Title: Towards Next-Generation Global IoT: Empowering Massive Connectivity with Harmonious Multi-Network Coexistence
Authors: Ziyue Zhang, Xianjin Xia, Ruonan Li, Yuanqing Zheng（The Hong Kong Polytechnic University）
Scribe: Ziyi Wang (Xiamen University)
Introduction
LoRaWAN is regarded as a key technology for next-generation global IoT infrastructure, providing low-cost network access for millions of IoT devices worldwide. However, when operators attempt to scale their services to meet growing IoT connectivity demands, they encounter bottlenecks where the number of concurrent users falls far below theoretical values (e.g., TTN supports a maximum of only 16 concurrent users). Existing LoRaWANs tend to exhibit contention rather than cooperation in shared network environments, leading to a significant decline in capacity as network density increases. To address this, the authors conducted an in-depth analysis of operational LoRaWANs and discovered for the first time that the capacity degradation is not due to traditionally assumed wireless interference or packet collisions, but rather the overlooked root cause: “decoder contention.” Based on this new insight, the authors propose design principles and present practical, implementable solutions to mitigate contention, optimize spectrum utilization efficiency, and promote spectrum sharing among network operators.
Key idea and contribution
The authors identify the “decoder contention problem” as the root cause of LoRaWAN’s capacity bottleneck. They first validate this bottleneck through large-scale real-world measurements, then propose eight strategies, from which four hardware-agnostic solutions fully compatible with existing network equipment are selected and integrated into a software plugin named AlphaWAN. Furthermore, two novel primitives are introduced to integrate the selected strategies into the LoRaWAN stack: intra-network channel planning, which coordinates channels and parameters between gateways and end-devices to minimize contention, and inter-network channel planning, which allocates frequency-misaligned channels to different operators to achieve physical-layer isolation and mitigate inter-network contention.
The contributions of this work are summarized as follows:
- Demonstrating that operational LoRaWANs fall short of capacity when supporting large-scale IoT connections, and revealing for the first time the root cause of this capacity gap: decoder contention.
- Systematically exploring eight strategies to address the decoder contention problem and thoroughly evaluating their feasibility for real-world LoRaWAN deployments.
- Designing and implementing the AlphaWAN system to comprehensively manage both intra-network and inter-network contentions.
Evaluation
Based on the open-source LoRaWAN platform ChirpStack, the authors implemented AlphaWAN and conducted comprehensive evaluations through real-world experiments and large-scale simulations. The results show that under identical spectrum and gateway resources, AlphaWAN supports up to three times more end-devices. With additional gateways deployed, the system capacity approaches the theoretical limit of LoRaWAN. In a six-network coexistence scenario, spectrum utilization demonstrates a remarkable improvement of 778.1%. These findings significantly indicate that AlphaWAN effectively addresses the decoder contention problem in practical deployments, thereby providing a viable pathway for large-scale implementation of LoRaWAN as global IoT infrastructure.
Q&A
Q1: My question is about how to assign different sequences to different devices. How can a device clearly specify the required sequence during the attachment process to access different channels?
A1: For the decoder contention problem, we can employ frequency planning as a solution. For instance, as shown in this figure, we can configure different gateways and devices to use different frequencies or different channels. This way, the gateways won’t all receive identical sets of data packets, and the packet arrival sequences will differ. Different gateways may receive different subsets of packets, which can potentially increase overall system capacity using the same number of gateways. For end-users, we can pre-configure their device frequencies: assign a specific subset of optional channels to different devices, allowing them to choose different gateways. This logically distributes users across gateways, enhancing overall capacity.
Q2: Is it necessary for all devices to follow a uniform policy to ensure they evenly select the correct frequencies and avoid contending for the same frequency again?
A2: Yes, such a policy is required. This can be achieved by configuring the gateways and end-nodes prior to deployment.
Q3: How can we get different network operators to work together to make this full deployment happen? What happens if only some agree to work together but others do not?
A3: We have thoroughly evaluated this scenario in our paper. If all operators join the coordinated system, every operator will achieve higher performance than before. Even if only a subset of operators participates, those who cooperate can still gain some capacity improvement, though the gain would be less significant compared to the scenario of full cooperation. Detailed data can be found in the paper.
Personal thoughts
Through empirical measurements and extensive experiments, this paper demonstrates that decoder contention, as opposed to traditional interference, is the root cause of capacity limitations. The design of AlphaWAN is pragmatic and maintains backward compatibility with existing hardware, demonstrating significant practical value for real-world deployment. However, the study does not delve into real-time scheduling under dynamic workloads, nor does it address performance boundaries in extreme-density deployment scenarios.