An increasing number of robotic systems feature multiple inertial measurement units (IMUs). Due to competing objectives-either desired vicinity the center gravity when used in controls, or an unobstructed field view integrated a sensor setup with exteroceptive for ego-motion estimation-individual IMUs are often mounted at considerable distance. As result, they sense different accelerations platform is subjected rotational motions. In this work, we derive method spatially calibrating single...

Achieving accurate, high-rate pose estimates from proprioceptive and/or exteroceptive measurements is the first step in development of navigation algorithms for agile mobile robots such as Unmanned Aerial Vehicles (UAVs). In this paper, we propose a decoupled Graph-Optimization based Multi-Sensor Fusion approach (GOMSF) that combines generic 6 Degree-of-Freedom (DoF) visual-inertial odometry poses and 3 DoF globally referenced positions to infer global robot real-time. Our casts fusion...

Abstract We present the development process behind AtlantikSolar , a small 6.9 kg hand‐launchable low‐altitude solar‐powered unmanned aerial vehicle (UAV) that recently completed an 81‐hour continuous flight and thereby established new endurance world record for all aircraft below 50 mass. The goal of our work is to increase usability such robotic by maximizing their perpetual robustness meteorological deteriorations as clouds or winds. energetic system models design methodology, implement...

This paper presents the design of small-scale hand-launchable solar-powered AtlantikSolar UAV, summarizes flight results a continuous 28-hour that demonstrated AtlantikSolar's capability for energetically perpetual flight, and offers model-based verification performance an outlook on energetic margins can be provided towards given today's UAV technology. is 5.6m-wingspan 6.9kg mass low-altitude long-endurance was designed to provide endurance at geographic latitude 45N in 4-month window...

We propose a novel scoring concept for visual place recognition based on nearest neighbor descriptor voting and demonstrate how the algorithm naturally emerges from problem formulation. Based observation that number of votes matching places can be evaluated using binomial distribution model, loop closures detected with high precision. By casting into probabilistic framework, we not only remove need commonly employed heuristic parameters but also provide powerful score to classify...

Abstract Large‐scale aerial sensing missions can greatly benefit from the perpetual endurance capability provided by high‐performance low‐altitude solar‐powered unmanned vehicles (UAVs). However, today these UAVs suffer small payload capacity, low energetic margins, and high operational complexity. To tackle problems, this paper presents four individual technical contributions integrates them into an existing UAV system: First, a lightweight power‐efficient day/night‐capable system is...

Full autonomy for fixed-wing unmanned aerial vehicles (UAVs) requires the capability to autonomously detect potential landing sites in unknown and unstructured terrain, allowing self-governed mission completion or handling of emergency situations. In this letter, we propose a perception system addressing challenge by detecting based on their texture geometric shape without using any prior knowledge about environment. The proposed method considers hazards within region such as terrain...

This paper presents the design of small-scale hand-launchable solar-powered AtlantikSolar UAV, summarizes flight results a continuous 28-hour that demonstrated AtlantikSolar's capability for energetically perpetual flight, and offers model-based verification performance an outlook on energetic margins can be provided towards given today's UAV technology. is 5.6m-wingspan 6.9kg mass low-altitude long-endurance was designed to provide endurance at geographic latitude 45N in 4-month window...

In this paper, we propose a resource-efficient approach to provide an autonomous UAV with on-board perception method detect safe, hazard-free landing sites during flights over complex 3D terrain. We aggregate measurements acquired from sequence of monocular images by Structure-from-Motion into local, robot-centric, multi-resolution elevation map the overflown terrain, which fuses depth according their lateral surface resolution (pixel-footprint) in probabilistic framework based on concept...

This paper proposes a computationally efficient method to estimate the time-varying relative pose between two visual-inertial sensor rigs mounted on flexible wings of fixed-wing unmanned aerial vehicle (UAV). The estimated poses are used generate highly accurate depth maps in real-time and can be employed for obstacle avoidance low-altitude flights or landing maneuvers. approach is structured as follows: Initially, wing model identified by fitting probability density function measured...

Accurate and robust real-time map generation onboard of a fixed-wing UAV is essential for obstacle avoidance, path planning, critical maneuvers such as autonomous take-off landing. Due to the computational constraints, required robustness reliability, it remains challenge deploy with an online-capable, accurate framework. While photogrammetric approaches have underlying assumptions on structure view camera, generic simultaneous localization mapping (SLAM) are computationally demanding. This...

In this paper, we propose a visual-inertial framework able to efficiently estimate the camera poses of non-rigid trinocular baseline for long-range depth estimation on-board fast moving aerial platform. The time-varying is based on relative inertial measurements, photometric pose optimizer, and probabilistic wing model fused in an efficient Extended Kalman Filter (EKF) formulation. estimated measurements can be integrated into geo-referenced global map render reconstruction environment...

This paper presents a framework for the localization of Unmanned Aerial Vehicles (UAVs) in unstructured environments with help deep learning. A real-time rendering engine is introduced that generates optical and depth images given six Degrees-of-Freedom (DoF) camera pose, model, geo-referenced orthoimage, elevation map. The embedded into learning-based six-DoF Inverse Compositional Lucas-Kanade (ICLK) algorithm able to robustly align rendered real-world image taken by UAV. To learn alignment...

Use of low-cost depth sensors, such as a stereo camera setup with illuminators, is particular interest for numerous applications ranging from robotics and transportation to mixed augmented reality. The ability quantify noise crucial these applications, e.g., when the sensor used map generation or develop scheduling policy in multi-sensor setup. Range error models provide uncertainty estimates help weigh data correctly instances where range measurements are taken different vantage points...

In this paper, we present our deep learning-based human detection system that uses optical (RGB) and long-wave infrared (LWIR) cameras to detect, track, localize, re-identify humans from UAVs flying at high altitude. each spectrum, a customized RetinaNet network with ResNet backbone provides detections which are subsequently fused minimize the overall false rate. We show by optimizing bounding box anchors augmenting image resolution number of missed altitudes can be decreased over 20...