Paper: Berlinger, F., Gauci, M., & Nagpal, R. (2021). Implicit coordination for 3D underwater collective behaviors in a fish-inspired robot swarm. Science Robotics, 6(eabd8668), 1–14. Retrieved from https://robotics.sciencemag.org/content/6/50/eabf4315
Schools of fish perform a mesmerizing underwater ballet, gracefully and harmoniously swimming alongside hundreds of other individuals. Schooling can be a tactic to avoid predators or hunt more effectively, but we know very little about how fish coordinate their behaviors to form a seamlessly swimming mass. If we could harness the collaborative nature of schooling fishes, scientists believe they could develop robots that could provide a better understanding of the behavioral and biological underpinnings of schooling and eventually be used to monitor at-risk environments or conduct search-and-rescue missions.
Engineers have already developed terrestrial and aerial robot swarms capable of coordinating actions through communication with one another or a centralized communication system. But the radio waves and satellite-based GPS signals that enable these robot collectives to cooperate are useless underwater, where signals cannot move as easily as they do through air. Underwater robots typically rely on communicating with a server via a direct cable or tethered connection (as seen in Remotely Operated Vehicles or Submersibles) or they must come to the surface of the water to broadcast a signal to a satellite or nearby communication hub. Some recent advances in marine robotics have focused on establishing networks of underwater base stations to enable underwater communication. Others have drawn on biologically-inspired electroreception (akin to the sensory system used by sharks to navigate and find food) to coordinate movements between two robots. But these fail to replicate the seamless coordination and agility of robot swarms above water.
Underwater swarming bluebots
Inspired by schooling behaviors in fishes, a team of researchers from Harvard University led by Florian Berlinger, developed a group of robots capable of swarming underwater. Their blueswarm coordinates actions by observing actions of other robots in the swarm (called implicit communication) in a manner that does not require communication with an external command center and is expected to be similar to the methods used by schooling fishes.
The 130 mm long robots (about the same length as a typical smart phone) even look like fish, with two large eye-like cameras for visualizing their surroundings and four independently moving fins (two pectoral fins, one caudal fin, and one dorsal fin) that help provide the bluebots with the agility required to form tight-knit underwater formations. Bluebots are also equipped with three blue LED lights, mimicking the spots and stripes (“schooling marks”) fishes use to detect their neighbors. Like the fish they are modeled after, bluebots rely on vision for schooling; the wide-angled cameras detect the blue LED “schooling marks” on nearby bluebots, allowing them to detect the position of and distance to neighboring robots up to 5 meters away (under ideal conditions).
Synchronized Swimming
Berlinger’s team was able to program bluebots to coordinate movements by developing an algorithm based on behaviors in fireflies, which are known to synchronously flash by matching the frequency of their own blinking to that of their neighbors. The researchers started by setting individual bluebots to flash their blue LED lights at random intervals. But when the bluebots detected light flashes from neighboring bots, they slowly began to synchronize their flashing. In less than 2 minutes, the entire swarm had synced and coordinated their flashing frequencies.
The team of researchers even attempted to mimic more complicated behaviors that are not fully understood in schooling fishes yet. They started by developing an algorithm that would allow the blueswarm to alternate between schooling and dispersing – much like fishes do in the wild, dispersing to forage individually but somehow retaining connection to one another so they can rapidly reform a school when danger is nearby. Building on this behavior, the researchers used an evolutionary-based algorithm that allowed bluebots to mimic milling – when fishes form a tightly-packed school and swim in a circle to avoid predators. Although not as complex as the milling schools observed in the ocean (which can be made up of 100s -1000’s of individuals cooperatively swimming at once), the researchers were able to program up to 7 bluebots to swim in a milling-like circular formation.
Search and Rescue
In a final test, the researchers developed a method to mimic cooperative hunting behaviors in schooling fishes by sending a swarm of bluebots on a search-and-rescue mission. The bluebots were programmed to locate a red LED light. Once the search began, individual bluebots dispersed to hunt for the light independently. When a bluebot encountered the red LED light, it was programmed to hold its position and flash an alert signal. As nearby bluebots noticed the alert signal, they were programmed to swim to the alerting bluebot and join the alert display. In this manner, more and more bluebots gathered near the red LED light and formed a stronger alert beacon to attract the reaming bluebots in the swarm.
Bluebots have only been tested under near-ideal conditions in a laboratory setting, but the researchers hope that improvements to camera systems will allow bluebots to detect other members of the swarm in murkier waters and natural environments. Soon, researchers could be using blueswarms to study how fish coordinate movements or to build more cohesive networks of underwater robots.
I received my Master’s degree from the University of Rhode Island where I studied the sensory biology of deep-sea fishes. I am fascinated by the amazing animals living in our oceans and love exploring their habitats in any way I can, whether it is by SCUBA diving in coral reefs or using a Remotely Operated Vehicle to see the deepest parts of our oceans.
