Autonomous Surface Vessel detection and tracking of static and dynamic obstacles

by Hashir Zahir

Obstacle detection and tracking is crucial for Autonomous Surface Vessels (ASVs) to navigate safely in a sea environment with other moving vessels and static obstacles. The Bumblebee ASV has been the culmination of 5 years of continuous effort by Team Bumblebee at the National University of Singapore. Over the years, many key capabilities have been added but obstacle avoidance was one of its key deficiencies of the software stack despite a stable platform with strong controls system. One of the key reasons behind this was the poor obstacle detection and tracking performance. The previously used grid-based approach also caused smeared grids when moving obstacles passed by the ASV, making the grid unusable. Moreover, over the years, the radar was largely left unused for the perception system due to difficulty in integration, lowering the ASV’s capabilities to detect obstacles.

In this project, an algorithm was developed to help the ASV detect and track static and dynamic obstacles. In particular, the focus was on near shore and near port scenarios with higher density of vessel movement as this was the primary deployment of the Bumblebee ASV. To accomplish this, an effective framework to represent static and dynamic obstacles is required. For this project, a combination of a probabilistic grid map for static obstacles and a centroid based tracker approach with state estimation was taken for dynamic obstacles. Camera images, lidar point clouds and radar data were fused to obtain the detections and tracks. The algorithm was evaluated on the Bumblebee ASV at the Republic of Singapore Yacht Club (RSYC) via 2 separate scenarios, where it showed good results in segmenting the static and dynamic obstacles, with a small number of false positives.

Computational Hydrodynamics of an Autonomous Underwater Vehicle (AUV)

by Steven Harta Prawira

This project explores the use of computational method to simulate the motion of an AUV that is acted upon by the force generated by its propulsion system. The primary objective is to study the setup of a CFD simulation that is coupled with free body dynamics available in StarCCM+ with the use of overset meshing technique. In this case, the overset meshing techique is used to model the 6 Degree-of-Freedom motion of the AUV with rotating propeller. This meshing technique allows for the visualization of the physical behaviour of the AUV experiencing various fluid forces, on top of the thrust generated by the propeller. Having the ability to visualize the physical behaviour of the AUV and to track the different parameters associated with it, this project can be used further in designing optimal control system for the AUV and many other purposes.

In achieving the above, the project is sub-divided into a few sub-projects to facilitate incremental learning within the CFD environment and the various techniques that comes with it. Also, instead of using complex AUV geometry, this project uses a simple ellipsoid AUV model as a proof-of-concept before moving further from it. Towards the end, the project also discusses how a PID controller can be interfaced with StarCCM+ for further development of the project. All in all, the project has been a fruitful learning journey to find out the possibilities of integrating CFD simulation and rigid body dynamics with StarCCM+.

Control of an Autonomous Surface Vessel (ASV) and Autonomous Underwater Vehicle (AUV) for Ocean Deployment

by Ng Ren Zhi

Recently, the focus in the marine robotics field has been shifting towards cross-platform systems, with competitions such as the Maritime RobotX Challenge and Shell Ocean Discovery XPrize encouraging multi-platform systems. This paper presents the development process for the propulsion systems and teleoperation of an integrated ASV and AUV system. The main challenge in surveying open sea waters is the ability of the vehicles to station keep and move adeptly against the strong sea currents. A fully-actuated propulsion system is implemented for greater manoeuvrability for the ASV. An Operator Control Station (OCS) is also implemented for teleoperation and status monitoring of the system during deployment. This design is implemented on the Bumblebee Autonomous Underwater Vehicle 3.5 (BBAUV 3.5) and Bumblebee Autonomous Surface Vessel 2.0 (BBASV 2.0), and has been operationally deployed in open sea trials in preparation for the Maritime RobotX Challenge 2018.

3D Object Localization using Forward Looking Sonar (FLS) and Optical Camera via Particle Filter based Calibration and Fusion

by Yaadhav Raaj, Alex John, Tan Jin

Underwater Object Localization is widely used in the industry in Autonomous Underwater Vehicles (AUV), both in sea and lake environments for various applications. Sonars and Cameras are popular choices for this, but each sensor alone poses several problems. Data extraction from Optical Cameras underwater is a challenge due to poor lighting conditions, hazing over large distances and spatio-temporal irradiate (flickering), while Sonars tend to have coarser sensor resolution and a lower signal-to-noise ratio (SNR) making it difficult to extract data. This makes false positives more likely. In this paper, we present a robust method to localize objects in front of an AUV in 3D space, using camera imagery, sonar imagery and odometry information from onboard sensors. This is done through various image processing techniques, and a hybrid sonar/camera particle filter based calibration step and fusion step.

Bumblebee AUV/ASV Development Work

by Alex Philipose John

In this report I present the design and development of core software components that run on board an Autonomous Underwater Vehicle (AUV) and Autonomous Surface Vessel (ASV). To achieve the ultimate goal of demonstrating autonomous launch and recovery (LARS) of the AUV from the ASV, we first develop the navigation and control system. First, a suitable under-water simulation stack is developed and the derivation of the hydrodynamic model is detailed. Second, the simulation stack is used to test the control system and a simple proof of concept neural network based real time PID tuner is developed. This enables the PID controller to adapt its constants to changing environmental conditions, particularly wind. Off the shelf sensors used in inertial navigation is benchmarked and compared. The inertial navigation system used on two separate vehicles was dismantled and re-developed as an Error State Kalman Filter that dynamically adapts to either GPS, Camera, USBL or DVL observations. Real world test results are showcased and analyzed.

Launch and Recovery of Autonomous Underwater Vehicle

by Tey Kee Yeow

In this thesis, the launch and recovery system of an Autonomous Underwater Vehicle (AUV) from an Autonomous Surface Vehicle (ASV) is presented. The primary objective is to determine the feasibility of launching an AUV from an autonomous surface vehicle (ASV) and implement it on the system. The LARS is primarily designed to cater for the Maritime RobotX Challenge but deployable for low sea state use. The AUV used in this system is the Bumblebee 3.0 which was used to participate in International RoboSub Competition 2016.

To meet this objective, industry LARS are examined closely with engineers, reading from paper works, and multiple digital sources. The most feasible method is determined to be using load-bearing tether coupled with the use of a telescopic arm to reduce the effect of splash zone transition. The concept was tested with series of prototype tests and proven to be useful in developing subsequent product. The literature review of LARS, designs, experimental results, and recommendations for future work are presented in this thesis.

Design and Manufacture of a Pick-and-Place Manipulator for Integration within an Autonomous Underwater Vehicle

by Tan Hui Juan Esther

Underwater manipulation systems make it possible to access and perform mechanical works in hostile and hazardous environments where humans cannot enter, such as the deep oceans, icy waters, natural disaster region or a man-made wreckage. They are highly sought after in industries ranging from the Oil and Gas Industry to Search and Recovery, Deep water Archaeology and Marine Science, where they are required to perform tasks such as welding, valve turning and connector plugging, retrieval of fragile corals or recovery of free-floating objects (Ridao, Carreras, Ribas, Sanz, & Oliver, n.d.).

These manipulation systems are typically installed on board an underwater vehicle, notably a Remotely Operated Vehicle (ROV) where tasks are mainly performed under human supervision, or an Autonomous Underwater Vehicle (AUV) where tasks are performed independently of human control. Today’s manipulator systems swings between the extremes of being either too heavy and expensive (Cooney, 2006), or too simple and lacking in functionalities. Also, the multi-purpose usage of manipulators in various facets demands for a robust and versatile gripping system.

Henceforth, this project will be on the research, design and fabrication of a manipulator, which serves a dual role of meeting the industry needs in manipulation systems and also for competitive use on the Bumblebee Autonomous Underwater Vehicle (AUV). In this thesis, the mechanical design and integration of a manipulator is presented, with versatile gripping achieved using a Jamming Gripper technique and precise positioning achieved via pneumatic actuations and high torque servo rotation.

Electrical System for Autonomous Underwater Vehicle and Autonomous Surface Vehicle

by Vanessa Cassandra

Team Bumblebee has been developing Bumblebee Autonomous Underwater Vehicle (AUV) version 3.0 since August 2014 and it finally made its debut during Singapore Autonomous Underwater Vehicle Challenge in March 2016. From the initial stage of conception until the deployment of the vehicle, there are problems that were not foreseen and only arose during operation. Improvements need to be made to make the system more robust and reliable.

Team Bumblebee has also started the development of an Autonomous Surface Vehicle (ASV) and it will be deployed during Maritime RobotX Challenge in December 2016.

This paper discusses the improvements that can be made for Bumblebee AUV version 3.0 to ensure a stable system and also about the electrical architecture of the ASV.

Direction Finding Arrays in Acoustic and Electromagnetic Domains

by Huan Yongchang

In this thesis, we consider the problem of direction finding in the acoustic and electromagnetic domain. In the electromagnetic domain, Maximum Ratio Combining (MRC) and Median Filtering have been proposed as the appropriate wideband DOA techniques that can be used in a frequency selective channel and in the presence of narrowband interferences. Simulation results have shown that these proposed techniques yield a more accurate Direction of Arrival (DOA) estimation in comparison to Equal Gain Combining (EGC). The narrowband MUSIC algorithm has also been tested in the antenna test chamber at Temasek Laboratories. Implementation issues involving the calibration of multiple software-defined radios have been resolved using a calibration signal for phase estimation and phase compensation. In the acoustic domain, sophisticated direction finding algorithms have been developed on an Autonomous Underwater Vehicle (AUV) to complete the acoustic localization task in the Singapore AUV Challenge (SAUVC). Results collected from extensive pool tests indicate that the high-resolution MUSIC algorithm presents a more robust solution compared to the Time Difference of Arrival (TDOA) algorithm.

Underwater Real-Time Object Recognition and Tracking for Autonomous Underwater Vehicle

by Tan Soon Jin

Design and Implementation of a Backplane Electrical System for an Autonomous Underwater Vehicle

by Hoang The Huan

The electrical system of Bumblebee Autonomous Underwater Vehicle version 2.0 has been tested extensively during pool tests and competitions from March 2014 till now. Several problems have been identified and analyzed for future improvements. As part of the development for Bumblebee Autonomous Underwater Vehicle version 3.0, an electrical backplane system is proposed which will incorporate changes to fix the existing problems, reduce the amount of wiring, enhancing reliability and testability.

This thesis discusses the existing problems in the electrical system of Bumblebee version 2.0, presents possible solutions for those problems as well as new features. It describes the development and testing methodologies of the solutions and new features. Next, the design considerations for incorporating the solutions and new features on the backplane are discussed. Finally, it presents the final electrical backplane product.

Real Time Implementation of Passive Acoustic Source Localization System on an Autonomous Underwater Vehicle

by Joshua Yow

This project focuses on the problem of localizing an underwater acoustic pinger in a reverberant environment under the constraints of a real time implementation on an Autonomous Underwater Vehicle.

This paper proposes a novel algorithm to make use of both distinctive features of an acoustic ping, the rising edge of the ping and the phase of the sine wave portion, to identify the location of the source. A combination of a step detector working on the envelope of the signal and the high resolution MUSIC algorithm is used to gain a more accurate and robust estimate of source location than can be obtained from each of these techniques alone. The algorithm is implemented on MATLAB and verified using data recorded by the hardware on BBAUV.

An implementation of the proposed algorithm is carried out on the Bumblebee Autonomous Underwater Vehicle (BBAUV). The resulting implementation is able to maintain the performance of the algorithm while maintaining low latency under a resource-constrained embedded hardware system.