Exploring the Potential of Multi-modal Sensing Framework for Forest Ecology

TL;DR

This paper discusses a multi-modal sensing framework utilizing drones and ground sensors to improve data collection in forest ecology, addressing accessibility, disturbance, and retrieval challenges in canopy environments.

Contribution

It introduces a novel multi-modal sensing approach combining aerial and ground sensors to enhance data collection efficiency and reduce risks in forest environments.

Findings

Swarm drones can navigate canopy autonomously and avoid collisions.

Multi-modal sensors improve data coverage in hard-to-reach forest areas.

The framework reduces human risk and resource expenditure in forest data collection.

Abstract

Forests offer essential resources and services to humanity, yet preserving and restoring them presents challenges, particularly due to the limited availability of actionable data, especially in hard-to-reach areas like forest canopies. Accessibility continues to pose a challenge for biologists collecting data in forest environments, often requiring them to invest significant time and energy in climbing trees to place sensors. This operation not only consumes resources but also exposes them to danger. Efforts in robotics have been directed towards accessing the tree canopy using robots. A swarm of drones has showcased autonomous navigation through the canopy, maneuvering with agility and evading tree collisions, all aimed at mapping the area and collecting data. However, relying solely on free-flying drones has proven insufficient for data collection. Flying drones within the canopy generates loud noise, disturbing animals and potentially corrupting the data. Additionally, commercial drones often have limited autonomy for dexterous tasks where aerial physical interaction could be required, further complicating data acquisition efforts. Aerial deployed sensor placement methods such as bio-gliders and sensor shooting have proven effective for data collection within the lower canopy. However, these methods face challenges related to retrieving the data and sensors, often necessitating human intervention.