SBIR Phase I: Robotic Forest Inventory and Mapping

点击翻译按钮获取中文摘要

The broader impact/commercial potential of this Small Business Innovation (SBIR) Phase I project is modernizing monitoring systems for the forestry industry, environmental sustainability, and nature conservation purposes. The estimated commercial potential on the $325 B global forestry industry is an additional $3.4 B of unlocked value through faster, cheaper, and more accurate inventory systems. Warmer temperatures have contributed to an explosion in pest epidemics that have destroyed over 120 million acres of timberland in the US since 1998. In addition, these temperatures and dead trees have exacerbated wildland fire, with annual economic damage estimated to be $350 B. The societal impact of these problems is pervasive, as smoke plumes can drift for thousands of miles and adversely affect human health and environmental pollution. One fundamental requirement to address these problems is an automated forest monitoring system, as current systems still heavily rely on manual measurements. This project will enhance scientific and technological understanding by developing autonomous and large-scale semantic mapping robotic systems for dense, natural forests to tackle these broader high-impact problems. In addition, this project will provide high-tech career opportunities in rural communities by training skilled operators to develop, deploy, and control the robot teams in forests. This Small Business Innovation (SBIR) Phase I project will develop the first commercially viable automated timber cruise to estimate forest volume from below the canopy level. The forestry industry still relies on manual measurements because, due to fundamental technical challenges, the technology to autonomously measure tree sizes under the canopy over long distances does not exist. This project will overcome two limiting challenges: 1) Robust Autonomy Challenge: No one has achieved robust autonomy in truly 3D, unstructured, GPS-denied environments where manual control (teleoperation) is not feasible; and 2) Large-Scale Semantic Mapping Challenge: No one has attempted semantic mapping at the scale and accuracy proposed, as most demonstrations have been for a few object instantiations and without the need for precise measurement. To tackle these challenges, the project anticipates three technical results: 1) Real-time tree detection to robustly detect trees in challenging conditions; 2) Continuous-Time Semantic Simultaneous Localization and Mapping to precisely model trees over vast distances; and 3) Fast Online Motion Planning with Deep Model Predictive Control to robustly navigate unmanned aerial vehicles in cluttered forest environments.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.