RoboSub is an international and interdisciplinary competition involving numerous student groups and universities from around the world. In the competition, each team must develop and present an Autonomous Underwater Vehicle (AUV) that has capabilities of completing several tasks that have real-life applications. These applications include underwater exploration, seafloor mapping, sonar localization, and underwater object manipulation. The week-long RoboSub competition takes place annually towards the last week of July or the first week of August in San Diego, California at the NIWC Pacific Transdec facility. However, because of the COVID-19 epidemic, the competition will be held remotely in 2020.
RoboSub is held by RoboNation, which is an organization that empowers students to generate innovative ideas in the field of marine sciences. It is a nonprofit that partnered with companies such as SolidWorks, Northrop Grumman, and Boeing, among others to promote STEM education through programs and 9 student competitions that engage more than 250,000 students per year.
This year, RoboSub added and modified several tasks to fit the theme of the roaring 20s and prohibition (Skidoo). In the competition, each team is allowed to join the bootleggers or the G-men (government men). This choice of persona dictates how the tasks should be approached for maximum points. The tasks this year include:
The gate task combines 3 smaller subtasks into this larger task. At the beginning of the robot’s run, the team is given the choice to decide its robot’s orientation relative to the gate (known path) or they can flip a coin (random path). The flip of a coin will decide whether your robot’s orientation will start 90 degrees or 180 degrees relative to the gate. Since the random path is the higher difficulty option, more points are rewarded for that task.
As the robot starts making its way to the gate, it will be greeted with two images on each side of the gate (the gate is split 50/50). There will be an image of a bootlegger and one of a G-man. The remainder of the tasks is based on the robot’s decision as to which side it will go through. In other words, there will be two courses (G-man and bootlegger) that have the same tasks but different images associated with the choice made.
Lastly, as the robot passes through the gate, the robot can gain style points for every 90-degree change in direction to a maximum of 4. For example, the robot can do a barrel roll (rotates 360 degrees on the roll axis) to gain the maximum amount of points.
The plan for Arctos is to start in a random orientation, choose the bootlegger side (we are dangerous), and complete our signature barrel roll to maximize our points.
The buoys are a simple task to understand but they require a well thought out plan before implementation. In the buoys tasks, there will be a 2-sided flat board with 2 feet by 4 feet image of a Tommy gun (bootlegger side) or a badge (G-man side). The objective of the task is to touch the image associated with the initial course decision made during the gate task. Arctos will choose the Tommy gun so it can ‘re-load its arsenal’ before completing the rest of the course.
In this task, there is a bin with 2 images, which are dependent on the side chosen. These images are placed on the bottom surface of the bin and can be seen using downwards facing vision. The objective of this task is to drop markers into the bin to gain points. Each robot is permitted to carry a maximum of 2 marker droppers but is permitted to pick up nearby ‘bottles’ and drop them into the bins. The more droppers that the bins receive, the higher the points the robot will achieve. However, to make this task harder, 1/3 of the bin is covered at all times by a lid that is placed randomly on the bin. Arctos has an on-board dropper system coupled with a new bottom camera that will be used to effectively drop the markers.
This is one of the most exciting tasks in the course since it involves shooting projectiles through a relatively small target. Similar to buoys, there will be an image of either a bootlegger or G-man on a 2 feet by 4 feet panel with two cut-outs. A robot has the choice of shooting onboard torpedo missiles through these targets, which is based on the initial course decision made after passing through the gate. Arctos will participate in the ‘shootout’ since we improved our torpedoes system immensely for this competition.
Lastly, there is the octagon task, which is split into several sub-tasks. As the robot travels through the pool, there will be a floating octagon which can be seen from the surface. In the middle of this octagon, there are 2 tables with either an ax (G-man) or a dollar sign (bootlegger). One option is to surface the robot through the octagon. Another option is to pick up bottles and place them on the appropriate table. Thirdly, a robot can pick up a bottle, surface with it, and then place it on the table to maximize points. Arctos will attempt to pick up bottles and place them on the bootlegger table. This will be the first time an ARVP robot has tried a manipulation task and is only possible because we have added a robotic arm to Arctos.
Pool logistics can be a nightmare when the team wants to test the robot. It is not an option to test the robot on land since the robot is designed for aquatic environments. Even after organizing pool tests, resources and supplies must be moved from the University of Alberta to the external venue to ensure we have the items needed to troubleshoot and test. However, it is always worth the pain to witness the team’s hard work in action!
Capturing high-quality video for computer vision is a difficult task out of the water and is much more difficult underwater. Pool water tends to have a green to bluish tint, which can skew the actual colors of obstacles or task components. When the water is murky with organic materials, this further prevents the robot’s ability to segment images for its computer vision as these materials decrease the shades, tones, and color differences between various parts of a competition component.
With underwater vehicles, GPS is not an option to determine the location of the robot. Therefore, an expensive and diverse sensor suite including depth sensors, cameras, a Doppler velocity logger, sonar, and an inertial measurement unit is required to map out the pool environment while ensuring a secure orientation. This sensor suite is easily the most expensive part of the robot.
Sealing is the largest issue that the mechanical team combats. Most electronics used in the world are not capable of functioning for prolonged periods underwater since they will be damaged. As a result, the mechanical team adheres to universal sealing standards when designing the sealing mechanism.
A second design constraint that the mechanical and software teams consider is the robot's buoyancy. Mechanically, this slightly constrains the design as the team must design for the robot to be 1 to 2 pounds over buoyant. If it is higher than this threshold, the thrusters will waste too much energy on keeping the robot underwater. For software, they must create a controller that prevents the robot from surfacing, which combined with the other thrusters becomes challenging. However, both the mechanical and software teams have incredible members working on finding these solutions!
Not all wires are made the same, so the team must get creative when connecting components to power supplies. Most of the time, electrical tape and epoxy does the trick but must be applied carefully to prevent component damage. Moreover, the electrical team sometimes splices wires or modifies existing connections to ensure a good fit.