SMART-AGENTS project update

SMART-AGENTS stands for Swarm collaborative Multi-Agent cybeR physical sysTems with shAred sensinG modalitiEs, 5G commuNication and micro-elecTromechanical Sensor arrays

The SMART-AGENTS project conducts research with a view to developing symbiotic networks of collaborative agents (drones) to solve challenges in logistics, agriculture and other areas. The project proposes a new multi-drone scenario to create a collaborative network. The system will then be demonstrated across multiple industrial environments.

SMART-AGENTS is an interdisciplinary project between the Bernoulli Institute for Mathematics, Computer Science and Artificial Intelligence and the Engineering and Technology Institute Groningen. Other partners include TNO, Werkman Hoofcare, Corvus Drones, Arox Logistics and Meewisse Roses.

In Spring 2023, we spoke with Dr. Kerstin Bunte, the Project Leader of SMART-AGENTS. Kerstin currently works at the Faculty of Science and Engineering at the University of Groningen, where she conducts research into machine learning, artificial intelligence, data mining and interpretive modelling. Projects include biomedical data analysis, dimensionality reduction and machine learning in technical applications.

Autonomous technical solutions

Industrial environments such as agriculture and logistics present challenges to the use of autonomous technical solutions. The vast scale of greenhouse crop production makes crop forecasting, pest control and monitoring of photosynthetic efficiency difficult, costly and dangerous. Similarly, large warehouses hold huge stocks that must be regularly managed and controlled to optimise the logistics processes. Both industries present challenges to the use of autonomous technical solutions.

Fully automated system

The SMART-AGENTS project will develop a robust, fully automated system using a collaborative network of small autonomous drones. The small drones will use their equipment and capabilities to help solve critical challenges. The research project involves several interdisciplinary aspects. The researchers will focus on developing novel biomimetic self-powered microsensors and integrating multiple sensors for robust navigation and control in fast-moving environments. Furthermore, they will develop real-time 5G communications and edge computing solutions to provide scalability and enable the integration of multiple agents.

Three test locations

The project collaborates with some small family-run businesses and startups. Testing is already underway with Werkman Hoofcare, Arox Logistics and Meewisse Roses.

Werkman Hoofcare: Development and improvement of racehorses using microelectromechanical systems (MEMS). "These are tiny but efficient self-powered, 3D-printed sensors. Different types of MEMS sensors exist for a wide range of applications. Werkman is particularly interested in pressure sensors and the technology to print them. Our testing here encompasses pressure sensors in the horse's hoof. This is a precarious process, of course, as the sensors must withstand the weight of the horse."

Meewisse Roses: Corvus Drones are used to take photos of plants in greenhouses as they grow to visualise any plant diseases. The idea here is to identify them and combat them. "We are using a lens that is also used in astronomy, a much cheaper - and lighter - solution than that currently used for monitoring. A spectral lens is cheaper and lighter than a multispectral camera, which costs €10,000 - €20,000 and is heavy. Multispectral imaging provides maximum information for agricultural monitoring.

Arox Logistics: MEMS logistics for stock-taking in warehouses with flow sensors. MEMS flow sensors are particularly suited to avoiding obstacles and for swarming, or flying in formation. The sensors must make it possible to detect changes in airflow caused by local interference. Arox Logistics is interested in robust drone navigation for cycle counting in a fast-moving environment (products go in and come out every day), making visual navigation difficult. A combination of sensors, including MEMS, radio frequency signals and visual signals, are therefore combined to navigate through this challenging environment.

Even when it is dark and hazy

The SMART-AGENTS platform will enable the integration of multiple sensors, including sensor arrays for microelectromechanical systems (MEMS), cameras, radio frequency signals and magnetic field measurements. 3D flow-based images of the environment allow control of connected agents even in dark and hazy conditions. Kerstin explains some of the challenges around sensing. "Deploying drones in greenhouses or warehouses is by no means simple. GPS signals are obstructed by buildings and are therefore not always readily available in indoor spaces. The environments also tend to be fast-moving, with daily changes taking place, meaning the images are constantly changing. However, everything looks similar, and you also have the added challenge of shadows and changing light. What's more, lenses are often set for far away rather than close up. In the warehouses, we are therefore looking at a combination of machine learning and robotic engineering solutions to find a robust and efficient solution using multiple sensors deployed in a swarm of drones."

Increasing demand for mobile agents

There is a growing demand for "mobile agents" that can perform various tasks. In particular, multi-platform cyber-physical systems (CPSs), where various agents contribute to a network that simultaneously solves multiple objectives. "We are working in parallel on different parts of the SMART-AGENTS project. Above all, we want to make everything smaller, more robust and lower cost; this is where the potential lies. Doing more with less energy, not just the algorithms, but also the robotics and sensing. We hope to be able to give demonstrations later this year."