A summary of all the prototypes and the project.
A summary of all the prototypes and the project.
In the open ocean
An autonomously generated map section
Autonomous sailing even in rough weather
What it does
A giant issue with coral conservation - reef mapping is done by divers moving and photographing a PVC quadrat for every unit area of the reef. MindoroBot is a swarm-robot which can sail and photograph+map reefs autonomously at a low cost with a laser quadrat.
Coral reef ecosystems are some of the diverse and valuable ecosystems on earth. They support more species per unit area than any other marine environment, including about 4,000 species of fish, 800 species of hard corals and hundreds of other species as well as provide goods and services worth $375 billion each year. However, due to climate change, only about 46% of the world’s coral were considered healthy in 2008 and this percentage has dropped further recently. Coral protection is key for the sustenance of human life on earth and we started Mindorobots to make it affordable and viable for local communities to map and conserve their reefs.
How it works
A combination of lightweight & strong acrylic, flexible bamboo & vulcanized rubber was used to attain a lightweight (3 kg) & durable frame. The materials combined with smart design allowed it cruise steadily even in rough seas with a speed of 9.7 knots. On average the drone can cover 360 m2/hr area. It can work up to a maximum depth of 5 meters & with very high image quality. Programming Language: Python & C++ were used for coding. Software: Pix4D, Google SketchUp. The ocean drone is relatively cheap & easy to build. The simple yet efficient design allows for easy modification. Cost: USD2375, a small investment compared to current submersible vehicles & equipment. The robot uploads the photographs into a cloud based photogrammetry suite where computer vision algorithms transform the Multispectral images to 3D maps. Successive maps are then compared across multiple pixels to gauge the extent of bleaching of the reef at different points and different dates.
A total of five prototypes were built - each incrementally better than the previous one till the desired product was made. The first prototype called 'Vertical' was a square in shape with the cameras+propulsion underwater and the radio communication above in two connected tubes. We found this design was unstable and required large power. We then made prototype 2-"Super U" which was in a "U" shape and floated on water. We got great results with this - Stable Design, Good Propulsion, Light Weight but with large turning radius. Then we built prototype 4-"Super H" from PVC with a waterproof section for the electronics & camera. It also had a gimbal for stability of the video. We found it to be: Durable Materials, Stable, Strong. However, much needed to be improved on the electronics to autonomously guide the robot and we still had waterproofing issues. Hence, the Pixhawk flight controller from a drone was hacked and modified to work on our boat system. Finally "“Bamboo Fans" - our final prototype was made from local bamboo so that local communities are more open to using and maintaining it and had propellers to sail so as to avoid waterproofing problems. The bot was tuned so that it can autonomously navigate and photograph reefs to generate the required maps. See the video for more!
How it is different
This robot updates the physical quadrat which is used today to a projected laser quadrat - eliminating the need to dive to the bottom of the sea without functional loss. The robot is capable of mapping larger areas of reef at lower time requirement than a human diver. To our knowledge, this is the first system to adapt air-based autonomous navigation capabilities to ocean-based systems while updating quadrat. Maps generated with manual photography take longer to generate by the order of 10^2. For comparison, Satellite: time = 1 second / cost = $100 million / 1 px =10 m Airplane-UAV: time = 1 minute / cost = $10 million/ 1 px =1 m Mindorobot: time = 1 hour / cost = $1000/ 1 px = 0.001 m Diver: time=1 day / cost = $100*no.of days/1 px=0.000001 m The ocean drone does seem to provide an optimum advantageous balance between cost, time and image quality with regard to mapping reefs. This is the only design for autonomous reef mapping robot in the world!
The resulting map generated by the robot is taken from the water surface, hence reducing the effective resolution with which the seabed is observed. On analysis we can say that this method generates a map of higher resolution than air/UAV based analysis by the order of 10^3. Hence, the current area of research would be the integration of underwater extension of the same system that can operate as an untethered underwater drone. Further improvements are to be carried out to the image capture systems to improve their resolution and make the autonomously generated map close to a manually developed one.
Gallant Ho Experiential Learning Fund, HKU Also broadcast in major television channels and newspapers in the Philippines (GMA News and Manila Bulletin)