Design and implementation of resilient networks for environmental monitoring and network operation models addressing the specific challenges in the regions involved, in particular harsh environment conditions and vandalism.

Sensor data must be collected from both static and mobile sensors in the field and then aggregated by nodes which are energy-constrained either because they rely on batteries or on local power sources such as solar panels. The dynamic and adaptive collection results in variable bandwidth and latency requirements that the network should adjust to in order to preserve spectrum and battery utilisation. The network is hence to be designed to be energy-efficient and resilient, tolerant to faults or manipulation of its components

Methods and Tools

Nodes in a Wireless Sensor Network (WSN) are generally energy, computation, and memory constrained. Therefore, WSN design should take into consideration the use of efficient and low power sensor modules, resource aware data processing algorithms, and energy aware routing protocols. 

Previous work in the design of WSNs starts from the design of the optimal sensing module and its firmware with the aim of eliminating some deficiencies such as short transmission distance due to electromagnetic interference as well as the failure of the node to meet the dynamic configuration of the network [2]. Other work involved deducing the distance between the node and coordinator based on the measured received signal strength indicator (RSSI) so as to derive a log-normal shadowing effect as well as investigating the path-loss exponent for an outdoor and indoor configuration [3].

In addition to a large body of research on existing architectures for wireless sensor networks, routing also in case of non-continuous connectivity, and energy efficiency, there is the need to explore modeling techniques and how each technique describes the node behaviour (hardware and software reconfiguration) and network behaviour (network topology modification) [1]. ‘As wireless devices become pervasive and essential in our daily life, security becomes a critical issue’ [4]. Denial - of - Service (DoS) detection needs to be investigated. The approach is to develop or apply a modeling concept such as the Model-Driven Engineering (MDE) methodology for the design of the WSN. After modeling and simulations, then deployment will be done. 

There is the need to use IEEE 802.15.4 while modifying upper layers to allow low-rate WPAN and low-power lossy network devices to communicate through the internet. There is also a need to investigate the use of Low Power Wide Area (LPWAN) technologies such LoRaWAN for data gathering. Thelong distance communication introduces new challenges to LoRaWAN, which call for new mechanisms of error correction, and energy aware mechanisms. Optimal gateway placement may be necessary for optimal results. 

Potential innovation: The main innovations are expected to be demonstrations of the practical use of advanced computing and communication technologies, methods and tools relying on commodity hardware and software wherever possible and the ability to optimise performance adaptively

Deliverables: Reports on solutions, methods and tools as specified in the task descriptions below.

Interdependencies: WP1 will interact with WP2 to define the balance between what communication is possible (WP1) and what communication is desirable due to limitations in computing for local analysis (WP2).

Task 1.1: Adaptive Network Architecture for Sensor Data Acquisition.

Partners: NTNU DIT, MAK, UoJ

Duration: M1-M36

Description of work: Design of a network architecture addressing the specific challenges of data communication in the regions involved. The architecture should accommodate delay-tolerant data acquisition for some sensor and aggregate data and traffic that is sensitive to quality of service, allowing buffering and prioritisation as well as transition to disruption-tolerant modes of operation. This must be possible to run on relatively low-performance equipment that is also energy-constrained.

Activities include

• Study existing protocols for low bandwidth (e.g. 6LoWPAN and DTN) (M1-M6)

• Develop algorithms for transitioning between operation modes and prioritisation, interacting with Task 1.2. (M6-M18)

• Development of requirements and test cases for network operation (derived from WP2 and WP3) (M6-M18)

• Design and simulation of a wireless sensor network and hybrid back-haul network for different deployment settings (M12-M24)

• Validation in simulation and experiments of network design in Network design to be validated in simulations before being deployed. These will rely on the test cases and generic urban as well as rural reference cases.

Potential innovation: Power-lean designs, cognitive network algorithms anticipating payload requirements

Deliverables:

D1.1.1: Report on a proposed network architecture using state of the art tools for modeling and analysis of Real-Time Embedded systems.

D1.1.2 Report methods for decomposition of the global applications to a set of local node behaviours for the implementation phase.

D1.1.3 Report on the implementation of the network architecture based on emerging next generation IoT operating systems, successors to contiki-os used in the WIMEA-ICT project.

Interdependencies: Co-Design with T1.2, requirements derived from WP2 and WP3 tasks.

Task 1.2: Network Resilience Monitoring and Maintenance

Partners: NTNU DIT, MAK, UoJ

Duration: M6-M42

Description of work: Ensure reachability of all sensors and gateway nodes, identifying faults and adverse interactions and ensuring resilience through re-configuration of network paths and components before disruptions and violations of quality of service parameters occur. This is to prevent loss of sensor data or disruption of time series -- both of which have adverse effects on data quality, particularly for automated analytical methods such as machine learning techniques.

Activities include:

• Simulation of network performance characteristics in the event of DoS attack or faults and failures to understand early indicators of impending faults and disruptions or planned maintenance.

• Design of protocols for both static and mobile network components which can pre-emptively reconfigure the network (cf. Task 1.1) to ensure continued availability while minimising local redundancy of network components.

Potential innovation: Predictive fault models and adaptive reconfiguration preventing quality of service degradation

Deliverables:

D1.2.1 Report and demonstration of secure and fault tolerant protocols for wireless sensor networks used in the project.

Interdependencies: In-dependency (concurrent) with T1.1

Resource requirements: 1 MSc candidate, shared equipment as specified above

Task 1.3: Energy-Aware Network Optimisation

Partners: NTNU DIT, MAK, UoJ

Duration: M12-M48

Description of work: Most sensors will be equipped with batteries, but some advanced sensor components and gateway systems will not be efficient to operate based on battery power. These will, however, typically still not be able to rely on mains power but will have local generation. Sizing generator capacity for peak loads is inefficient, hence the goal is to design protocols and algorithms to monitor and schedule energy consumption so as not to exceed peak capacity, thereby minimising the sizing of both generating and storage capacities.
These problems are to be represented as both local and distributed optimisation problems which must also take incomplete information into account

Activities include

• Study energy harvesting techniques suitable for powering wireless sensor nodes in various deployment settings (cf. Task 1.1)

• Construct local and distributed optimisation algorithms for energy storage, consumption, and generation under uncertainty and efficient adaptation to changes such as reconfiguration or faults

• Integrate results from Task 1.1 and 1.2 (as well as opportunistic demand from other tasks, e.g. for charging of batteries of sensors or other devices) to obtain load profiles. 

• Develop simulation including mobility models for validation based on test cases in Task 1.1

• Develop scheduling algorithms for constraint-based optimisation of energy consumption.

• Validation of results obtained in simulation against mathematical models, limited field validation to ensure correct parameterisation.

Potential innovation: Use of new developments in battery technologies, including hybrid ultracapacitors, distributed optimisation problems under uncertainty.

Deliverables:

D1.3.1 Optimized energy harvesters / generators designed for wireless sensor nodes and drones.

D1.3.2 Energy aware algorithms developed to extend the lifespan of a battery or any other energy source.

D1.3.3 Distributed optimization algorithms for energy consumptions and storage

Interdependencies: Task 1.1 and Task 1.2; Ph.D. candidate to be tasked in Task 1.4

Task 1.4: Mobile Gateway Nodes

Partners: NTNU DIT, MAK, UoJ

Duration: M6-M48

Description of work: Some sensors in the project will have substantial bandwidth requirements, but may themselves be deployed dynamically, resulting in high peak requirements in "hot spots". This task is to investigate if the use of drone-mounted mobile network components is an efficient way of addressing these transient requirements.

Activities include:

• Investigate the sizing of drone-mounted network components and constellations to allow the temporary transmission of bursty traffic (e.g. visual and multispectral imaging) to higher-capacity gateway systems with virtual routes consisting of drone constellations. 

• Drones may also rely on the energy and charging infrastructure investigated in Task 1.3.

• Development of edge computing AI techniques for pest detection

Potential innovation: [Describe the potential innovation that the task/wp will result into]

Deliverables:

D1.4.1 Self organizing algorithm to control and coordinate drone constellation developed

D1.4.2 Designed algorithm for monitoring energy levels from all drones in a constellation and autonomously controlling drones to their respective charging infrastructure.

D1.4.3 Deployment of self organizing drone constellation in fields for pest detection.

D1.4.4 AI based Pest Detection  algorithm developed and deployed

Interdependencies: Tasks 1.1 – 1.3

Resource requirements: 1 Ph.D. candidate (together with Task 1.3), 2 M.Sc. candidates, shared equipment as specified above