A Multi-Layer Feedback System Approach to Resilient Connectivity of Remotely Deployed Mobile Internet of Things

Muhammad Junaid Farooq, Quanyan ZhuIEEE Trans. Cogn. Commun. Netw 2018 · DOI

TL;DR

This paper introduces a novel cognitive feedback system for remotely deployed Internet of Things (IoT) devices, specifically addressing the challenge of maintaining resilient connectivity in dynamic environments with limited traditional infrastructure. The proposed multi-layer framework allows mobile access points (MAPs), like drones, to autonomously adapt their positions and configurations to ensure continuous communication with both each other and the underlying smart devices, even amidst mobility and device failures.

What Problem Does This Paper Address?

The proliferation of the Internet of Things (IoT) has created a demand for ubiquitous connectivity, especially in remote or disaster-stricken areas where traditional communication infrastructure is unreliable or non-existent. In such scenarios, a multi-layer network architecture is needed, where mobile access points (MAPs), such as unmanned aerial vehicles (UAVs), provide coverage to ground-based smart devices (MSDs). However, designing and managing these overlay networks presents significant challenges:

  • Lack of Backhaul Connectivity: Remote environments often lack stable connections to the wider internet.
  • Device-to-Device (D2D) Communication Reliance: MAPs must communicate with each other to form a cohesive network, relying on D2D links.
  • Dynamic Connectivity: The network's connectivity is constantly affected by the mobility of MSDs and the potential failure or adversarial attacks on MAPs.
  • Complex Deployment: Traditional placement strategies (like circle packing or p-median) are often static, NP-hard, and don't adequately address the need for inter-MAP connectivity or dynamic adaptation.

This paper aims to bridge this gap by proposing a framework that enables MAPs to autonomously and adaptively maintain resilient connectivity in these challenging conditions.

Methods

The core of the proposed solution is a cognitive feedback loop that allows each MAP to independently adjust its behavior based on local information and communication with its neighbors. This loop consists of several key components:

  • MAP-MSD Matching: At each step, MSDs are matched to the nearest available MAP, considering MAP capacity constraints. This ensures that devices are served efficiently.
  • MAP Dynamics and Objective: MAPs are modeled with kinematic dynamics (position, velocity, acceleration). Their objective is to achieve a desired configuration that maximizes coverage and maintains connectivity.
  • Cognitive Controller Design: Each MAP computes a control input based on three main forces:
  • Attractive/Repulsive Forces: MAPs are attracted to areas with high MSD density and repelled from each other to maintain a minimum distance, preventing excessive overlap and ensuring coverage. They also attract each other if a MAP is overloaded.
  • Velocity Consensus: MAPs align their velocities with their neighbors to ensure smooth movement and prevent disconnections due to relative velocity.
  • Individual Goal: Each MAP aims to move towards the centroid of the MSD cluster it serves, ensuring it's optimally positioned.
  • Cluster Centers: To handle MSD mobility, the MSDs are clustered, and their centroids are used as dynamic "goal positions" for the MAPs. This allows MAPs to follow the general movement of the user population.
  • Discretized System: The continuous-time dynamics are discretized for practical implementation.

The framework is designed to be distributed, meaning each MAP makes decisions based on its local observations and information exchanged with immediate neighbors, making it robust to single points of failure.

Datasets

Not discussed in the paper. The evaluation is based on simulations with synthetically generated data representing MSDs and MAPs.

Key Insights and Ideas

  • Cognitive Autonomy is Key: The paper demonstrates that a distributed, cognitive approach where MAPs independently adapt their behavior is highly effective for maintaining connectivity in dynamic and uncertain environments, outperforming static, centralized methods.
  • Balancing Coverage and Connectivity: The framework cleverly balances the need for MAPs to spread out to cover MSDs with the necessity of staying close enough to maintain D2D communication links.
  • Resilience to Failures: The feedback loop allows the system to quickly reconfigure and recover connectivity even when a significant percentage of MAPs fail, highlighting its robustness.
  • Dynamic Adaptation: The use of cluster centroids for MSDs allows the MAPs to fluidly follow the movement of users, a critical advantage over static deployment strategies.

Key Results

The simulations showcase the effectiveness of the proposed cognitive framework:

  • High MSD Coverage: The framework achieves high proportions of covered MSDs, comparable to or better than baseline algorithms, with significantly fewer MAPs.
  • MAP Reachability: The system maintains high reachability (algebraic connectivity) between MAPs, even under significant MAP failure events (up to 30% failure).
  • Information Penetration: The probability of information dissemination within the MAP network remains high, indicating effective communication.
  • Performance Comparison: Compared to circle packing and p-median approaches, the proposed dynamic approach requires fewer MAPs to achieve comparable or superior coverage and connectivity, especially in dynamic scenarios. For instance, it achieves high coverage with around 20 MAPs, whereas circle packing requires over 100.

Limitations and Future Work

  • Centralized Cluster Center Estimation: While the MAPs operate distributively, the paper notes that obtaining cluster centers might require a centralized entity or local estimation with distributed consensus, which adds complexity. The authors suggest future work could explore fully distributed consensus mechanisms for cluster centers.
  • Parameter Tuning: The performance is sensitive to parameters like communication range and failure thresholds, which need to be carefully determined based on specific radio propagation models and reliability requirements.
  • Future Directions: The authors suggest exploring more sophisticated distributed consensus algorithms for MAPs to agree on MSD locations and investigating multi-mode operation for MAPs to handle diverse MSD distribution patterns.

Summary updated

BibTeX
@article{farooq2018multi,
  author = {Muhammad Junaid Farooq and Quanyan Zhu},
  title = {{A Multi-Layer Feedback System Approach to Resilient Connectivity of Remotely Deployed Mobile Internet of Things}},
  journal = {IEEE Trans. Cogn. Commun. Netw},
  year = {2018},
  doi = {10.1109/tccn.2018.2837907},
}
Figure