1. Introduction
The purpose of this project is to research cooperative robotics and more specifically swarm robotics. Following this, a swarm robotics system will be designed, developed and implemented using swarm intelligence principles to demonstrate the concept of cooperative robotics.
1.1. Inspiration
Swarm robotics research has taken inspiration from areas such as amorphous computing, self-assembly of materials and the behaviours of biological swarms in nature. Many animals, insects and organisms which are individually simplistic, inefficient and incapable can exhibit complex behaviour when in large groups, also known as “swarm behaviour”. For example, Research has shown resistance to antibacterial agents in a large group of bacteria is 500 times greater than that of individual bacteria of the same strain [1]. While termite colonies which have an individual height of millimetres can build large, complex structures meters tall [2] [3]. The foraging behaviour of ant systems is a common research area [4] [5] [6], often focusing on the cooperative group behaviour, and how individuals produce pheromone trails in order to communicate the shortest path to travel to other ants. The more ants that follow the same trail, the greater the pheromone strength and the greater the likelihood of other ants to follow it [7]. The Ant Colony Optimisation (ACO) algorithm is a technique that can computationally simulate this group behaviour and find solutions to optimisation problems, which is useful for experimentation and application in the area of swarm robotics. For example, autonomous robots exploring an unknown area can use ACO to make a navigation decision based on pheromones, perceiving less pheromones to be an area that is less likely to be explored [4]. The diagram in Figure 1 shows how ants optimise foraging behaviour to find the shortest path to food.
Figure 1.1.1. Ant foraging behaviour [8].
There are other commonly used swarm optimisation techniques such as Particle Swarm Optimization inspired by flocks of birds and schools of fish [9], and the Firefly Algorithm which is inspired by the attractive light flashing behaviour of firefly swarms [10].
1.2. Fields of Application
The fields of application for swarm robotics include:
• Area covering tasks: Environmental monitoring, such as crops in farming [11], oil spill tracking [12], marine environmental monitoring [13], and monitoring for the immediate detection of dangerous substances [14].
• Dangerous tasks: The detection and disposal of mines [10] or searching for survivors in the event of an earthquake or tsunami [15].
• Scaling tasks: Tasks that require up or down-scaling over time. For example, a robotic swarm to contain an oil spill will require upscaling as the spill size increases [14].
• Redundancy tasks: Jobs that place an importance in a robust system that can operate as normal, even in the event of damage or destruction to individuals in the system. For example, a dynamic communication system in a battlefield [14].
1.3. Characteristics of Swarm Robotics
In order to decide upon objectives for this project, a clear set of criteria must first be defined for characterising a swarm robotic system :
1. Autonomous: The robots must be able to physically interact with the environment and work completely autonomously.
2. Individual robot simplicity: The individual robots are programmed with simple instructions but can collectively carry out complex tasks.
3. Homogeneous: Every robot is equal in its instructions and capabilities.
4. Large numbers of robots: Groups of 10 or 20 are generally considered to be a swarm, however, groups are scalable meaning numbers can be scaled up or down in a range, from tens up to millions and beyond. This reduces cost of manufacture but poses challenges with experimentation and maintenance.
5. System redundancy: If one robot fails or malfunctions, another will compensate for it, making individual robots dispensable.
6. Every robot must have sensing and communication capabilities: Stigmergy is known as communication and coordination through the environment, which is one of the most important concepts of swarm robotics [3].
7. Decentralized control: Every robot has their own communication capabilities, allowing for cross communication between robots, adding robustness and redundancy to the system. Individual robots are dispensable, meaning if one robot fails the system is not affected. In centralised control, however, a central controller controls all robots with the disadvantage that if the controller breaks, the whole system will be unable to operate.