Automating Aircraft Scanning For Inspection With A UAV And Reinforcement Learning Technique

Visual inspections of aircraft exterior surface are usually required in aircraft maintenance routine. It becomes a trend to use mobile robots equipped with sensors to perform automatic inspections as a replacement of manual inspections which are time-consuming and error-prone. The sensed data such as images and point cloud can be used for further defect characterization leveraging the power of machine learning and data science. In such a robotic inspection procedure, a precise digital model of the aircraft is required for planning the inspection path, however, the original CAD model of the aircraft is often inaccessible to aircraft maintenance shops. Thus, sensors such as 3D Laser scanners and RGB-D (Red, Green, Blue, and Depth) cameras are used because of their capability of generating a 3D model of an interested object in an efficient manner. This paper presents a two-stage approach of automating aircraft scanning with a UAV (Unmanned Aerial Vehicle) equipped with an RGB-D camera for reconstructing a digital replica of the aircraft when its original CAD model is not available. In the first stage, the UAVcamera system follows a predefined path to quickly scan the aircraft and generate a coarse model of the aircraft. Then, a full-coverage scanning path is computed based on the coarse model of the aircraft. In the second stage, the UAV-Camera system follows the computed path to closely scan the aircraft for generating a dense and precise model of the aircraft. We solved the Coverage Path Planning (CPP) problem for the aircraft scanning using Monte Carlo Tree Search (MCTS) which is a reinforcement learning technique. We also implemented the Max-Min Ant System (MMAS) strategy, a population-based optimization algorithm, to solve the CPP problem and demonstrate the effectiveness of our approach.

Seasonal Prediction of Tropical Cyclones over the North Atlantic and Western North Pacific

In this study, Tropical Cyclones (TC) over the Western North Pacific (WNP) and North Atlantic (NA) basins are analysed in seasonal forecasting models from five European modelling centres. Most models are able to capture the observed seasonal cycle of TC frequencies over both basins; however, large differences for numbers and spatial track densities are found. In agreement with previous studies, TC numbers are often underestimated, which is likely related to coarse model resolutions. Besides shortcomings in TC characteristics, significant positive skill (deterministic and probabilistic) in predicting TC numbers and accumulated cyclone energy is found over both basins. Whereas the predictions of TC numbers over the WNP basin are mostly unreliable, most seasonal forecast provide reliable predictions for the NA basin. Besides positive skill over the entire NA basin, all seasonal forecasting models are skillful in predicting the interannual TC variability over a region covering the Caribbean and North American coastline, suggesting that the models carry useful information, e.g. for adaptation and mitigation purposes ahead of the upcoming TC season. However, skill in all forecast models over a smaller region centred along the Asian coastline is smaller compared to their skill in the entire WNP basin.

A physics perspective on lidar data assimilation for mobile robots

This paper presents a new algorithm for lidar data assimilation relying on a new forward model. Current mapping algorithms suffer from multiple shortcomings, which can be related to the lack of clear forward model. In order to address these issues, we provide a mathematical framework where we show how the use of coarse model parameters results in a new data assimilation problem. Understanding this new problem proves essential to derive sound inference algorithms. We introduce a model parameter specifically tailored for lidar data assimilation, which closely relates to the local mean free path. Using this new model parameter, we derive its associated forward model and we provide the resulting mapping algorithm. We further discuss how our proposed algorithm relates to usual occupancy grid mapping. Finally, we present an example with real lidar measurements.

Towards monitoring CO₂ source-sink distribution over India via inverse modelling: Quantifying the fine-scale spatiotemporal variability of atmospheric CO₂ mole fraction

The prospect of improving the estimates of CO2 sources and sinks over India through inverse methods calls for a comprehensive atmospheric monitoring system involving atmospheric transport models that make a realistic accounting of atmospheric CO2 variability. In the context of expanding atmospheric CO2 measurement networks over India, this study aims to investigate the importance of a high-resolution modelling framework to utilize these observations and to quantify the uncertainty due to the misrepresentation of fine-scale variability of CO2 in the employed model. The spatial variability of atmospheric CO2 is represented by implementing WRF-Chem at a spatial resolution of 10 km × 10 km. We utilize these high-resolution simulations for sub-grid variability calculation within the coarse model grid at a horizontal resolution of one degree (about 100 km). We show that the unresolved variability in the coarse model reaches up to a value of 10 ppm at the surface, which is considerably larger than the sampling errors, even comparable to the magnitude of mixing ratio enhancements in source regions. We find a significant impact of monsoon circulation in sub-grid variability, causing ~3 ppm average representation error between 12–14 km altitude ranges in response to the tropical easterly jet. The cyclonic storm Ockhi during November 2017 generates completely different characteristics in sub-grid variability than the rest of the period, whose influence increases the average representation error by ~1 ppm at the surface. By employing a first-order inverse modelling scheme using pseudo observations from nine tall tower sites over India and a constellation of satellite instruments, we show that the Net Ecosystem Exchange (NEE) flux uncertainty solely due to unresolved variability is in the range of 6.3 to 16.2 % of the total NEE. We illustrate an example to test the efficiency of a simple parameterization scheme during non-monsoon periods to capture the unresolved variability in the coarse models, which reduces the bias in flux estimates from 9.4 % to 2.2 %. By estimating the fine-scale variability and its impact during different seasons, we emphasise the need for implementing a high-resolution modelling framework over the Indian subcontinent to better understand processes regulating CO2 sources and sinks.

In-Depth Comparison of Dry Particle Coating Processes Used in DPI Particle Engineering

High-shear mixer coatings as well as mechanofusion processes are used in the particle-engineering of dry powder inhalation carrier systems. The aim of coating the carrier particle is usually to decrease carrier–drug adhesion. This study comprises the in-depth comparison of two established dry particle coating options. Both processes were conducted with and without a model additive (magnesium stearate). In doing so, changes in the behaviour of the processed particles can be traced back to either the process or the additive. It can be stated that the coarse model carrier showed no significant changes when processed without additives. By coating the particles with magnesium stearate, the surface energy decreased significantly. This leads to a significant enhancement of the aerodynamic performance of the respective carrier-based blends. Comparing the engineered carriers with each other, the high-shear mixer coating shows significant benefits, namely, lower drug–carrier adhesion and the higher efficiency of the coating process.

Automating Aircraft Exterior Inspection With Robotics and Machine Learning Technologies

Most aircraft exterior inspections require human workers to visually detect defects such as dents, cracks, leaking, broken or missing parts, etc., and manually measure the parameters of the identified defects, which is a time-consuming process and it is also error-prone if the human inspector is not fully focused. This situation can be alleviated by the advance in computer vision and robotics for saving time and relieving human workers from such repetitive and stressful tasks. However, the challenge of automated robotic inspection for aircraft exterior remains due to the very large inspected area, full-coverage requirement and sometimes insufficient digital model available for planning the inspection path. This paper presents a two-stage approach to automate visual inspection of aircraft exterior surface in a static environment such as a MRO shop by a mobile manipulator equipped with a consumer-grade RGB-D camera following an optimal inspection trajectory learned from a low-resolution point cloud model of the aircraft when a proper CAD model of the aircraft is unavailable. In the first stage, a low-cost RGB-D camera is used to acquire a coarse point cloud model of the whole or interested area of the aircraft. In the second stage, a full coverage inspection path is learned based on the coarse model using a reinforcement learning process, which is our focus of this work.

Parareal with a learned coarse model for robotic manipulation

A key component of many robotics model-based planning and control algorithms is physics predictions, that is, forecasting a sequence of states given an initial state and a sequence of controls. This process is slow and a major computational bottleneck for robotics planning algorithms. Parallel-in-time integration methods can help to leverage parallel computing to accelerate physics predictions and thus planning. The Parareal algorithm iterates between a coarse serial integrator and a fine parallel integrator. A key challenge is to devise a coarse model that is computationally cheap but accurate enough for Parareal to converge quickly. Here, we investigate the use of a deep neural network physics model as a coarse model for Parareal in the context of robotic manipulation. In simulated experiments using the physics engine Mujoco as fine propagator we show that the learned coarse model leads to faster Parareal convergence than a coarse physics-based model. We further show that the learned coarse model allows to apply Parareal to scenarios with multiple objects, where the physics-based coarse model is not applicable. Finally, we conduct experiments on a real robot and show that Parareal predictions are close to real-world physics predictions for robotic pushing of multiple objects. Code (https://doi.org/10.5281/zenodo.3779085) and videos (https://youtu.be/wCh2o1rf-gA) are publicly available.

THREE-DIMENSIONAL PATH PLANNING OF UAVS IMAGING FOR COMPLETE PHOTOGRAMMETRIC RECONSTRUCTION

Lightweight unmanned aerial vehicles (UAVs) have been widely used in image acquisition for 3D reconstruction. With the availability of compact and high-end imaging sensors, UAVs can be the platform for precise photogrammetric reconstruction. However, the completeness and precision of complex environment or targets highly rely on the flight planning due to the self-occlusion of structures. Flight paths with back-and-forth pattern and nadir views will result in incompleteness and precision loss of the 3D reconstruction. Therefore, multiple views from different directions are preferred in order to eliminate the occlusion. We propose a 3D path planning method for multirotor UAVs aiming at capturing images for complete and precise photogrammetric 3D reconstructions. This method takes the coarse model from an initial flight as prior knowledge and estimates its completeness and precision. New imaging positions are then planned taking photogrammetric constraints into account. The real-world experiment on a ship lock shows that the proposed method can acquire a more complete result with similar precision compared with an existing 3D planning method.