MapForest: A Modular Field Robotics System for Forest Mapping and Invasive Species Localization

Sandeep Zachariah; Francisco Yandun; Sachet Korada; Abhisesh Silwal

cs.ROMar 23, 2026

MapForest: A Modular Field Robotics System for Forest Mapping and Invasive Species Localization

Authors: Sandeep Zachariah, Francisco Yandun, Sachet Korada, Abhisesh Silwal

Organizations: Robotics Institute, Carnegie Mellon University, Pittsburgh, USA.

Abstract

Monitoring and controlling invasive tree species across large forests, parks, and trail networks is challenging due to limited accessibility, reliance on manual scouting, and degraded under-canopy GNSS. We present MapForest, a modular field robotics system that transforms multi-modal sensor data into GIS-ready invasive-species maps. Our system features: (i) a compact, platform-agnostic sensing payload that can be rapidly mounted on UAV, bicycle, or backpack platforms, and (ii) a software pipeline comprising LiDAR-inertial mapping, image-based invasive-species detection, and georeferenced map generation. To ensure reliable operation in GNSS-intermittent environments, we enhance a LiDAR-inertial mapping backbone with covariance-aware GNSS factors and robust loss kernels. We train an object detector to detect the Tree-of-Heaven (Ailanthus altissima) from onboard RGB imagery and fuse detections with the reconstructed map to produce geospatial outputs suitable for downstream decision making. We collected a dataset spanning six sites across urban environments, parks, trails, and forests to evaluate individual system modules, and report end-to-end results on two sites containing Tree-of-Heaven. The enhanced mapping module achieved a trajectory deviation error of 1.95 m over a 1.2 km forest traversal, and the Tree-of-Heaven detector achieved an F1 score of 0.653. The datasets and associated tooling are released to support reproducible research in forest mapping and invasive-species monitoring.

Figures & tables

Fig. 1 : System overview. Multi-modal sensing and logging feed a GLIM-based [ 1 ] LiDAR-inertial mapping backbone and a YOLOv8 detector. Detections are fused with the estimated trajectory to produce georeferenced map layers (e.g., GeoTIFF/KML). Optional modules include GNSS priors, and aerial-terrestrial map alignment.

Panel of Figure 2 — Fig. 2 : Sensor payload deployed across multiple carriers, demonstrating platform-agnostic operation.

Source	Train	Val	Test
iNaturalist (web)	700	–	–
Gascola (field)	436	54	56
Hazelwood (field)	330	31	32
Total	1466	85	88

TABLE I : Tree-of-Heaven detection dataset composition and splits (images).

Panel of Figure 4 — Fig. 3 : Likelihood fields used for aerial–terrestrial alignment. Top row: likelihood fields for (a) aerial, (b) terrestrial, and (c) aligned results. Bottom row: the corresponding point-cloud maps for each case.

Site	ToH present	Traversal length (km)	RTK GNSS
Gascola	Yes	1.40	Yes
Bike Trail	Yes	24.10	Yes
Chatham	Yes	1.20	Intermittent
Flagstaff	No	0.80	Yes
Frick Park	No	2.22	Intermittent
CMU Campus	No	0.48	No

TABLE II : Summary of evaluation sites and data collection runs.

Method	Flagstaff
FastLIO	$15.77\pm 13.91$
FasterLIO	$12.83\pm 9.02$
FastLIMO	$18.70\pm 11.59$
GLIM	$8.67\pm 4.77$

TABLE III : Trajectory error (m) w.r.t. RTK GNSS ground truth on the Flagstaff site.

Fig. 4 : Estimated vs. ground-truth DBH for 11 Flagstaff trees (FSCT on reconstructed map vs. tape-measured).

Method	Error (m)
Without GNSS	$8.67\pm 4.77$
Constant-scaling GNSS	$5.80\pm 3.66$
Covariance-aware GNSS	$2.83\pm 2.51$

TABLE IV : SLAM trajectory error (m) for GLIM variants

Panel of Figure 9 — Fig. 5 : Reconstructed maps on the Flagstaff site for two GLIM variants.

Method	Error (m)
Baseline (no robust kernel)	$2.28\pm 2.24$
Robust kernel	$1.95\pm 1.88$

TABLE V : Effect of robust optimization on SLAM accuracy. Trajectory Error (m) for GLIM with and without a robust kernel

Panel of Figure 11 — Fig. 6 : Tree-of-Heaven detections from our YOLOv8 model in a bike trail site, showing predicted bounding boxes and confidence scores for Ailanthus altissima.

Panel of Figure 12 — Fig. 7 : Aerial–terrestrial map alignment results on two patches. Green points show the terrestrial (under-canopy) map and red points show the aerial (above-canopy) map; the strong overlap indicates accurate alignment.

Method	Patch 1		Patch 2		Patch 3
	Trans	Yaw	Trans	Yaw	Trans	Yaw
ICP	1.71	1.49	3.53	2.83	5.55	2.32
Fast Global Registration	4.13	28.07	23.87	9.43	64.79	30.51
Clique	6.00	-53.54	9.62	-75.02	9.37	-81.78
Tensor-MI (our approach)	0.27	0.70	0.73	0.61	0.83	3.92

TABLE VI : Horizontal alignment error (translation in m, yaw in deg) on three patches.

Panel of Figure 14 — Fig. 8 : Geo-referenced Tree-of-Heaven outputs. Top: Point-cloud maps with detections highlighted in red. Bottom: GeoTIFF detection layers overlaid on Google Earth imagery.

Panel of Figure 15 — Fig. 9 : Marked tree used for DBH evaluation. (Left) AprilTag provides a unique ID and reflective tape yields high-intensity LiDAR returns. (Right) Point cloud of the corresponding tree trunk; the pink region indicates high-intensity returns from the reflective tape.

Simultaneous Localization and Mapping

Simultaneous localization and mapping estimates a robot’s motion while building a map of an unknown environment.

See all 197 papers

DTIF: Robust Loop Closure Detection via Delaunay Triangle Topology in Complex Forests
Jul 23, 2026Xin Zhao, Jianping Li, Qin Zou +3Point Cloud Registration Loop Closure Detection
School of Computer Science, Wuhan University, Wuhan 430079, China
TreeLoc++: Robust 6-DoF LiDAR Localization in Forests with a Compact Digital Forest Inventory
Mar 4, 2026Minwoo Jung, Dongjae Lee, Nived Chebrolu +3Robotics LiDAR Localization
Department of Mechanical Engineering, Seoul National University, Seoul, South Korea · Department of Computer Science and Engineering, Indian Institute of Technology Bombay, India · Oxford Robotics Institute, Department of Engineering Science, University of Oxford, Oxford, UK
Integrating national forest inventory, airborne lidar, and satellite imagery for wall-to-wall mapping of forest structure with computer vision
Jun 18, 2026Luke J. Zachmann, David D. Diaz, Vincent A. Landau +8Remote Sensing Forest Aboveground Biomass Estimation
Vibrant Planet Public Benefit Corporation, Truckee, CA, USA

Jul 23, 2026J/K move · Enter open · S save

DTIF: Robust Loop Closure Detection via Delaunay Triangle Topology in Complex Forests

Xin Zhao, Jianping Li, Qin Zou, Fuxun Liang+2

School of Computer Science, Wuhan University, Wuhan 430079, China

Point Cloud Registration Loop Closure Detection

Open paper AskPDF arXiv

Accurate forest inventory and large-scale mapping are essential for ecosystem monitoring and sustainable forest management. Multiple low-cost edge platforms enable efficient large-area data acquisition, but merging independently constructed local maps in GNSS-denied understory environments still requires initialization-free loop closure detection and global registration. This task is challenging because low-cost LiDAR point clouds are sparse and noisy, while repetitive trunk layouts and the lack of distinctive geometric landmarks lead to severe perceptual aliasing and false correspondences. To address these issues, we propose DTIF (Delaunay Triangulation in Forests), a lightweight trunk-topology-based framework for forest loop closure detection and global registration. Tree trunks are first extracted as stable landmarks and encoded using a Delaunay topology for compact scene representation. Candidate submaps are then screened using edge-length and radius statistics, followed by edge--radius consistency verification and strong/weak vertex support aggregation to construct weighted vertex correspondences. Finally, topology-derived reliability weights are incorporated into a decoupled robust pose estimator that separately estimates yaw, horizontal translation, and elevation translation under gravity alignment. Experiments on simulated and real-world forest datasets demonstrate that DTIF achieves accurate registration with low computational overhead, providing a favorable balance among robustness, efficiency, and deployability on resource-constrained edge platforms.

Abstract

Simultaneous Localization and Mapping

Explore similar work

DTIF: Robust Loop Closure Detection via Delaunay Triangle Topology in Complex Forests

TreeLoc++: Robust 6-DoF LiDAR Localization in Forests with a Compact Digital Forest Inventory

Integrating national forest inventory, airborne lidar, and satellite imagery for wall-to-wall mapping of forest structure with computer vision