Swarms of smart sensors explore the unknown

Monday, May 18, 2020

There is a lack of technology for exploring inaccessible environments, such as water distribution and other pipeline networks. Mapping these networks using remote-sensing technology could locate obstructions, leaks, or faults to deliver clean water or prevent contamination more efficiently. The long-term challenge is to optimise remote-sensing agents in a way that is applicable to many inaccessible artificial and natural environments.

The EU-funded PHOENIX project addressed this with a method that combines innovations in hardware, sensing, and artificial evolution, using small spherical remote sensors called motes.

‘We integrated algorithms into a complete co-evolutionary framework where motes and environment models jointly evolve,’ say project coordinator Peter Baltus of Eindhoven University of Technology in the Netherlands. ‘This may serve as a new tool for evolving the behaviour of any agent, from robots to wireless sensors, to address different needs from industry.’

 

Artificial evolution

The team’s method was successfully demonstrated using a pipeline inspection test case. Motes were injected multiple times into the test pipeline. Moving with the flow, they explored and mapped its parameters before being recovered.

Motes operate without direct human control. Each one is a miniaturised smart sensing agent, packed with microsensors and programmed to learn by experience, make autonomous decisions and improve itself for the task at hand. Collectively, motes behave as a swarm, communicating via ultrasound to build a virtual model of the environment they pass through.

The key to optimising the mapping of unknown environments is software that enables motes to evolve self-adaptation to their environment over time. To achieve this, the project team developed novel algorithms. These bring together different kinds of expert knowledge, to influence the design of motes, their ongoing adaptation and the ‘rebirth’ of the overall PHOENIX system. Artificial evolution is achieved by injecting successive swarms of motes into an inaccessible environment. For each generation, data from recovered motes is combined with evolutionary algorithms. This progressively optimises the virtual model of the unknown environment as well as the hardware and behavioural parameters of the motes themselves.

As a result, the project has also shed light on broader issues, such as the emergent properties of self-organisation and the division of labour in autonomous systems.

For more information, please click here

Image source: © Bart van Overbeeke, 2019

Further Information: External Link