA/CA) and a contention-free period with guaranteed time slots. Additionally, the standard provides a persistent CSMA/CA with priority channel access (PCA) to enhance the delivery of critical data. PCA is a backoff algorithm, where a node performs a persistent random backoff after a failed clear channel assessment (CCA), in order to gain access to the channel as soon as possible. Also, two modes of channel diversity are supported: time slotted channel hopping (TSCH) and deterministic and synchronous multichannel extension (DSME).
Interference is a well-known cause of signal degradation. There are quite a few approaches in the literature regarding interference mitigation. An early attempt to improve the timely delivery of critical data was [5], which added queuing strategies to CSMA/CA. The combination of scheduled and random access is desirable as well, see [6]. As described in [7], a way of mitigating collisions due to hidden nodes is clustering. Also, a collision avoidance algorithm was proposed in [8], which can be useful in mesh networks. Moreover, contention can be reduced using adaptive transmission power, see [9]. TSCH has been exploited in several works to date, such as [10] and [11]. DSME is also an active area; see [12]. Indeed, channel diversity is effective, but has a significant cost of complexity due to scheduling and channel hopping. Our goal is to present a simple out-of-the-box function, compatible with the standard.
The rest of this letter is structured as follows. In Section II, we present and discuss MAAS. Our system model is presented in Section III. In Section IV, we discuss the results of the simulations. Finally, Section V concludes this letter.

II. MUTUAL AID AMONG SENSORS
MAAS is an emergency MAC function for IEEE 802.15.4-2020, which reduces contention and enhances the timely delivery of critical data. As a MAC capability, it may be enabled or disabled adaptively in a beacon enabled superframe, as described in [13].
The coordinator transmits beacons periodically to synchronize the nodes and to inform them about the upcoming CAP. Amy, a node that has critical data, receives a beacon and transmits a critical data packet. The coordinator receives the critical data packet from Amy and sets an emergency flag to the following beacon. Jon, a node that does not have 2475-1472 © 2020 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See https://www.ieee.org/publications/rights/index.html for more information. critical data, receives the beacon and schedules a timer to switch to sleep mode and another one to enter CAP. Then, Jon starts overhearing.
In the meantime, Amy transmits a critical data packet again. Jon detects the critical data packet and cancels its CAP timer. Therefore, it switches to sleep mode and wakes up just on time to receive the next beacon. Jon may not detect any critical data. In this case, it switches to sleep mode for a while and enters CAP as scheduled. This flow of events is depicted in Fig. 1. MAAS is inspired by the cooperative methods in networking, based on overhearing and forwarding, and offers a simple and intuitive alternative: a node may provide help by sleeping instead of forwarding. This idea suits well many applications in sensor networks, particularly critical systems suffering from high contention. When a node does not have any critical data, it may be more useful to sleep instead of transmitting. It mitigates contention and it saves some energy as well. Each node is rewarded with some sleep time for its help.
The proposed protocol does not require any additional frame transmissions. Instead, we attach one field to the beacon frame and one to the data frame. In the case of the beacon frame it indicates whether there is an emergency or not, while in the case of the data frame it indicates whether the frame contains critical data or not. We can assume with confidence that this overhead is negligible.
The duration of a critical event is assumed to span across multiple superframes. This refers to the cases where the nodes sense and transmit continuously. In case of extremely bursty traffic (e.g., one critical packet), the notification of the nodes at the beginning of the following superframe would be too late, since the critical packet has been already transmitted successfully. However, we expect that MAAS will be beneficial in most cases.

III. SYSTEM AND CHANNEL MODEL
Our system model was implemented with Castalia [14], in OM-NeT++ [15]. Castalia is a modular and extensible framework, which allows us to test MAAS using its infrastructure.  The wireless channel module uses a lognormal model to calculate pathloss as a function of distance: PL(d ) dB = PL(d 0 ) + 10η log 10 (d/d 0 ) + X (μ, σ 2 ), where d is the distance, PL(d 0 ) = 55 dBm is the pathloss at distance d 0 = 1 m and η = 2.4 is the pathloss exponent. X is a normally distributed variable with μ = 0 and σ = 4 dB. Apart from interference, pathloss is also a cause of signal degradation. A harsh pathloss model (see Fig. 2) helps us draw realistic conclusions about the performance of MAAS. Pathloss affects every signal, so it mitigates interference as well.
The simulated radio module is CC2650, by Texas Instruments. Its power consumption is calculated using the provided values by the data sheet [16]. The level of interference is calculated using the default additive interference model of Castalia. The estimation of bit errors for each received signal takes place after the subtraction of the received power from each other node.
The MAC module with MAAS capability is based on the baseline IEEE 802.15.4-2020 implementation of Castalia. It supports beacon enabled superframes and a CAP with slotted CSMA/CA, in compliance with the standard. MAAS requires the scheduling of two special timers. Upon the reception of a beacon with an emergency flag, sleeping and CAP timers are scheduled as depicted in Fig. 3. Overhearing is limited to the first quarter of the superframe, sleeping takes place during the second quarter and CAP during the second half of the superframe. If Jon does not detect any critical data packets from Amy, it follows the predefined schedule: It switches to sleep mode and enters CAP in the second half of the superframe. Otherwise, if Jon detects a critical data packet from Amy, its CAP timer is canceled. Thus, it switches to sleep mode and stays in this state until the end of the superframe.

IV. SIMULATION RESULTS
We evaluate the baseline IEEE 802.15.4-2020 implementation and MAAS, using a variable number of nodes that have critical data and a variable number of total nodes. Our simulation represents a 5 s critical   Table 1 for more details. The provided results are the mean values of 1000 realizations. Fig. 4 depicts the received packets after one realization of the simulation using 20 nodes in total where one node has critical data, denoted in red. The area matches the typical range of a coordinator in low-power networks, about 10 m. Each node transmits data packets to the coordinator. At first glance, MAAS seems effective, since packet reception from the red node is increased.

event. See
Next, we evaluate MAAS using 20 nodes in total and a variable number of nodes that have critical data. In Fig. 5, failed packets due to interference are reduced significantly when a few nodes have critical data and increase as the number of nodes increases. Similarly, in Fig. 6, received critical packets are increased when a few nodes have critical    data and decrease as the number of nodes increases. Fig. 7 shows that the received data packets with latency below 100 ms are increased only when a few nodes have critical data. Specifically, there is a peak at three nodes, indicating that interference from nodes that do not have critical data is minimized. This result depends on the range of each node. Fig. 8 shows that energy consumption is not affected. In fact, it is slightly reduced due to the applied sleep time. The aforementioned results of MAAS converge with the baseline. Essentially, MAAS has identical results to the baseline when every node has critical data, since every node participates in CAP.
Scalability should be investigated as well. Figs. 9 and 10 show the failed packets due to interference and the received critical packets, respectively, when five nodes have critical data and the number of total nodes is variable, from five to 50. We notice that the performance of the baseline declines rapidly due to contention. In contrast, MAAS maintains a low number of failed packets and high packet reception,   even though the error bars indicate an increased level of uncertainty towards 50 nodes in total. Finally, Fig. 11 shows the received data packets with latency below 100 ms, when five nodes have critical data. Note that, in Fig. 7, the performance of MAAS when five nodes out of 20 have critical data is low. This is also confirmed in Fig. 11 for a variable number of total nodes. Despite this, the performance of MAAS is consistently better than the baseline.
Our results, using a variable number of nodes that have critical data and a variable number of total nodes, show that the proposed function performs well. One may argue that the nodes that do not have critical data are treated somewhat unfairly. However, this is not the case. In sensor networks, the nodes usually have common goals, and MAAS guarantees that each one can rely on its neighbors' help when needed. Everyone has the opportunity to choose whether to stay silent or not. In the long run, everyone wins.

V. CONCLUSION
In this letter, we introduced MAAS, an emergency function for interference mitigation in sensor networks, where the nodes may overhear critical data packets and switch temporarily to sleep mode. MAAS is compatible with IEEE 802.15.4-2020, it can be used as an out-of-the-box function and is very effective regarding timely delivery of critical data. It seems that artificial intelligence will be a key feature of future networks. We hope that mutual aid, a noble aspect of human intelligence, will be the first choice.