A field dataset for legged robotics, pairing synchronized multimodal sensing with precise ground truth across 49 missions and varied real-world conditions.

Radar–IMU–LiDAR odometry that adjusts Doppler constraints to the directions in which LiDAR geometry provides weak localization information.

LiDAR odometry that captures small geometric details with oriented height images and concentrates registration on informative parts of the map.

A general factor-graph framework for smooth local motion estimation and globally consistent localization across sensors, reference frames and robot platforms.

Camera, radar and LiDAR measurements are combined to estimate full 3D velocities at individual points on moving objects.

A field comparison of degeneracy mitigation methods for ICP, examining when constraints and regularization improve LiDAR localization.

A unified review of continuous-time estimation, connecting spline and Gaussian-process formulations with robotic applications and open research questions.

An open sensor payload and design study showing how synchronization, calibration and hardware choices affect localization and mapping.

Forest place recognition using attention across LiDAR density images at multiple heights to handle repetitive vegetation and changing conditions.

A diffusion model predicts multiple plausible LiDAR locations in buildings with repeated rooms and floors, using training scans generated from a mesh.

A logistics robot with steerable wheels, articulated legged suspension and autonomous loading for carrying heavy payloads over uneven terrain.

Localizability-aware ICP detects weakly constrained directions and controls the registration update to keep LiDAR localization reliable in difficult geometry.

A coverage planner for orchards and vineyards that respects vehicle maneuverability while finding feasible routes within a chosen computation budget.

An integrated legged-robot system for autonomous forest surveying, combining navigation with online tree detection and measurement.

A team of legged robots combines terrain traversal, mapping and complementary scientific instruments for exploration in planetary analog environments.

Autoencoder and LSTM architectures for learning compact nonlinear models that support both time-series prediction and free-running simulation.

A detailed design and experimental study of a quadcopter that folds and deploys its arms using one onboard actuator.

The original FOLLY prototype introduces automatic arm folding and deployment to improve quadcopter portability and autonomy.