Coral reefs support half a billion human lives, but are increasingly damaged and dying; we want to build robots that repair them!
Coral reefs support over 500 million people across the globe. But reefs are being damaged on a global scale by storms, destructive fishing, ship groundings and careless tourists. It can take many years to decades for them to heal. We can speed up the process by using scuba divers that re-attach healthy pieces of coral back onto the reef like these pictures show from the "Fragments of Hope Coral Nursery in Belize" (photos by Lisa Carne) using skilled paid fisherman and tour guides that are an integral part to the reef restoration process. But this method cannot be used to repair reefs in deeper waters because of depth limits to humans diving.
The answer: ‘coralbots’.
Coral-bots are a team of robots that intelligently navigate across a damaged coral reef, transplanting pieces of healthy corals along the way. The big job of developing and testing the robots at sea has already been done. All that remains is to embed the robots with computer vision to “see” healthy bits of coral, and configure appropriate manipulator arms for each robot to pick up and put down the pieces in the right spots. Kickstarter funds will let us purchase and assemble this kit, and allow us to conduct our first live demonstration of the robot team on a coral reef in a public aquarium. This will provide a conservation solution that paves the way for coral reef restoration across the globe.
Here are some examples of autonomous underwater vehicles (i.e underwater robots) that have been developed by the Ocean Systems Laboratory (part of the coralbots project team) for purposes such as monitoring of underwater pipelines.
The image immediately above, 'Nessie 4' is the platform we intend to adapt for the first missions to repair coral reefs in Belize, and for demonstrations prior to that in public aquaria. This platform is easily equipped with onboard camera, computing, and flexible arms and grippers. A prototype first coralbot design can be seen in the main image that accompanies our kickstarter presence.
What makes our vision work is our idea of using swarm intelligence methods to control robot behaviour. Swarm intelligence explains how simple behaviours in a group of creatures can lead to complex and functional structures – this is how bees build hives, and termites build complex mounds, and beavers build dams. See here for a basic tutorial on swarm intelligence:
Borrowing ideas from swarm intelligence, we can build on state of the art expertise in each of coral biology (Lea-Anne Henry), autonomous underwater robots technology (David Lane runs the Ocean Systems Laboratory at Heriot-Watt University, Edinburgh, and Dick Blidberg runs the Autonomous Undersea Systems Institute in New Hampshire), swarm intelligence (David Corne), and video/image processing (Neil Robertson – giving coralbots the ability to distinguish coral and understand the other key things in its environment).
With $107,000, in six months – with your help – we can demonstrate this technology in public with two specially adapted robots.
Every extra $$ will extend the number of public demonstrations, bringing us closer to our ultimate goal of increasing the team size to eight robots and using them on coral reefs around the globe.
As you may imagine, it certainly doesn’t stop there. The concept of swarm intelligence combined with what we know is possible in underwater robotics and computer vision, all make for a powerful combination that could contribute immensely to a swathe of issues in the marine environment. Help us in that longer term programme by first backing coral-bots!
Risks and challenges Learn about accountability on Kickstarter
Getting robots to do things autonomously is hard, and getting them to do it underwater is even harder. But the coralbots team is a highly motivated group of individuals whose collective track record of successful underwater robotics projects is very hard to beat. We already know the typical obstacles and how to surmount them. What we want to do in this project (and in the wider programme) is to change the way robots are used and perceived in marine science, and in environmental science as a whole: at present, robots are invariably used just to do monitoring and/or surveillance. But, modern artificial intelligence coupled with underwater experience can enable them to do so much more, and make a big difference to many global challenges. In this project that means we need to get the machine vision, arm/gripper control and associated navigation algorithms right, as well as tackle and solve various issues (that the team has solved in other projects) of communication between robots and between an onshore, or on-surface team and the robots. We know we can do all of this with current tried and tested technology - the major risk is that the performance of the system we will be able to build with the kickstarter resources will be quite slow -- but one slow robot that takes several days to restore a small reef system is better than nothing at all (when that reef is too deep for scuba divers) and better than risking human lives. And, as we further improve the hardware and software technologies in parallel, we can speed this up by adding more robots (which is one of the benefits of the 'swarm intelligence' approach).
Thanks to Anton to be the first to ask this question.
Definitely, everything we generate with kickstarter funds and a whole lot besides will be open source!
There is an issue to be aware of though: to be able to get the real work of reef restoration started as soon as possible, we expect we will need to exploit some h/w and s/w that has already been designed, including some off the shelf equipment (some we have already and some we don't). It could be that not all of that is open source yet. Basically, the fastest way to get these problems addressed means (for example) that we stand on the shoulders of some visual servoing ware that comes packaged with a particular robot gripper (that sort of thing). But I repeat, of course all code we generate, and any new h/w, will be open. This certainly includes all of the swarm intelligence aspects and machine vision for coral (and similar) recognition, and so on. And we will advise on open alternatives for anything we happen to make use of that isn't, until we have the resource to develop/adopt that ourselves too.
Thanks to Herbert to be first to ask this question and suggest it as an FAQ.
Various approaches to attachment have been tried on reefs by SCUBA divers, and research has monitored the success of the re-generation of different attachment methods. A promising approach seems to be 'underwater epoxy' - basically, glue that works in that environment. Divers also sometimes use wires/ties and nets. We have robotics approaches that will be able to operate each of these techniques. In each case (and with the current state of technology) our approaches will likely operate slowly at first. But of course, they can keep going for several hours, and are multiplied in speed by having a robot team. The epoxy approach is particularly interesting in the light of swarm robotics -- likely we'll have specially prepared plugs of epoxy lowered to the seabed nearby on a pallet; half of the bot team would then pick up and find good places for these - the other half would handle coral and find epoxy plus to put them on. There is also the simpler case where divers have to transplant nursery-grown corals - these are typically placed in concrete 'pots' that are then placed in a suitable area on the reef. The bots will pick up the pot, and place it, with little or no need for further fixing.
This was DavidC