Research in LoRa: The Disaster Communication Networking Space
There has been a lot of interest in LoRa over the last few years in the IoT space for its long range, low power, and low cost of both hardware and electrical requirements. Platforms (like SigFox, The Things Network (TTN) and now the ClusterDuck Protocol), have made getting started with LoRa and IoT easy by providing wireless hardware and software infrastructure. We can easily connect our device to services we build ourselves. The benefits of LoRa intrigued the interest of people in the natural disaster sector. The natural disaster sector has been stuck with a problem: “How will we be able to communicate in the event of a disaster when mobile communication networks aren’t available?” And they believe that LoRa might be a viable solution. Current research is being done to leverage LoRa technology solve this problem. We will explore this subject using a few research papers that propose network architectures pursuant to this problem.
LoRa Modules, Meshed
Let’s start with how LoRa’s characteristics and how it can be used in a mesh network. In our first paper, Pham, Kisel, et al [1]. examine LoRa’s capabilities in mesh networks using OMNET++, a popular piece of network simulation software.
LoRa modules can be configured using a few key parameters, which are spreading factor (SF), coding rate (CR), and bandwidth (BW). Explained nicely by Josef Matondang, these three components can be used to determine the data rate (DR), given by this equation:
Noted by the authors, high BW increases the data rate, but decreases coding sensitivity. High SF increases decoding sensitivity but decreases the data rate. The typical LoRa mesh is going to look like any other wireless mesh you’ve encountered. The authors give this figure to illustrate.
Routing selection is done via Received Signal Strength Indicator (RSSI), a common method of metric for measuring the strength of a signal. Here it’s being used to determine the cost of a link between two nodes, which can be used to determine the total cost of any route. The algorithm always chooses the link with the lowest cost.
When simulating the network, a propagation model that considers an urban setting where devices have structures in between them. Payload is generated randomly and there is a gateway node in the center of the network. For completeness, you can see the simulation parameters here in one of the paper’s tables.
Now let’s look at their results. Below are Figures 3 and 4 graphing the influence of load intensity and number of relay nodes respectively, on delivery latency.
High spreading factor on increasing loads leads to considerable jumps in latency but can be counteracted by increases in bandwidth. This tracks with the original equation that model’s data rate. Interestingly latency numbers seem to converge on SF = 9, BW = 125 and SF = 10, BW = 250. The number of relays is also quite important here, showing a similar arrangement of linear relationships. Notice again the convergence of the previously mentioned spreading factor-bandwidth pairs. The relationship between spreading factor and latency is further intimated by the authors when increasing the node count to 100, showing the high SF means longer transmission time.
The number of nodes in the network makes more interesting data. When there are more nodes, there are more packets generated. With more nodes to travel through, the latency is higher and there are more chances for a packet to get lost. The authors find this in Figures 6 and 7 below.
This should give us some good intuition about LoRa and how a mesh network would operate using it. Next, we’ll look at some problem-based solutions being researched today.
Phone, Meet Network
Our second paper is out of the University of the Philippines from Macaraeg, Hilario, et al. [2] It looks directly at our topic of off-grid communications by evaluating LoRa’s ability to serve as a text message delivery medium. The work is accomplished by implementing an ad-hoc on-demand distance vector (AODV) routing protocol using received signal strength indicator (RSSI) for the routing metric.
The network assumes a mobile phone is available to the user, and a LoRa transceiver is present. The LoRa transceiver is on an ESP32 development board along with a battery and bluetooth submodule. This will allow the phone to exchange information over the network as shown in Figure 1 of the paper.
The AODV routing mechanism using RSSI as the distance metric (RSSI-AODV) is chosen by our authors because of it correlates well with peak data rate (PDR). The routing is described by Figure 3 of the paper.
The experimentation includes two sets of tests; one point-to-point test and a multi-hop test. The point-to-point test has two nodes on the ground, with clear line of sight between nodes. The authors have shown reliable communication beyond 800 meters using packet size groups of x<25 bytes and 256>x>25 bytes. Results of the PDR performance are detailed in Figure 4.
However, in a more “real-world” scenario with obstructions in between nodes, the authors indicate a PDR of only 50% with a distance of up to 300 meters.
In the second set of tests using a multi-hop network, the authors implement two network structures shown in Figure 7 of the paper.
The tests are then run using both AODV and RSSI-AODV and monitor the route from source to destination. In confirming the route, the packets do take path with higher RSSI as expected. In Figure 8, it’s shown that PDR deteriorates wit hop count, consistent in AODV and RSSI-AODV and across both spreading factors evaluated.
The authors note that this type of network could cover a 5-square km area for around USD$240 at the cost of USD$40 per node. Since smartphones are ubiquitous in the Philippines, the deployment of such a network should be negligible with the assistance of local governments.
What If the Users Could Help Each Other?
The next paper in our research battery examines a network that links each person into a sort of mutual aid infrastructure using LoRa as the wireless network. They call this strategy LOCATE.
This work out of University of Bologna in Italy [3] seeks to disseminate emergency requests (E-REQs) throughout a LoRa network where each node is phone connected via USB to a LoRa radio. The E-REQ should eventually encounter another node that can solve the emergency.
The messaging system follows a dissemination strategy that uses a timer set when a node in the network receives an E-REQ. When the timer is up and no emergency solver (EV) is identified, the node passes the message to other nodes in the network. The authors described the strategy visually in their Figure 5.
The timer is referred to as a “contention window” and is described by
Where CWacc is the acceptance contention window, or the time a node will wait for an EV. Our exponent in e determines how close or far away priority should be given to potential solvers, as d(as, aj) is defined as the distance between the emergency source and the receiver node. For this contention window, priority is given to nodes closer to the source to win the contention. since nearby solvers might be able to reach the emergency source sooner.
For forwarding the E-REQ, there is a forwarding timer with a very similar equation.
This prioritizes the nodes further from the source node, so the message can be spread as fast as possible.
This was all done in phase 1 of a transmission. In phase 2, those that have received a message give an emergency response (E-REP), disseminated over this same scenario. This entire protocol is tuned using an epidemic diffusion model, where whether or not to actually send out the E-REQ message is determined probabilistically so as to not flood the network. How to tune this p∗ value is discussed further in the paper, but that’s enough math for now.
The simulation results of this strategy look promising. Several strategies were evaluated, but we’re going to look at the LOCATE scheme vs flooding and continuous dissemination in the graphs from Figures 11 and 12.
Here, the emergency resolution rate (ERR) is far and away an improvement over continuous dissemination, which is the probabilistic dissemination scheme without any analytical tuning. Even flooding is outdone here, as the ERR really picks up at 20 nodes and beyond.
For the emergency response time (ERT), the LOCATE scheme employing the probabilistic tuning again outpaces both continuous dissemination and flooding, indicating that not rebroadcasting the E-REQ from every node is key to good performance of the network. But wouldn’t it be nice to have an overhead metric? Well, University of Bologna does happen to report that as well.
The overhead of LOCATE does spike at 20 nodes as the other metrics do, but tapers off the more you have. This makes sense given that greater number of node means fewer actually have to disseminate the message. Even so, the overhead of LOCATE at its worst is better then both continuous dissemination and flooding at their best.
Now what about actual experimental results? Empirical data is always best, so let’s see those results.
Again, the LOCATE scheme performs better on each metric, although emergency overhead (EO) seems to grow much closer to flooding’s overhead. It should be noted that varying the number of solvers in the LOCATE scheme has a noticeable impact, shown in the paper. It should also be noted that the attempts to use 802.11 with the LOCATE scheme did not go well, further enhancing our desire to use LoRa.
Closing Thoughts
We’ve seen three different implementations of the LoRa technology for disaster relief scenarios. Beyond the development of Doomsday Survival Tech, there are other more mundane scenarios LoRa is being explored such as smart cities. The concept of smart cities is to provide promising opportunities for water safety monitoring, positional sensing, responsive wireless-enabled technology and more. We could have a city where current kilowatt-hour prices are given straight to vehicles so that smart electric cars are aware of the cost of charging up whenever they’re near a charging station. And even smart traffic lights that can detect accidents and communicate with other traffic lights using LoRa to reroute traffic to avoid the accident area? In these scenarios LoRa would be a possible solution since it is a long-range, low cost, low power communication standard. There are some limitations to LoRa such as the limiting power transmission and low bandwidth, and that is why there is still research being done to navigate through these limitations. I am very interested in the implementations of LoRa and hope to be in the forefront on these research topics, whether it be for disaster relief or improving the daily lives of people.
The graphs and tables are from the research papers summarized within the article.
- Pham, V. D., Kisel, V., Kirichek, R., Koucheryavy, A. & Shestakov, A. Evaluation of A Mesh Network based on LoRa Technology. 2021 23rd International Conference on Advanced Communication Technology (ICACT) 1–6 (2021) doi:10.23919/icact51234.2021.9370792.
- Macaraeg, K. C. V. G., Hilario, C. A. G. & Ambatali, C. D. C. LoRa-based Mesh Network for Off-grid Emergency Communications. 2020 Ieee Global Humanit Technology Conf Ghtc 1–4 (2020) doi:10.1109/ghtc46280.2020.9342944.
- Sciullo, L., Trotta, A. & Felice, M. D. Design and performance evaluation of a LoRa-based mobile emergency management system (LOCATE). Ad Hoc Networks (2020) doi:10.1016/j.adhoc.2019.101993.
