A Middleware Solution for Wireless IoT Applications in Sparse Smart Cities

Paolo Bellavista, Carlo Giannelli, Stefano Lanzone, Giulio Riberto, Cesare Stefanelli, Mauro Tortonesi

TL;DR

This paper introduces RAMP-WIA, a middleware designed to simplify the development and deployment of wireless IoT applications in "sparse" smart city environments. It enables devices like smartphones to dynamically form multi-hop networks, manage data transmission even with intermittent connectivity, and reliably control sensors, all while allowing for on-the-fly software updates.

What Problem Does This Paper Address?

The proliferation of smart devices and sensors is leading to a massive increase in connected devices, particularly in urban environments. However, building robust IoT applications for these "sparse" smart cities presents significant challenges. These environments are characterized by:

  • Infrastructure-less Networks: Devices often connect directly to each other (peer-to-peer) rather than relying on fixed Wi-Fi access points or cellular towers.
  • Dynamic Topologies: Users and their devices are mobile, leading to frequent changes in network connectivity and availability.
  • Intermittent Connectivity: Devices may become temporarily unreachable due to network partitioning or battery limitations.
  • Heterogeneous Devices: A mix of devices with varying capabilities (smartphones, Raspberry Pis, sensors) needs to interoperate.
  • Application Management: Deploying, updating, and managing applications across these dynamic and distributed devices is complex.

Existing solutions often focus on specific aspects (e.g., routing, sensor management) but lack a comprehensive approach to handle the combined challenges of dynamic connectivity, opportunistic networking, and application lifecycle management in these sparse environments.

Methods

RAMP-WIA is built upon a three-layer reference architecture:

  1. Single-hop Connectivity Layer: This layer focuses on managing direct device-to-device (e.g., Wi-Fi, Bluetooth) connections. It allows for both manual configuration and autonomous management of IP subnets, abstracting away the complexities of different wireless protocols. A key component here is the Link Connectivity Component (LCC), which dynamically switches between client and hotspot roles and connects to available hotspots to maximize connectivity opportunities.
  2. Multi-hop Networking Overlay Layer: This layer builds on the single-hop connections to enable opportunistic multi-hop packet delivery. It transparently dispatches packets across nodes without requiring OS-level routing table modifications. A crucial addition is the RAMP Opportunistic Networking (RON) component, which extends the middleware's ability to handle network failures and partitions. RON persists packets and retries transmission when connectivity is restored, using techniques like "spray and wait" to maximize delivery probability.
  3. Coordination Wireless IoT Application Layer: This layer simplifies the development of applications by abstracting away node coordination. It facilitates the transparent dispatching of commands and software updates. For actuation, it introduces a quorum-based distributed actuation solution where commands are only issued if a sufficient number of sensor nodes confirm their availability. This layer also supports a resiliency policy for sensor nodes to adopt a standard behavior when controller connectivity is lost.

The paper also discusses dynamic processing updates, proposing solutions like distributing Java classes and scripts, or leveraging containerization (Docker) for more complex deployments.

Datasets

Not discussed in the paper. The evaluation relies on experimental testbeds using off-the-shelf devices.

Key Insights and Ideas

  • Opportunistic Networking is Key: In sparse smart cities, leveraging intermittent connectivity opportunities is not just a fallback but a core strategy for reliable communication.
  • Abstraction is Crucial: Hiding the complexity of dynamic wireless links and multi-hop routing from application developers is essential for rapid development.
  • Dynamic Software Updates are a Must: The ability to push new code or configurations to devices on-the-fly is vital for adapting to changing needs and hardware in smart cities.
  • Quorum-Based Actuation Balances Reliability and Dynamism: Ensuring a command is only sent when a sufficient number of sensors are available provides a practical balance between guaranteed delivery and the reality of intermittent connectivity.
  • Resiliency Policies Prevent Stalling: When controllers fail, sensors need a mechanism to maintain basic functionality, preventing complete system failure.

Key Results

The paper presents experimental results from small-scale testbeds using Android smartphones and Raspberry Pi devices.

  • Two-hop Scenario: Demonstrated successful packet dispatching in a Bluetooth-IEEE 802.11 chain with manual configuration, achieving an average end-to-end dispatching time of 115.8 seconds (with significant overhead from manual configuration). When excluding manual configuration time, the dispatching took only 5.8 seconds.
  • Four-hop Scenario with LCC Policies: Using automated LCC policies (lazy and hectic), packet dispatching times were reduced to 365.4 seconds (lazy) and 158.9 seconds (hectic), showing the effectiveness of dynamic link management. The hectic policy significantly lowered dispatching time.
  • Quorum-Based Actuation: The presence detection application successfully demonstrated:
  • Controller receiving "here-I-am" messages from sensors.
  • Command issuance only when the quorum (all sensors in one test case) was met.
  • Command abortion when the quorum was not met.
  • Sensor nodes detecting controller unavailability and activating a resilience policy.

Limitations and Future Work

  • Scale of Experiments: The current experiments are small-scale. The authors plan to conduct wider and more in-depth performance analyses on larger, real-world testbeds with thousands of devices.
  • Containerization Feasibility: While Docker is discussed as a promising solution for dynamic updates, its large size and deployment challenges on off-the-shelf smartphones are acknowledged. Future work will investigate its feasibility.
  • QoS Management: The authors are exploring the adoption of Software Defined Networking (SDN) to improve Quality of Service (QoS) by managing traffic flows and considering network capabilities.
  • Containerization of Middleware: They are also working on containerizing the RAMP-WIA middleware itself.

Summary updated

Abstract

The spread of off-the-shelf mobile devices equipped with multiple wireless interfaces together with sophisticated sensors is paving the way to novel wireless Internet of Things (IoT) environments, characterized by multi-hop infrastructure-less wireless networks where devices carried by users act as sensors/actuators as well as network nodes. In particular, the paper presents Real Ad-hoc Multi-hop Peer-to peer-Wireless IoT Application (RAMP-WIA), a novel solution that facilitates the development, deployment, and management of applications in sparse Smart City environments, characterized by users willing to collaborate by allowing new applications to be deployed on their smartphones to remotely monitor and control fixed/mobile devices. RAMP-WIA allows users to dynamically configure single-hop wireless links, to manage opportunistically multi-hop packet dispatching considering that the network topology (together with the availability of sensors and actuators) may abruptly change, to actuate reliably sensor nodes specifically considering that only part of them could be actually reachable in a timely manner, and to upgrade dynamically the nodes through over-the-air distribution of new software components. The paper also reports the performance of RAMP-WIA on simple but realistic cases of small-scale deployment scenarios with off-the-shelf Android smartphones and Raspberry Pi devices; these results show not only the feasibility and soundness of the proposed approach, but also the efficiency of the middleware implemented when deployed on real testbeds.
BibTeX
@inproceedings{bellavista2017middleware,
  author = {Paolo Bellavista and Carlo Giannelli and Stefano Lanzone and Giulio Riberto and Cesare Stefanelli and Mauro Tortonesi},
  title = {{A Middleware Solution for Wireless IoT Applications in Sparse Smart Cities}},
  booktitle = {Sensors},
  year = {2017},
  url = {https://openreview.net/forum?id=pegt5qb1Uu},
}
Figure