Teaching Swarms to split up tasks efficiently
Swarm of security drones with surveillance camera flying in the sky. 3D rendering image

Teaching Swarms to split up tasks efficiently

The Technology Innovation Institute in the United Arab Emirates is developing a new paradigm for teaching robot swarms how to efficiently devote themselves to activities. The concepts may pique interest in developing more effective methods for commanding a large number of simple robots to perform more complex jobs autonomously, with little or no external communication.

Numerous previous studies have been conducted on coordinating swarms of drones to execute amazing tasks, such as a coordinated light show. However, these techniques frequently relied on centralised coordination, expensive equipment installed on each drone, or a combination of the two. The TII researchers are investigating several methods for scaling a swarm by combining a large number of simple robots with a few intelligent ones.
The basic premise is that this approach would enable swarms to operate in confined areas, such as deep underground, behind enemy lines, or in outer space. Eliseo Ferrante, a senior researcher at the TII, predicts additional futuristic situations, including the control of augmented natural cells used as robotic components.

This novel technique establishes a method for guiding the swarm through the use of a few specially equipped implicit leaders. “By lowering the required sensing and communication thresholds, these types of applications may become feasible even at the nanoscale,” he explained. This could pave the way for the development of fleets of nano-drones capable of eradicating cancer, repairing tissue, and curing disease in areas with little external connectivity.


Teaching Swarms to split up tasks efficiently
aerial surveying with drones


Flocks and bots

Eliseo’s interest in the mathematical features of swarms stems from his fascination with how birds and bees flock. Similar fascinations over the years have prompted scholars to build a hybrid theoretical and experimental discipline devoted to understanding collective behaviour. This science explains how individuals paying attention to a few small qualities of their neighbours can result in coordinated behaviour, such as a flock of birds’ perfect triangle.

Other researchers have been investigating ways to operate multiple robot swarms in such a way that complexity is efficiently propagated to all the robots. However, these systems are frequently inefficient in terms of scalability, as each robot requires sophisticated sensors, external networking infrastructures, and computers. “We’re presenting a method for combining these two approaches to robotics,” Eliseo explained.

One line of research has concentrated on constraining multi-robot systems one constraint at a time. This may entail disabling the network, lowering sensing capabilities, or restricting computing.

Collective decision making

This burgeoning new field of collaborative decision-making blends swarm behaviour and collective control. The area began by investigating various ad hoc approaches to communal decision-making and subsequently progressed toward a more generic framework.

At one extreme of this paradigm, robots are taught to make decisions based on what is optimal for some aspect of the environment, such as the amount of food or light available. The robots are unaware that they are contributing to the solution of a common problem. The researchers were attempting to determine ways to motivate the robots to solve little tasks that added up to a larger task. A distinct issue is to reach a consensus on a course of action that minimises expenses such as energy or travel distance.

The researchers included a few knowledgeable individuals with a broader understanding of the environment in their study to help steer collective decision-making in a certain direction. Earlier studies examined how to coax robots to congregate in one location. This occurs when ants are drawn to a large pile of food by the trail of pheromones left behind by other ants.
However, what if you want the ants to form two groups in order to transport a whole slice of watermelon and the smaller cherry adjacent to it? The goal is for the ants to automatically determine how to proportionately divide themselves between the much bigger group required for the heavier watermelon and the much smaller group required for the cherry. Nanomedical robots may entice additional robots to focus on a larger tumour, with a few branching off to work on a smaller one.

Eliseo’s intuition told him that they needed to develop a feedback mechanism to ensure the proper balance was maintained.


No votes yet.
Please wait...