1 December 2018 in Water Practice & Technology
Inspection of deep tunnel networks is extremely challenging due to their inaccessibility, and them being an unknown and potentially hazardous environment. Unmanned aerial vehicles (UAVs) provide a viable alternative for access, and are unaffected by debris or sewer flow. However, commercial UAVs are designed for high altitude aerial imagery and are not appropriate for short-range detailed imaging of tunnel surfaces. In addition, autonomous flight is usually achieved using GPS, which is not available underground. This paper presents the design and development of a smart UAV platform, Surveyor with Intelligent Rotating Lens (SWIRL), customized for autonomous operation in tunnels. It can capture high resolution images for subsequent image processing, and defect detection and classification. An innovative rotating system enables undistorted imaging of the tunnel's inner circumference surface using a single camera. The proposed location method using limited data resulted in substantial unit weight and power consumption reductions, compared to existing systems, making more than 35 minutes of autonomous flight possible.
14 January 2019 in IEEE Robotics and Automation Letters
DSDanial Sufiyan bin Shaiful
Danial Sufiyan bin Shaiful
In this paper, the design optimisation and localisation with sparse sensing array of range-based sensors is presented. This optimisation focuses on an optimised sparse sensing array for localisation and autonomous navigation in tunnel environments with right rectangular prism geometry and the autonomous ascent and descent in vertical cylindrical shafts. A optimisation is conceived to optimise sensor configuration to traverse both tunnels and shafts with constraints from the field of the view of the camera system and the mechanical layout. The optimised sensor configuration solved using a genetic algorithm performed on average 21 times better than configurations that were placed heuristically. The optimised design and the localisation algorithm is implemented on SWIRL: Surveyor with Intelligent Rotating Lens. The proposed localisation method with sparse sensing results in at least four-fold reductions in the weight and eight-fold reduction in power consumption of the sensing system compared to existing ones. SWIRL was autonomously flown in a series of experiments, consisting of tests in shaft and tunnels in both indoor mock-up and field environment. Lastly, in an endurance test in a mock-up shaft, SWIRL achieved over 35 minutes of autonomous flight using the proposed sensing placement and scheme, making extended aerial robot navigation and inspection of deep hazardous tunnels possible.
27 June 2022 in IEEE Robotics and Automation Letters
In this paper, we propose a ground-based monocular UAV localisation system that detects and localises an LED marker attached to the underside of a UAV. Our system removes the need for extensive infrastructure and calibration unlike existing technologies such as UWB, radio frequency and multi-camera systems often used for localisation in GPS-denied environment. To improve deployablity for real-world applications without the need to collect extensive real dataset, we train a CNN on synthetic binary images as opposed to using real images in existing monocular UAV localisation methods, and factor in the camera’s zoom to allow tracking of UAVs flying at further distances. We propose NoisyCutout algorithm for augmenting synthetic binary images to simulate binary images processed from real images and show that it improves localisation accuracy as compared to using existing salt-and-pepper and Cutout augmentation methods. We also leverage uncertainty propagation to modify the CNN’s loss function and show that this also improves localisation accuracy. Real-world experiments are conducted to evaluate our methods and we achieve an overall 3D RMSE of approximately 0.41 m.
29 November 2021 in IEEE Sensors Journal
DSDanial Sufiyan Bin Shaiful
Danial Sufiyan Bin Shaiful
This paper presents the Flying Lidar (Flydar) aerial robot for simultaneous localization and mapping (SLAM). The Flydar integrates a single laser and capitalizes on the rotating dynamics of a nature-inspired aerial robot for omnidirectional scanning. This paper presents a 2.5D SLAM approach using the Flydar and does not assume that the Flydar’s hovering plane, and thus the laser scan, is parallel to the ground plane, which is a key assumption in our previous work. A quaternion-based filter performs active correction of the Flydar’s laser scan due to its transient attitude. A dual-accelerometer approach was incorporated to estimate the Flydar’s scan rate during high angular rates flight beyond the gyroscope dynamic range. The proposed attitude filter output was experimentally evaluated statically on a bench-top setup and dynamically in flight, with an rms inclination estimate error of up to 1.1° and 0.38° respectively. The attitude-corrected lidar scan was used to estimate the robot pose for 2.5D SLAM. The 2.5D SLAM was experimentally validated on the Flydar and demonstrated to be superior to pure 2D SLAM in loop closure, position estimate drift, and rms error. Significantly, the 2.5D SLAM using the Flydar reports a low rms Euclidean error of up to 0.083 m, which is a 32.0% error reduction compared with our previous work, which uses 2D SLAM.