(5 min read)
Digital Twins are increasingly making enormous contributions to our understanding of complex physical systems, be it the optimization of manufacturing design or plant in industry 4.0, the prediction of traffic and pedestrian flows in smart cities or the refinement of comfort and experiences of our physical spaces. One area where Digital Twins could lead to breakthrough understanding is in the natural environment. Here they could help identify the more subtle links between subsystems not obviously intertwined or assist governance policies of varying timescales, such as dealing with invasive species or wildlife diseases, reducing erosion and increasing biodiversity.
Of course, simulating the natural environment is an entirely established field in its own right. However, the defining characteristic of a Digital Twin is that it is not a simulation, i.e., imitation derived from extrapolating theory, but rather a depiction of real data in sufficient detail to draw meaningful insights. And whilst capturing data in homogeneous, predictable, secured and controlled environments, such as factories, is one thing, doing so in the wild is quite another!
The Swarmchestrate demonstrator takes place in an abandoned quarry just outside Madrid, Spain, that is being returned to nature. It is managed by a small collective that employs light touch practices to nudge recovery with minimum direct interference. Thus, for example, we find the once harsh, bulldozed and terraced landscape previously traversed by massive dump trucks and excavators has been softened, with a myriad of gentle meandering slopes that interweave to create varying amounts of shade, moisture and wind protection. These channel rainwater into a pond that grows in winter and ebbs in summer, leaving smaller pools as it retracts. Different species colonise different areas of the site based on these slight factors. Yet, beyond those initial earthworks, restoration is mostly down to nature itself – leaving the flowers, vines, trees, bushes, reptiles, birds, mammals and amphibians to do their thing, at their pace.
Our challenge is to model this in real-time without disturbing any of these natural fluxes taking place and to do so with both accuracy and resilience. Cables cannot be laid. Concrete structures cannot be put up and whatever sensors we use must stand up to whatever is thrown at them: blistering heat or driving snow, extreme wind or torrential rain. The area is alive, and our installation must withstand roosting birds, muddy water and curious creatures – all without fixed power or cable internet!
The only infrastructure available on site is a small office equipped with solar panels and the general mobile network coverage. Much of the time there is no one on site, or a small number of volunteers tending to tasks such as butterfly counts or maintenance of the perimeter fence or other features. However, at times the site receives visits from large groups, such as school excursions. Consequently, the available resources are affected not solely by the supply side (light coverage) but also unpredictable demand.
During the research project our focus is predominantly on these challenges, and how Swarmchestrate enables this by orchestrating the resources most optimally, however belying the demonstrator is a credible business case that extends into several different domains. For instance: education – bringing science to life for young learners; academic study for specialist university departments; new policies for zoning or environmental impact; and more laterally: new ways of implementing IoT generally in harsh environments such as active quarries, oceans or inhospitable landscapes.
The sensors used in the demonstrator include multiple connected weather stations to monitor particular variations in rainfall and humidity across the site, water height sensors on the static pond, and water quality sensors. These will communicate their data via 5G and be powered by battery and/or small solar panels. In the small site office, a cluster of Raspberry Pi’s will facilitate a local version of the Digital Twin, whilst a duplicate will be provided via standard cloud. The former can be used by the naturalists and maintenance crew on site, and for when there is a group tour. The latter can be used by the landowner and scientists working remotely.
Aerial image of the abandoned quarry just outside Madrid, Spain and locations of installed IoT devices (© Swarmchestrate consortium 2024-2026)
Swarmchestrate is expected to provide a significant benefit, enabling resources to be optimised according to changing supply and demand conditions. For instance, one of the barriers to scalability of IoT sensors in natural environments (and other situations without power sources) is the time and cost of battery replacement. Spread over a large enough area, the mere task of changing batteries for each sensor every few weeks would become a full-time occupation. With Swarmchestrate we expect the IoT swarm to automatically adapt. When there is no need for high granularity data, sensors can be switched to less frequent readings and communications, spending much greater proportion of their time dormant. Then, when atmospheric changes occur, the pond water level is rising and consequently the water quality variable, much more frequent readings are required. There are a number of other optimizations anticipated, such as bandwidth, preprocessing data, and similar.
In conclusion, the demonstrator is poised to both further our understanding of optimizing device swarms under difficult constraints, at a practical level bring benefit to the community behind the quarry initiative and their userbase, and at a business level expand our ability to deploy IoT and digital twins in hostile environments.
Editor: Daniel Field, UST