Project topics and materials
Get Complete Project Material File(s) Now! »
Tethers and Deformable Objects Management
This section focuses on existing strategies of tether management for underwater and mobile robots in general. Since the tether can also be seen as a deformable object, this section will also address some techniques of deformable objects handling that could be applied to tether management.
Underwater Applications
This subsection deals with tether management systems (TMS) that are typically used for work-class and middle-class robots. As depicted in Figure 1.6, the TMS is placed at the junction between the end of the umbilical and the beginning of the tether (Hawkes and Jeffrey, 1987; Abel, 1994). This junction is made by a simple clump/depressor weight or by a vehicle handling system (a protection cage or a top hat mechanism), as it is shown in Figure 1.7. The use of a clump weight is prevalent in the observation-class category. The clump weight absorbs the cross-section drag of the current in a passive way, relieving the submersible of the umbilical drag from the surface to the working depth. Therefore, the ROV only needs to drag a portion of the tether length between the clump weight and the vehicle. Similarly to clump weights, cages also function as a negatively buoyant anchor to overcome the drag imposed by the umbilical, and additionally, they are used to protect the vehicle against abrasions and deployment damage due to the instability of most surface vessels. Clump weights and cages can be used without a TMS for reasons of simplicity. The addition of a TMS is considered by operators as a similar complexity of having a second ROV concurrently in the water.
Terrestrial and Aerial Applications
Tether and deformable objects handling is a challenge also found in some fields of application other than underwater environment. In the context of planetary exploration, tethers are used to connect a base station to a rover dedicated to explore hard-to-reachterrains (Krishna et al., 1997; McGarey et al., 2016b; Brown et al., 2018). In such applications, the tether works as both a mechanical support and a transmission link of power and data. In Iqbal et al. (2008), the tether also serves as a guide line so that the rover can autonomously track it in order to find the way back to the base station. Frontal infra-red sensors mounted on the front of the rover are used to track the tether while it is rewound into the rover reel. The tether lays on the ground and entanglement with obstacles was an issue unfortunately not addressed in this paper. When the terrain is to steep, the tether can be kept under tension and function as an anchor for the rover to explore planetary craters. In Tsai et al. (2013), the problem of vision-based tether-assisted docking of a daughter rover to its base station (central module) is addressed. The docking strategy relies on an algorithm running on the rover that uses stereo cameras to detect fiducials markers mounted on the central module and then estimate its relative pose. The rover uses a motion planner to position and orient itself such that it aligns its arm with the docking cone to be retracted back to the mother station at the end of the mission. Proper tether tension is maintained during motion thanks to the vision-based relative pose estimation that enables to geometrically calculate the instantaneous exposed tether length from the tip of the rover’s arm to the mother station. The fact that the tether itself is not used to provide information for the docking process limits the robot workspace since the docking station should permanently stay in the rover’s field of view. Objects in the environment can also be seen as anchor points instead of obstacles. In Vishnu et al. (2008) and Rajan et al. (2016), force sensors were used to detect tether anchor points. On-board tilt sensors together with the exposed tether length measurement allowed to estimate the robot position in a line-of-sight. The disentanglement technique proposed was based on an algorithm of tether-following that allows the robot to bypass the obstacles without being clung to them. Experiments with two wheeled robots serially linked together by a tether to a base station were carried out. The robots were organized in a leader/slave configuration and performed tether disentanglement around a single obstacle. Tether aided localization was then further improved by gathTethers and Deformable Objects Management 25 ering multiple sensory feedback. In Murtra and Tur (2013), wheel odometry, tether length measurements, and an IMU (inertial measurement unit) were used to localize a pipe-inspection robot, where tether length was used to limit uncertainty about the distance traveled in a pipe. This work was only concerned with localization of the robot and no attempt was made to detect and map tether contact points (i.e., the anchor point) within the pipe. In McGarey et al. (2016a) and McGarey et al. (2017), the pose of a tethered robot and the positions of the intermediate tether anchor points were estimated using tether length, bearing-to-anchor angle, and odometry gathered along the trajectory. The objective of this work was the formulation of a tethered simultane- ous localization and mapping (TSLAM) problem whose solution would allow the robot to safely return to its base along an outgoing trajectory while avoiding tether entanglement. The motivation was to use TSLAM as a building block to aid conventional, camera and laser-based approaches to SLAM, which tend to fail in dark and or dusty environments.
Multiple mobile robots can be used for cooperative transportation of large objects, that can be rigid (Huntsberger et al., 2004) or flexible. In Echegoyen et al. (2010), three terrestrial robots were used to transport a flexible hose modeled by Geometrically Exact Dynamic Splines (GEDS). A camera with a global view of the scene was used to detect the robots and the hose by color segmentation. The leader robot pursued a predefined trajectory while the follower-robots’ command velocities were computed from a fuzzyheuristic local controller. The curvature of the hose segment in front of each robot was used as a visual feature by the controller that was regulated in order to avoid the hose of being taut.
Aerial robots were also used in the transportation of flexible cables (Estevez and Gra˜na, 2015; Estevez et al., 2015). The criteria for the transportation was that all the robots should carry the same load. Thus, the hose was modeled by a catenary and the IMU of the robots were used to estimate their relative height, which was regulated by a PID controller with the aim of evenly distribute the hose weight among the robots. Tethers are used by unmanned aerial vehicles (UAV) for long-term missions with high-speed communication between the operator and the robot in an wide range of applications, such as robot-assisted search and rescue (Pratt et al., 2008) and coastal and environmental remote sensing (Klemas, 2015). Most published work in the field of tethered flight are restricted to the taut tether case. In these systems no tether management is employed while the UAV maintains tension. Otherwise, a winch mechanism placed in a fixed or mobile base station continuously reels in any slack tether length (Nicotra et al., 2017). The dynamics and control of a quadrotor unmanned aerial vehicle connected to a fixed point on the ground via a tether was addressed in Lee (2015). The tether was considered as a collection of an arbitrary number of rigid links that are serially interconnected via ideal ball joints. The motion equations of the system (robot and tether) were obtained from Hamilton’s principle. A control law based on inertial sensing was designed through feedback linearizion and used the UAV pose with the aim of maintaining a desired tether state (orientation and tension). This control law was evaluated in two numerical examples of station keeping and predefined path tracking. A reactive tether management approach, where the tension and departure angle are measured at the winch, showed moderate winch controller results that can be further improved by incorporating knowledge of the UAV position (Zikou et al., 2015). Another work used the measured tether length, tension, and departure angle as a means for non-GPS position estimation of the UAV based on a catenary cable model (Kiribayashi et al., 2017). In Talke et al. (2018), the quasi-static catenary curve of a hanging tether between an essentially stationary UAV and a small unmanned surface vehicle (USV) is investigated and characterized. The objective is to develop in a near future a winch controller that could maintain the tether slack and compensate the USV heave in order to minimize the tether traction and the risk of being in contact with the water surface. A multi-agent extension of the tethered aerial robot problem was investigated through numerical simulations in Tognon and Franchi (2015), where a chain of two flying robots was considered. The goal was to independently control the elevation angles of the two tether segments as well as their internal stress.
Our Scientific Focus: Vision Servoing of a Pair of Robots in a Chain of Mini-ROVs
The long term objective of this project is to design an active tether management solution for mini-ROV missions of long range displacements within cluttered environments and shallow waters. We choose to control the tether shape by adding several robots linked together all along it. We call this concept the chain of mini-ROVs (see Figure 1.9). The robots play the role of actuators and change the whole tether shape depending on the situation at hand. Our strategy is to avoid any contact with obstacles. If the depth is too narrow, it would be better to maintain the tether more taut in order to prevent it from dragging on the seabed. Otherwise, if the environment is more spacious, the tether can be more slack in order to give more freedom of motion to the robots. The robots that compose the chain are compact, lightweight and with a limited sensory payload. Thus, we choose to investigate the use of the onboard camera to perceive and estimate a parameterized geometric model of the cable. The main focus of this thesis is to manage the tether linking two robots through visual feedback. These robots are named leader, for the front vehicle, and follower, for the rear vehicle. The thesis objective can be therefore summarized by the following question:
Robots Model and Configuration
Small teleoperated underwater robots, classified as mini-ROVs, are preferable to the execution of missions in confined environments or in shallow coastal waters thanks to their high maneuverability and small size. In this thesis, we choose to use the 6- thruster model of mini-ROVs inspired from the BlueROV experimental platform from Blue Robotics. There are two models, the first one is related to the BlueROV1 and the second one to the BlueROV2 (see Figure 2.3). A comparison between both robots is presented in table 2.1. The architecture of the robots is composed of a main board Raspberry Pi 3 Model Broadcom BCM2837 64bit, 1GB of RAM), which is connected via USB to a general purpose microcontroller, namely a Pixhawk autopilot microcontroller that was designed for aerial drones. The microcontroller has the following internal sensors: a 3-axis accelerometer and gyroscope (Invensense MPU6000), a 3-axis gyroscope (ST Micro L3GD20H), a 3D e-compass (ST Micro LSM303D) and a barometer (MEAS MS5611). Both robots are equipped with an external pressure sensor (the Bar30 model from BlueRobotics) that is connected to the Pixhawk and used to measure water pressure. The Pixhawk controls the six ESCs (electronic speed controllers) that manage the power and rotation velocity commands sent to the thrusters. Each robot is also equipped with a Raspberry Pi Camera. The onboard computer communicates with a remote workstation via a Fathom-X Tether Interface that provides robust high-speed Ethernet connection. The tether attached to the robot only provides data exchange with the workstation. The power is delivered by an embedded battery (a LiPo 3S 11.1V 5000mAh) that ensures an average autonomy of 1 hour. A schematic overview of the robot components is presented in Figure 2.4.
Table of contents :
Contents
Glossary
Introduction
1 State of the Art on Tether Management
1.1 The Underwater Environment
1.2 Underwater Robots
1.2.1 Autonomous Underwater Vehicles
1.2.2 Remotely Operated Vehicles
1.3 Tethers
1.3.1 Utility
1.3.2 Buoyancy
1.3.3 Cross-section
1.3.4 Models
1.4 Tethers and Deformable Objects Management
1.4.1 Underwater Applications
1.4.2 Terrestrial and Aerial Applications
1.4.3 Synthesis of Existing Cable Management Strategies
1.4.3.1 Passive and Active Cable Management Strategies
1.4.3.2 Classification according to Cable Perception and Modeling techniques
1.5 Our Scientific Focus: Vision Servoing of a Pair of Robots in a Chain of Mini-ROVs
2 System Modeling
2.1 Tether Model
2.1.1 Catenary Equation
2.1.2 Catenary Parameter
2.1.3 Catenary Parameter Constraints
2.2 Robots Model and Configuration
2.2.1 Thruster Configuration and Allocation Matrix
2.2.2 Kinematic Model of a mini-ROV
2.3 Pair of Robots Connected by a Tether
2.3.1 Catenary Model Applied for Tethered Robots
2.3.2 Tether Attachment Points and Robots Kinematics
2.4 Conclusions
3 Underwater Perception of the Tether
3.1 Camera Configurations and Assumptions
3.2 Camera Modeling
3.2.1 Image Formation
3.2.2 Camera Exposure and White Balance
3.2.3 The Pinhole Camera Model
3.2.4 Intrinsic Parameters
3.3 Tether Detection
3.4 Catenary Features Estimation
3.4.1 Catenary Equation in the Camera Frame
3.4.2 Catenary Projection on the Image Plane
3.4.3 Catenary Curve Fitting
3.4.3.1 Gauss-Newton Algorithm
3.4.3.2 Study of the Gauss-Newton Jacobian Singularities .
3.4.3.3 Particular Case of Remote Points
3.4.3.4 Gauss-Newton Improved Algorithm
3.4.3.5 Initial Guess of Catenary Shape for Gauss-Newton
3.5 Results
3.5.1 Focus on Feature Vector Estimation Error
3.5.2 Study Cases in a Simulated Environment
3.5.3 Discussion
3.6 Conclusions
4 Tether Shape Control
4.1 State of the Art on Vision-Based Control
4.1.1 Rigid Objects
4.1.2 Deformable Objects
4.2 General Control Scheme
4.3 Catenary-Based Interaction Matrices
4.4 Follower Robot Control using Tether Visual Feedback
4.4.1 Preliminary Results with Terrestrial Robots
4.4.1.1 Catenary-Based Visual Servoing for Terrestrial Robots
4.4.1.2 Comparison of Catenary-Based Control with Image-Based Visual Servoing
4.4.2 Underwater Tether Shape Regulation while Leader Robot is Motionless
4.4.2.1 Visual Servoing Control with 3 × 4 Interaction Matrix .
4.4.2.2 Sum of Controllers
4.4.2.3 Hierarchical Task Control
4.4.2.4 Comparing Follower Robot Trajectories
4.4.3 Underwater Tether Shape Regulation while Leader Robot Moves
4.4.3.1 Neglecting the Leader Velocity on the Follower Robot Command
4.4.3.2 Including the Leader Velocity on the Follower Robot Command
4.5 Follower Robot Control using Tether Visual Feedback from Both Camera- Underwater Case
4.6 Discussion
4.7 Conclusions
5 Conclusions
5.1 Summary
5.2 Perspectives
A Position and Orientation Representation
B Expression of the catenary parameter C
C Kinematic Equations with Twist Matrix: General Case
D Catenary Derivatives
E Preliminary Results with Terrestrial Robots
F Interaction Matrix Test Protocol
G Scientific publications, Workshop Participations and Scientific Popularization Activities
H R´esum´e en fran¸cais
Bibliography
