KiwiSense: Bio-inspired vibration and computer vision probe for early pest detection

Overview

KiwiSense is an early pest detection system inspired by the kiwi bird, which locates insects in darkness using vibration-sensitive mechanoreceptors in its beak. The system embeds a 5x5 mm MEMS vibration chip in a stake that is inserted into the soil like a garden ornament. Each stake is designed to monitor a 10–15 m radius area, capturing micro-seismic signatures produced by insects moving through soil or grass.

The core value proposition is early detection: image-based methods can only spot visible pests, so infestations are often well established by the time they are seen. Vibration sensing can detect insects underground or hidden under foliage, providing a first-signs warning. A companion camera then focuses on the flagged area for visual species identification. This two-stage design combines the sensitivity of vibration sensing with the specificity of computer vision, reducing false positives, unnecessary pesticide use, and consumer confusion.

Potential applications include residential lawn and garden pest management, turf care, specialty agriculture, and smart-home pest monitoring. Because the stake has no external infrastructure and is intended to look like a garden ornament, it is positioned as a low-barrier, consumer-accessible solution.

Technical specifications

How it works:

• MEMS vibration chip (5x5 mm) detects micro-seismic signatures from insect movement in the substrate.
• On-device AI continuously classifies vibration patterns to detect pest activity, even when insects are underground or hidden by vegetation.
• On positive detection, the system directs a companion camera to the flagged area for computer-vision species identification.
• The two-stage pipeline alerts the homeowner with actionable information on species, location, and severity via a smartphone app.
• A BLE radio and cloud alert pipeline are planned for data transmission.

Development roadmap:

• Phase 1 (months 1–3): procure MEMS vibration chips, design stake housing, collect vibration signatures from target pests such as white grubs, cutworms, and mole crickets across soil types; build labeled training dataset.
• Phase 2 (months 4–7): train on-device ML models for species discrimination; integrate chip and BLE into prototype stake; build iOS/Android app and cloud alert pipeline.
• Phase 3 (months 8–11): field-validate 20 units at residential sites; measure detection rate, false-positive rate, and battery life against a camera-only baseline; iterate on firmware and model.

Planned future commercialization steps include documenting manufacturing specifications and preparing co-development agreements with pest-control companies.

Technology readiness level

KiwiSense is at an early research and development stage. The concept and validation plan are defined, but the hardware prototype, labeled vibration dataset, and field testing have not yet been completed. The proposed 12-month development program includes laboratory characterization, prototype integration, and residential field validation of 20 units, after which the technology would be positioned for partnership and co-development discussions.

Estimated budget for the 12-month program is $50K, covering personnel, hardware, field validation, and dissemination. This indicates an early TRL (approximately 2–4) rather than a commercially deployed product.