Swarm interaction and group robotics in solving various tasks

A swarm is a self-coordinating system of multiple robots that act as a single organism, performing a common task. In fact, group robotics is an attempt to adopt the strengths of a true bee swarm, in which each insect plays its part to benefit the whole community. Remarkably, the term initially took us to an even lower level of life: simple human cells that build one giant organism. In 1988 the Japanese researcher Toshio Fukuda used the term "cellular robots" or "cellular robots" to refer to a dynamically reconfigurable robotic system. Then the term was "picked up" and modified by the American Gerardo Beni, who was looking for a "buzz word" and eventually adopted the idea of his colleague Alex Meistel:

"The discussion was quite lively, and I remember Alex Meistel saying that 'cellular robot' is an exciting concept, but the name doesn't appeal; you need a fancy [lit. 'buzzing'.-note] word to describe 'this kind of 'swarm."

Gerardo Beni, University of California

Goals and Objectives​​

Since the advent of the concept of swarm intelligence, researchers have articulated the primary attributes of a swarm. Typically, it consists of robots of the same type that can interact with their "fellows" and the environment using simple sensors but are not rigidly controlled "from above" and do not affect the global level system.

The main advantages of this approach are:

  • scalability
  • fault tolerance;
  • flexibility in problem statement and methods of achieving it;
  • the use of inexpensive, often miniature robots;
  • the possibility of parallel execution of the same type of operations;
  • the unpredictability of individual robot behavior.

These benefits have prompted the creation of complex distributed systems designed to solve problems that cannot be solved by a single individual robot or where swarming is more effective in achieving the goal. Swarming thus "closes" three of the four categories of robotics tasks:

  1. Fundamentally single-agent tasks.
  2. Tasks require multiple agents.
  3. Principally multi-agent.
  4. Winning by using several agents.

The distributed approach simplifies the robot design considerably to make it cheaper to turn it into a "universal soldier." Having limited sensory, communication, and computational resources, such an agent alone is not worth much, but together with its "fellows," can solve very complex problems. The main goal of group robotics is to create and improve such self-managed distributed systems.

Where and how it is used

The capabilities of distributed systems also determine their application areas. Group robotics is seen as an essential tool in the following cases:

  • in rescue, military, and other operations in which direct participation is dangerous for humans (demining, search actions, cleaning up nuclear or toxic waste, military operations);
  • when the scale of the task is initially unknown, and it is necessary to scale the system by adding or removing robots (e.g., when fighting an oil spill);
  • When working in environments where telecommunications and other infrastructure to control robots is not available, such as underwater, in space, or the open ocean;
  • when conditions change rapidly and swarm flexibility is needed (patrolling terrain, rescue during floods or after a hurricane, autonomous cattle grazing in the mountains);
  • when the factor of the cheapness of one robot and high speed of their joint work is used (mining industry, weeding in agriculture, 3D-printing of large objects).

Military departments of different countries are actively developing swarming systems.

In China, such a solution is used for autonomous management of a fleet of small vessels that conduct joint patrols of the territory. Unmanned boats can change their behavior strategy when foreign objects appear on their course; for example, they successfully circle an island or a ship.

Similar solutions are available from the U.S. Navy, which conducted boat swarm demonstration tests on the James River in Virginia in 2014. During the trial, five boats accompanied the flagship, providing cover. The Control Architecture controls the vessels' actions for Robotic Agent Command and Sensing (CARACaS) system, built on an "extraterrestrial" design. CARACaS is based on software and hardware developed initially to control NASA Mars rovers.

Good use of swarm systems is to control groups of drones for search, military operations, mapping, tracking moving objects, managing farmland, and a host of other tasks. Swarm interaction can be helpful at the micro- and nanoscale, for example, in diagnosing compact equipment or even in the human body.

The main challenges of creating swarming systems

Despite the apparent simplicity of robots, unsupervised group interaction between them is pretty tricky. It is laid down at the individual level through algorithms controlling the robot's and its fellows' activity. It is crucial to determine how a personal agent reacts to local changes (approaching another robot, obstacles, environmental changes) and affects global transformations.

At the same time, it is pretty tricky to calculate how a single object should behave for the whole system's optimal actions. The concept of scaling introduces additional difficulties because it is expensive and challenging to conduct many experiments with a natural swarm. Researchers use mathematical modeling to solve this problem and "artificial intelligence" and so-called "artificial evolution" to train the algorithm.

A separate group of problems is hardware. For example, the autonomous operation of robots can be solved by using more "advanced" batteries and by creating automated recharging stations or even a system in which one agent charges a "comrade-in-arms."

Another challenge is communication within the swarm, exchanging data with the external telecommunication network and uploading information to the cloud if the concept provides such capabilities. "Cloud robotics" is an even higher level of challenges that will allow more efficient use of the robot swarm, increase its responsiveness to events, and may generate new and somewhat unexpected niches for swarm technology applications. Yes, and the other problems, as they are solved, are highly likely to become a source of progress for swarm robotics and related industries, which will radically change IT and all of our lives.