Innovation Spring 2026

A customized silicone suction cup attached to the system’s actuator ensures the right amount of pressure can be applied to grip the mushroom without causing damage. P hoto : C ourtesy of 4AG

Sensing devices in agriculture have typically focused on creating a picture of a plant’s health through devices that measure certain variables in a plant’s environment. Substrate sensors – probes that plug into coco coir, soil, stone wool, and other grow mediums – are commonly deployed in the industry and can provide information about factors like moisture content, electrical conductivity, and pH. However, they may not accurately reflect a plant’s real time health. Alternatively, sampling a plant’s chemical condition in more detail typically involves a multi-day laboratory analysis. For high-value crops, this delay in monitoring can mean the difference in a thriving batch and one that provides sub-optimal yields. Kim experiments with a variety of sensing options, ranging from image processing to detecting plant electrical signals, or “electrophysiology,” to gather and process real-time data from crops as they develop. “Image processing is dominant right now in AI-based prediction, but electrical signals may let us detect problems earlier,” he said. “Before water stress, nutrient stress, or pathogen attacks become visible on the leaf through RGB cameras [which accurately capture colour

images], there may already be electrophysiological (EP) signals. Filling that gap with early detection is the key thing we want to do.” Kim’s recent work introduces several iterations of robotic systems designed to measure faint EP signals from plants in real time. “The challenge is not just measuring the signal,” Kim said, “but measuring it without disturbing the plant.” Traditional EP probes are costly and invasive, often altering the very signals they aim to capture. By contrast, Kim’s approach emphasizes low-cost, minimally intrusive sensors that can be deployed densely across growing environments. One option, presented in the July 2025 issue of Advanced Intelligent Systems , describes a prototype combining a mobile robot, an EP sensor, and a portable Faraday cage (an enclosure that blocks electromagnetic fields) to isolate plant signals from environmental noise. As the robot moves from plant to plant in a greenhouse, it collects electrical data that is analyzed by a Convolutional Neural Network (CNN), a type of machine-learning algorithm, trained to classify irrigation levels.

18

Spring 2026

Innovation