Mastering from A to Z
Most ROVs available in the market are closed systems in terms of mechanical structure and firmware/software. Upgrading them to perform new functionalities is thus difficult. To experimentally validate our homography-based control algorithms, we have developed an underwater robotic platform (named I3S-UV) at the I3S-OSCAR team.
A BlueROV1 kit without control parts was first purchased and then modified by adding a longer watertight tube to accommodate the main electronic components. The original tube of the BlueROV1 was implanted right below for containing separately a bigger battery. In addition, a shorter tube with spherical dome end cap was dedicated for a camera (either forward-looking or downward-looking). New structural components were designed and printed by using a 3D printer with plastic filament.
The I3S-UV software system has been developed for implementing control architecture including high- and low-level controls. At the high level, our homography-based control algorithms are implemented in the form of a ROS package in C++ and run in the companion computer using ROS as a middleware. The low-level control mainly in charge of motor control allocation is run in a Pixhawk. The code source of the low level has been adapted from the open-source autopilot PX4 which is based on Mavlink communication. The MAVROS package in ROS providing communication between Mavlink and ROS is thus used to link the two control levels. A low-cost and compact single-board computer (SBC) is used as companion computer for running high-level control, image acquisition, homography estimation using HomographyLab, and communication over Ethernet with the ground station computer. It was a Hardkernel XU4 which has been recently upgraded to an Nvidia Jetson Nano in May 2019. The high level control is set to run at 100Hz even though the rate of homography estimation (in station keeping mode) is lower. The low level control rate is 100Hz.
Additional packages available in ROS such as online graphical analysis of data, image view, etc. can be run on the ground station computer for directly debugging while performing the experiments. Sensor data, images and values of control inputs can be recorded in Rosbag files which are then useful for post-experiment data analysis. In this software system, ROS with its decentralized architecture allows for all processes (nodes) to run independently either in the companion computer or the ground station computer. The homography estimation node thus can be run in different computers for comparing the performance (see Table I).