Mimics the movement of Real Insects



Download 68.4 Kb.
Date03.05.2017
Size68.4 Kb.
#17130
Colony of Robots

Mimics the movement of Real Insects
Deepshikha Varshney*, Sachin Agrawal#, Sreesh Gaur% and Arun Agarwal^

Computer Sc. & Engg. Deptt.

J. P. Institute of Engineering & Technology, Meerut

* deepshikhavarshney1@gmail.com

# er.sachin.agl@gmail.com

% gaursreesh@gmail.com

^ arun.261986@gmail.com


Abstract- This paper presents an introduction to the world of swarm robots and adumbrates its applications and how it is applicable in a real life scenario, in what scenarios the robotic swarm can be used, what the benefits and draw-backs are.

A lot of interest has been put in robotic swarm applications in the past. Despite this there is still a lot to be developed in the field. Swarm robotics is currently one of the most important application areas for swarm intelligence. Swarms provide the possibility of enhanced task performance, high reliability (fault tolerance), low unit complexity and decreased cost over traditional robotic systems. They can accomplish some tasks that would be impossible for a single robot to achieve.

Swarm robots are more than just networks of independent agents, they are potentially reconfigurable networks of communicating agents capable of coordinated sensing and interaction with the environment. Robots are going to be an important part of the future. In the near future, it may be possible to produce and deploy large numbers of inexpensive, disposable, meso-scale robots. Although limited in individual capability, such robots deployed in large numbers can represent a strong cumulative force similar to a colony of ants or swarm of bees. Once robots are useful, groups of robots are the next step, and will have tremendous potential to benefit mankind. Software designed to run on large groups of robots is the key needed to unlock this potential.
Keywords— Swarm robots, cumulative force, networks, adumbrates,meso-scale.
I. Introduction

Swarm-bots are a collection of mobile robots able to self-assemble and to self-organize in order to solve problems that cannot be solved by a single robot. These robots combine the power of swarm intelligence with the flexibility of self-reconfiguration as aggregate swarm-bots can dynamically change their structure to match environmental variations. Swarm robotics is a very recent topic of research which involves creating artificially intelligent systems that can communicate and share information with each other.

The research of swarm robotics is to study the design of robots, their physical body and their controlling behaviors. It is inspired but not limited by the emergent behavior observed in social insects, called swarm intelligence.

What is "Swarm"?

As robots become more and more useful, multiple robots working together on a single task will become common place. Many of the most useful applications of robots are particularly well suited to this “swarm” approach. Groups of robots can perform these tasks more efficiently, and can perform them in fundamentally difficult to program and co-ordinate.

Swarm robots are more than just networks of independent agents they are potentially reconfigurable networks of communicating agents capable of coordinated sensing and interaction with the environment.
II. Evolution of Swarms

Inspired by the collective behaviour observed in natural insects, swarm robotics is a new approach of design to demonstrate swarm intelligence in a large group number of robots. Some researchers in robotics are trying to mimicking human intelligence, while other robotics researchers are taking inspiration from nature. It is not only the brains of animals and insects that they are trying to mimic but also their shapes and behaviour in unpredictable situations and environments. Swarm robotics draws inspiration from self-organizing behaviour observed in social insects1 like ants, bees and of other animals, called swarm intelligence (SI) [9].


One example of SI is bird flocking. By evading collisions, staying close to each other and aligning to local neighbours, birds in a flock avoid predators. These simple rules give rise to a very complex behaviour [1]. Fish schooling follows the same principle behaviour pattern. In social insect colonies, individuals may be very simple but when working together they can do remarkable things. For example, an object much too heavy for a single ant might not be for a group of ants working to reach a common goal. Ants are also very good at finding the fastest way to a food source from the nest by using pheromone trails [8].
This behaviour can be applied to robotics. The key concept is to get multiple simple robots following simple rules to carry out complex tasks. This is done by building a network of communication between the individuals in a robotic swarm, to enable them to share information amongst each other. Each individual coordinates by using de-centralized control and self-organization.
However, developing swarm software from the “top down”, i.e., by starting with the group application and trying to determine the individual behaviours that it arises from, is very difficult. Instead a “group behaviour building blocks” that can be combined to form larger, more complex applications are being developed. The robots use these behaviours to communicate, cooperate, and move relative to each other. Some behaviours are simple, like following, dispersing, and counting. Some are more complex, like dynamic task assignment, temporal synchronization, and gradient tree navigation. There are currently about forty of these behaviours. They are designed to produce predictable outcomes when used individually, are when combined with other library behaviours, allowing group applications to be constructed much more easily.
This would imply there is no central „brain‟ controlling the swarm, each robot needs to act independently [18]. They do so by following a set of simplified rules and algorithms. These rules and algorithms produce complex swarm behaviour. Algorithms like ant colony optimization (ACO) and particle swarm optimization (PSO) are some examples of algorithms that can be used in robotic swarm application [8], [19].

Ant Colony Optimization

Ant colony optimization or ACO is a meta heuristic optimization algorithm that can be used to find approximate solutions to difficult combinatorial optimization problems. In ACO artificial ants build solutions by moving on the problem graph and they, mimicking real ants, deposit artificial pheromone on the graph in such a way that future artificial ants can build better solutions. ACO has been successfully applied to an impressive number of optimization problems.
B. Particle swarm Optimization

Particle swarm optimization or PSO is a global optimization algorithm for dealing with problems in which a best solution can be represented as a point or surface in an n-dimensional space. Hypotheses are plotted in this space and seeded with an initial velocity, as well as a communication channel between the particles. Particles then move through the solution space, and are evaluated according to some fitness criterion after each time step. Over time, particles are accelerated towards those particles within their communication grouping which have better fitness values. The main advantage of such an approach over other global minimization strategies such as simulated annealing is that the large numbers of members that make up the particle swarm make the technique impressively resilient to the problem of local minima.


In near future, it may be possible to produce and deploy large numbers of inexpensive, disposable, meso-scale robots. Although limited in individual capability, such robots deployed in large numbers can represent a strong cumulative force similar to a colony of ants or swarm of bees.
III. Robot Design

The source of inspiration for a robot design is taken from nature and is shaped by environmental requirements. It gives some understanding of the general considerations needed to build a robot for robotic swarm applications. The general concept of a swarm robot design is that it should be very simple and relatively cheap to produce. For example, if one robot would get lost or broken be-yond repair during a mission it would not matter much. Quantity is the key to success in a robotic swarm to effec-tively utilize SI algorithms. A robot should also be somewhat robust, simple and effective [3]. There are several designs in robotic swarm applications that satisfy these terms. One example of such a robot is the swarm-bot, called S-bot [15], [7], [14], [13]. The S-bot as the name illustrates is designed for swarm applications and is suitable for swarm functions. The S-bot is shown in figure 1. The S-bot is 19cm high, has a diameter of 12cm and weighs around 700g, about the size of a handball.


The S-bot design was intended to mimic the ant‟s ability to grasp onto objects [3]. S-bots work as a team to find their prey2 and transport it back using the shortest path to the nest. The S-bot is a fully autonomous and mobile robot capable of clinging on to other S-bots similar to itself by using its grippers. By doing so the swarm can take any formation it wishes in the two dimensional space. The ability to do so makes the S-bots versatile. For example, the swarm can take a plough shaping formation to push an object forward or cling on to each other when crossing a gap much too wide for a single S-bot to cross [3], just like ants do [5]. The S-bot uses a two layer neural network onboard to teach the system and implement algorithms with simplified rules to produce the complex behaviours. Not all robotic swarm applications utilize neural networks [11].


Fig 1 S-bot mobile robot.

  1. Manoeuvrability

Robots have different forms of liberty to move around in the environments they were constructed for. That manoeuvrability given is called degrees of freedom (DOF) [20]. In this case each moving part or joint is a DOF. The S-bot was given five DOFs [18] to maximize movement ability in respect to the S-bots size and power consumption properties. The S-bot uses two DOF‟s for the traction system. It consists of a combination of two external wheels and a set of tracks, called treels. The treels gives the S-bot good traction in rugged terrain. This gives the advantage to overcome small obstacles. The S-bot uses another DOF to rotate the upper part of the S-bot, called the turret. The same DOF is used for the lower part, called the chassis. The 360 degree rotation gives the S- bot further freedom to grasp objects without having to reposition itself on the field. On the S-bots front the grasping mechanism can be found, that is used to produce the gripping behaviour mentioned earlier. For this purpose, one DOF is used. An additional DOF is used to move the arm or joint where the grasping mechanism is attached to. These two DOF‟s and associated parts are collectively called the gripper. Even though the S-bot does not resemble an ant it is a very robust design. All DOF‟s are set in motion by Direct Current (DC) motors. This gives enough manoeuvrability to perform the complex behaviours depending on the sensors, algorithms and rules implemented on the system.

  1. Sensors and communication

To avoid collisions with other objects or sense edges the S-bot uses infrared (IR) sensors.When a collision is about to occur the S-bot emits a sound to warn the other S-bots using loudspeakers. The sound emitted is then received with one of the omnidirectional4 microphones on-board. Further sensing is used in the gripper; it has sensors to sense if a grip has been performed successfully or not.
The S-bot utilizes a Video Graphics Array5 (VGA) cam-era to capture images for the vision system to image process and colour detect. On the centre of the turret there is a vertically placed transparent tube with a spherical convex mirror on the top. This gives the VGA camera a good overview of the surrounding environment. See figure 2. Communication is a key factor for any robotic swarm application. This is needed to give an individual in a robotic swarm better understanding of its environment and give a better sense of orientation. To communicate with other fellow S-bots the ring mounted on the turret is of importance. The ring is equipped with eight evenly distributed multi-coloured LEDs capable of emitting red, green and blue light. Different colours emitted by the LED-ring could mean different things depending on how the system is programmed. As an example red could indicate a request to grasp on to each other. There are other forms of communication like sending information in data packages wirelessly.
The S-bot has many other sensors and functionalities on-board to further help it error adjust its DOF‟s and manoeuvre with greater precision. For a more detailed over-view of the design and hardware specifications see the paper by F. Mondada, L. M. Gambardella, D. Floreano, S. Nolfi, J.-L. Deneubourg and M. Dorigo, [3].

IV. SWARM DESIGN

Looking at a swarm it works as a collective, by a common behaviour to complete a task rather than a robot performing an assignment. A swarm could be seen as one large entity. The design of a robotic swarm or multi-robot system consists of different highly integrated parts. Looking not at the physical design of the swarm robot, the swarm design is more the method a swarm functions with. Physical parameters are a condition of the swarm design, though it does not play a vital role in its construction. Emphasis is put on communication due to the fact that it is the main advantage of the swarm concept [10]. A swarm uses communication to share information among its members. If a robot gets „injured‟, it could call out for assistance [6] from others in the swarm, if such approach was deemed effective. A single robot that finds a coveted object can call out for other robots in the swarm to assist in transporting of the object, if it is too heavy for one robot to move [6]. Communications is the key to designing a functional swarm where the advantages of a flexible teamwork, much like the way humans work in groups, are visible. The issue can be encapsulated into the need of determining a functional way for the robots to communicate and react.


A. Communication

Communication is often desired to be wireless so as not to limit the moving capability of the swarm. The re-sources of the robotic platform can be limited, due to power restrictions [10], or physical size [11]. A natural approach is to lower transmission strength, processing power and memory capability. This would create a wire-less ad hoc network suffering from intra-flow interference leading up to inefficient use of power [10]. To lower the hardware requirements, processing could be moved to a centralized system that would deprive the swarm of flexibility and fault tolerance. In the article “Robot Swarm Communication Networks: Architecture, Proto-cols, and Applications” [10], there is a suggestion to create a wireless communication network in specific areas to function as a backbone ‒ a channel with high throughput for inter-robot communication, tracking and coordination. Such system could be supervised using centralized control software or a decentralized approach. The decentralized approach would have protocols to handle data fusion, swarm partitioning, energy efficiency and movement coordination. There is a need for network support such as communication load balancing, network reconfiguration, and quality of service. A functional decentralized system of this sort would result in very high fault tolerance, efficiency and flexibility, though it would still have high power consumption and need large processing capabilities. The centralized system would lower the computational and power restrictions, however lower the advantages given by the decentralized approach. Future work in this area could be of importance. The flexibility and fault tolerance of a robotic swarm depends on high numbers of robots. In a system consisting of many robots working as a group the burden of performing a task does not depend on a single robot but rather depends on the group. A robotic swarm should be constructed to be able to handle a sudden loss of a robot. Therefore there should not be a single point of failure within the swarm. Centralized processing causes a single point of failure. An alternative could be to create several points of main processing. This would increase fault tolerance and increase flexibility due to load balancing. Such main nodes could be made mobile to follow the swarm in an area where a backbone is not available.


B. Navigation

Closely coupled to the subject of communication is the topic of navigation. For a robot as well as the whole swarm it is essential to understand the surroundings in order to facilitate interaction. Navigational systems for swarms are dependent on the individual robots sensors regardless of the type of navigation. Here the need for communication to exchange knowledge of the environment is essential. Different sensor systems could be used when readings are combined within the swarm. Several robots can work together to confirm a location, lowering fault probability. Robots could be designed to exchange mapping information [4], risk-analysis and the efficiency of different methods and strategies learnt through experience.


Pre-mapped environments have been a usual navigational method and give a high efficiency level already from initiation [11]. Pre-mapped environments could also be compared to location systems that use beacons or markings such as the cricket indoor location system [17].
These methods are efficient but can require extensive preparations depending on complexity and size of the environment. Most importantly they are inflexible and must be adapted to each environment they are applied in. Methods could be implemented to simplify the preparation method and make it faster, such as a technician scanning a simple map by a using elementary skill. Such speculations are interesting but flexibility like self-learning is one of the most important aspects, especially when focusing on tasks like exploration.
C. Behaviour

Neural network implementations can be used to teach a robotic swarm system to behave according to given inputs from sensors and other onboard systems. Neural networks can be constructed in software or hardware. There are different ways of constructing a neural net-work, most of them relate to giving a winning strategy more credibility and therefore picking the solution with most credibility in a similar situation [12]. Neural training is a method that could improve problem solving and efficiency in a swarm. The technique builds on the concept of one robot teaching another [6]. A simple example of this is where positions of recharging are not known. If one robot finds a recharge station it could call out for others to gather and then lead them to the point of interest. The others could remember this point of interest, remembering the information when recharging is needed. Ant colony optimization (ACO) is an algorithm that can be used in swarm robot AI. ACO is used to solve problems regarding path finding optimizations from point A to point B. The general concept of ACO is not surprisingly taken from the ant world. Ants are generally very good at finding the shortest path to a food source by using pheromone trails while exploring their environment. Imagine a set of paths leading to the same food source with each path having different travelling distances to it. Ants normally start to explore their environment by choosing a random path emitting their pheromone as they go. The stochastic outcome would lead to a stronger concentration in pheromone in the shortest path due to the fact that a higher concentration of pheromone is the more likely to attract more ants thus giving the result of the shortest path to the source. [8]


Birds in a flock use very simple rules to keep together and avoid predators. Each bird acclimatizes and corrects themselves according to their neighbours. Birds tend to fly in alignment with other birds within their spatial awareness. If the distances to other individuals becomes too great or close, the bird adjusts the distance keeping the group together. These sets of simple rules create a bird flock‟s complex behaviour. Fish schools follow the same principle. To explain this behaviour the particle swarm optimization (PSO) algorithm has been of great importance and has been successfully implemented in reconfigurable walking robots [1], [19]. Combining algorithms like ACO and PSO with artificial neural networks can give us an efficient AI with learning capabilities. The ability to teach a system gives a huge advantage and more versatile system.
There are special applications where the robotic swarm‟s borders to multi-robot scheme like self-assembling robotic swarms. Self-assembly is a concept of a swarm constructing a larger entity out of the smaller robots it consists of, the finished product acts as an individual robot even though it consists of many small robots and can be reconfigured according to changing requirements. In such a swarm robots could react and work as group communicating over a wireless network. After self-assembly to a larger more complex entity new design principles apply. In an assembled state, were individuals connect to each other new method of locomotion such as crawling, walking on legs or rolling can be used. Communication is another example where the method changes, in an assembled connected state it would be beneficial to run communication over a bus instead of wireless communication [5]. In this state the existence of a swarm is eliminated until the assembled entity disassemblies. These are some of the aspects of designing a swarm. A swarm will grant flexibility in its work but at the same time increasing the design complexity by equal extent.
V. RECENT DEVELOPMENT

Theoretical examples are interesting; real-life properties are often different. It can be hard to actualise an idea. Further work into robots and swarms can open new paths for the robotic swarm concept. In some cases the ideas are old but until shown possible to work practically often only admired as unreal dreams. Work that has shown potential for opening the way of new technology in inspection, surveillance and reconnaissance is the mechanical fly made by Robert J. Wood. He created a mechanical fly with a wingspan of 3cm [16]; it can be seen on figure 2. If such a mechanical fly could be equipped with an internal power supply, microcontroller and sensors it could become a very powerful tool for reaching small spaces while being difficult to detect. One could directly draw the relation to Alice shown in figure 3. Alice is a wheeled robot created for the evaluation of machinery inspection by N. Correll and A. Martinoli. Their creation was designed for the purpose of inspecting machinery, foremost aircraft turbines. Hassle with drive is related to rugged terrain, Correll and Martinoli propose magnetic drive wheels or adhesive drive wheels as a solution [11]. This solution is limited and if Woods robotic fly would become a reality it might be a very usefully in this application, moving freely in all directions.



Fig 2. The mechanical fly, image from “The First Takeoff of a biologically inspired At-Scale Robotic Insect” by R. J. Wood [16].
A field of great interest is self-assembly. A good example of a self-assembling robots is Sambot [5] seen in figure 5. Sambot is a robot that can potentially be used for search and rescue in collapsed buildings. It can reconfigure and move like a worm through scrambles and then reassemble as another suitable shape such as a walker if space is given. The technique is highly flexible where the swarm can adapt to the environment. The main differences in-between Sambot and S-bot self-assembling is Sambot have the capability to self assemble in a 3D space. Sam-bot on the other hand is limited to smooth surfaces due to small thin wheels and it cannot assemble in arbitrary position, but only in perpendicular angles. ROBOTRAK is an example of a real-time system for monitoring and controlling a swarm. This system was designed as a centralized system operating on a backbone like the Internet. Even if other software for these pur-poses has been made, the ROBOTRAK software differs from many others on the basis of it controlling and monitoring a swarm in real-time [10].
One of the most fascinating areas of robotics which has been gathering pace in the last couple of years is swarm robotics. Inspired by the natural phenomenon of swarm intelligence found among social insects such as bees and ants, swarm robot technology could serve a range of functions in society from farming, searching for victims of natural disasters and medicine – if it doesn’t make you feel too queasy, imagine miniaturized robots swimming through your veins to deliver medicines.


Fig 3. Alice, the inspection robot, image from “Multirobot Inspection of Industrial Mach
Humans Invent spoke to robot expert, Antonio Espingardeiro to find out how far scientists have come in the field of swarm robotics.

VI. GOALS AND APPLICATIONS



    1. Swarm Tech

Espingardeiro says, “Swarm robotics is a very recent topic of research which involves creating artificially intelligent systems that can communicate and share information with each other. For example, machines could explore a certain room with a complex environment.  When exploring a certain room, one robot could detect there are certain obstacles in one corner of the room, upload that information to a cloud server so the other robots in that corner of the room will know where the obstacles are. All this sharing of information makes a swarm intelligent behaviour that we find in nature.”

This ability for robots to communicate with each other imitates the way ants work. A single ant is not a particularly intelligent being but when they all work together they can achieve great organizational feats. For example, by releasing pheromones, certain ants who find a good supply of food can communicate this message to the rest of the colony.

As Espingardeiro points out, the goal is for the swarm robots to communicate via the cloud but at the moment scientists are experimenting with local sharing. He says, “For the time being, the first experiments are just shared locally, i.e. there is a shared wireless connection between these robots, they are all equipped with computers so they share this information but it is not really a cloud server yet.”


    1. Robots on the cloud

Espingardeiro believes there are a huge number of applications for these swarm robots to help in dangerous situations. He says, “All situations where you need large cooperation to achieve an objective could be a potential scenario for swarm robotics. For example, if you are searching a forest in search of any traces of fire, sharing information between drones would make the system much more efficient than a single human helicopter hovering around.”

SensorFly is a project led by Professor Pei Zhang at Carnegie Mellon University, where they have created a flock of lightweight, miniature helicopters. The idea is that these could be deployed in an area where there has been a recent earthquake, searching for survivors as well as communicating with each other and human rescuers about what they have observed.



    1. Power in numbers

Another project called Symbrion (Symbiotic Evolutionary Robot Organisms) is currently working on robots that can join up like Lego in order to navigate over and around certain objects which they wouldn’t be able to do as individual units.  This is not too dissimilar to fire ant colonies that, when faced with flooding, join together to create a raft which drifts along the water until they reach dry land.

The risk of malfunction in technology is fairly high but with swarm robotics this problem is overcome by sheer numbers. As there is no one lead robot that the system relies on, if a few of the robots break down the network can carry on without too much hindrance.

Ultimately swarm robotics aims to create robots that are individually fairly unsophisticated and cheap to produce but en masse create a sophisticated system on the macro level – and it seems we have nature to thank for this idea.
VII. SCOPE

Both miniaturization and cost are key-factors in swarm robotics. These are the constraints in building large groups of robotics; therefore the simplicity of the individual team member should be emphasized. This should motivate a swarm-intelligent approach to achieve meaningful behavior at swarm-level, instead of the individual level.

Potential applications for swarm robotics include tasks that demand for miniaturization (nanorobotics, microbotics), like distributed sensing tasks in micromachinery or the human body. On the other hand swarm robotics can be suited to tasks that demand cheap designs, for instance mining tasks or agricultural foraging tasks. Also some artists use swarm robotic techniques to realize new forms of interactive art.

Robots that look more like ping pong balls could one day help to colonize Mars, so thinks their developer. The robots would work together in swarms of thousands to construct habitats for humans and perform gardening tasks.

The ping pong bots, or, “droplets” as they are known by their creator Nikolaus Correll, an Assistant Professor at the University of Colorado Boulder, are still in the prime of their development. The 20 that have been built so far have RGB color and infrared sensing, get around with vibrating motors and communicate via wireless. Down the road Correll hopes to have a completely autonomous swarm robot platform through distributed sensing, actuation, computation and communication that can be used for just about any kind of remote sensing or even construction tasks.

He views the swarm as a “liquid that thinks” with virtually no limit to what the robots could potentially accomplish, and likens them to the “swarm” of cells that make up our bodies that can serve a bewildering array of diverse functions. The robots could be deployed to contain an oil spill, says Correll, or assemble into useful bits of hardware after being launched individually into space. They could even assemble to construct a habitat on an alien planet, if humans were to explore Mars, for example.

But the droplets are still in the early developmental stages and are far away from becoming the intelligent, multi-faceted swarm that Correll foresees. The software that allows their sensors and processors to communicate is still being written and the only swarming the robots have done so far has been through computer simulations (in which hundreds were shown to coordinate). For now Correll plans to use the droplets to demonstrate self-assembly and emergent swarm behavior to carry out pattern recognition, sensor-based motion and adaptive shape change. These behaviors could eventually be scaled up to perform the same tasks in the 3D space of air or underwater.

http://singularityhub.com/wp-content/uploads/2012/12/image4a1.jpg

Fig4 Through emergent behavior swarm robots achieve behaviors much more complex than any single robot could achieve alone.

The versatile platform will be modified with different sensors depending on the task required. Correll has published the software code that facilitates communication between droplets online so other developers can build on what’s already been established. He also hopes to come up with a methodology that would allow for complex aggregations such as assembling parts of a space telescope, for example, or airplanes and jets.

The droplets should be incorporated as part of the natural evolution of a project Correll and colleagues began in 2008 to create a robotic greenhouse. The project originally involved iRobot’s Roomba-like but more programmable Create robots. The continuing goal of the project is to use a robot network to create a precision agriculture system of maximum efficiency. Sensors on the robots monitor the “local environment conditions” of the plants so that water and nutrients are delivered locally and on-demand and fruit is harvested in an optimal fashion. Combine the greenhouse with its self-assembling capabilities and the droplets could one day, Correll envisions, be used to establish “habitats and lush gardens for future space explorers.”

A swarm of tiny, simple robots probably won’t ever be as sexy as the humanoid robots like Asimo and Nao or the tireless assembly line bots like those being installed at Foxconn, but the strength of swarm robotics lie in their numbers. With hundreds, thousands, even tens of thousands of robots the amount and type of complex behaviors that could possibly emerge from a simple set of rules is virtually limitless. Their sheer numbers also means they’re disposable. If a swarm includes a thousand robots, losing even a hundred of them (because they were cheaply made) won’t appreciably affect overall performance. Swarm robotics is a field in its infancy, but thanks to ambitious developers like Correll trying to push the boundary, this exciting field of robotics is moving forward.

Imagine if you could harness the productivity of an insect colony-hundreds, if not thousands of miniature agents working together towards a larger goal – that's the future promised by swarm robotics. Potential applications, such as intelligent sensor networks, could have a wide-ranging impact on various industries. Researchers at the University of Colorado Boulder (CU-Boulder) are developing the technology with prototypes about the size of a ping-pong ball, which they have called "droplets."

The robots contain RGB color and IR (infrared) sensing, skitter about thanks to vibrating motors, and can communicate using analog/digital IR sensors. Each droplet contains an Atmel XMega 128-A3 microprocessor capable of executing code.

“Every living organism is made from a swarm of collaborating cells,” said Assistant Professor Nikolaus Correll. “Perhaps someday, our swarms will colonize space where they will assemble habitats and lush gardens for future space explorers.” Correll began working on a robotic garden at MIT in 2009, which he continues to develop. Part of that project is a model of a long-term space habitat maintained by green-thumbed robots.

Currently NASA is sending individual rovers to Mars about once a decade, and other space agencies around the world are exploring similar possibilities. Instead of a mission riding on the success or failure of one robot, however, swarm robotics may one day offer a different approach. An ideal swarm robot is cheap and disposable, so that even if a hundred of them fail, there would be still be hundreds more to take their place.

Swarms would have a multitude of uses. "Swarms of robots could be unleashed to contain an oil spill or to self-assemble into a piece of hardware after being launched separately into space," Correll explains. For now, his team is working out the basics of swarm pattern recognition, sensor-based motion, communication, and grouping into various shapes. They can test their work on thousands of robots in a computer simulation before they attempt to run code on the real droplets.

The same group worked on a large-scale interactive art exhibit called the Swarm Wall, which had 70 nodes that would move and change color, light, and sound when they detected movement. Each node would communicate with its nearest neighbors, leading to patterns of emergent behavior. You can see these two projects in the following videos.



VIII. CHALLANGES

Swarm robotics is an idea that in recent time has given rise to some interesting works. The whole field is open for discussion and development. Real life applications for robotic swarms using present technology are limited, where supervision, interactions with a swarm and energy accumulators are of large concern [11]. Systems like ROBOTRAK are needed to transform the concept of robotic swarm to a functioning and useful innovation. N. Correll and A. Martinoli touched on a very interesting topic. To increase safety and limit downtime in machinery, an intelligent sensor system would be effective. This system could consist of a swarm. This swarm could be placed inside a piece of machinery to work as an inspectional tool. While the equipment is idle rather than using fixed sensors, indicating when a problem in the machiery has occurred [11].





Fig 5. Several Sambot's that are in a self-assembled state moving like a worm. Image from “Sambot: A Self-assembly Modular Robot for Swarm Robot” by H. Wie, Y. Cai, H. Li, D. Li, T. Wang [5]

IX. CONCLUSION
When committing to the development and dispatching of such complex systems as a robotic swarm the question must be raised what the benefits are and if it is worth the effort. In fact swarm technologies have the potential to lower costs and increase efficiency. This is due to the scalability, flexibility and fault tolerance of the swarm. To reduce downtime in machinery, a robotic swarm can be used. Today much machinery needs manual inspection especially if the equipment is related to human safety. Downtime due to manual inspection can be lowered using a swarm of robots working in parallel, reaching machinery that would require cameras and dismantling of parts [11]. Remote equipment could be reached by means of a swarm. An example of this could be power line inspection where a long distance of cable needs close and visual inspection. A swarm could travel along the power lines checking them for damage. If damage is found a larger separate robot could be sent for repairs. This would speed up the inspection process, increase efficiency and lower cost. When searching for something it often requires covering large surfaces. Whether it is under water, from the air or on the ground, more units searching imply faster cover-age due to its scalability. Here the use of robots can lower cost by replacing manned search. Were robots search with a range of sensors beyond human capabilities. This would implicate in the event that the desirable object or objects are humans that lives could be saved.
These swarms could also be used in hazardous environments that are unsuitable for human interaction. Several robots also have the capability to transport objects as a group giving support one robot could not achieve [6], [18].
In surveillance of buildings and larger aerials, swarms give better coverage, fault tolerance and navigation than single robotic surveillance. In addition a swarm, com-pared to human surveillance gives lower costs [10]. From these applications we can see some of the advantages of applying a robotic swarm to an assignment that has high cost and is difficult to perform for humans. If the assignment requires a need for scalability and fault-tolerance the gain is even higher.

X. REFERENCES
[1] B. Jakimovski, B. Meyer, E. Maehle,“Swarm Intelligence for Self-reconfiguring Walking Robot”, IEEE Swarm Intelligence Symposium, St. Louis, MO, pp. 1 – 8, September 2008.

[2] C. Schroeter, H.M. Gross, “A sensor-independent approach to RBPF SLAM – Map Match SLAM applied to Visual Mapping”, IEEE/RSJ International Conference on Intelligent Robots and Systems, Nice, pp.2078 – 2083, September 2008.

[3] F. Mondada, L. M. Gambardella, D. Floreano, S. Nolfi, J.-L. Deneubourg, M. Dorigo, “The cooperation of swarm-bots: Physical interactions in collective robotics”, IEEE Robot. Automat. Mag., vol. 12, no. 2, pp. 21–28, June 2005.

[4] H. Jacky Chang, C. S. George Lee, Y. Charlie Hu, Yung-Hsiang Lu, “Multi-Robot SLAM with Topological/Metric Maps”, Intelligent Robots and Systems, 2007. IROS 2007. IEEE/RSJ International Conference on, San Diego, CA, pp.1467 - 1472, October - November 2007.

[5] H. Wie, Y. Cai, H. Li, D. Li, T. Wang, “Sambot: A Self-assembly Modular Robot for Swarm Robot”, IEEE International Conference on Robotics and Automation, Anchorage, AK, pp. 66-71, May 2010.

[6] J. S. Jennings, G. Whelan, W. F. Evans, “Cooperative Search and Rescue with a Team of Mobile Robots”, 8th International Conference on Advanced Robotics, Monterey, CA, pp. 193 - 200, Jul 1997.

[7] M. Dorigo, “SWARM-BOT: An experiment in swarm robotics”, in Proc. of the 2005 IEEE Swarm Intelligence Symposium. IEEE Computer Society Press, Los Alamitos, CA, pp. 192–200, June 2005.

[8] M. Dorigo, M. Birattari, T. Stutzle, “Ant Colony Optimization”, IEEE Computational Intelligence Magazine, vol. 1, no. 4, pp. 28 – 39, November 2006.

[9] M. G. Hinchey, R. Sterritt, C. Rouff, “Swarms and Swarm Intelligence”, IEEE Computer Society Computer, vol. 40, no. 4, pp.111 - 113, April 2007.

[10] M. Li, K. Lu, H. Zhu, M. Chen, S. Mao, B. Prabhakaran,”Robot Swarm Communication networks: Architectures, protocols, and Applications”, Third International Conference on Communications and Networking in China, Hangzhou, pp.162 – 166, August 2008.

[11] N. Correll, A. Martinoli, “Multirobot Inspection of Industrial Machinery”, IEEE Robotics & Automation Magazine, vol. 16, no. 1, pp. 103 - 112, March 2009.

[12] P. Lenander, A. Fosselius, “A survey of the SpiNNaker Project: A massively parallel spik-ing Neural Network Architecture”, IRCSE'10 conference, Västerås, October 2010.

[13] R. Groß, E. Tuci, M. Dorigo, M. Bonani, F. Mondada, “Object Transport by Modular Ro-bots that Self-assemble”, Proceedings of the 2006 IEEE International Conference on Robot-ics and Automation, Orlando, FL, pp. 2558–2564, May 2006.

[14] R. Groß, F. Mondada , M. Dorigo, ”Transport of an Object by Six Pre-attached Robots Inte-racting via Physical Links”, Proceedings of the 2006 IEEE International Conference on Robot-ics and Automation, Orlando, FL, pp. 1317–1323, May 2006.

[15] R. Groß, M. Bonani, F. Mondada, M. Dorigo, “Autonomous Self-Assembly in Swarm-Bots”, IEEE transactions on robotics, vol. 22, no. 6, pp.1115-1130, December 2006.

[16] R. J. Wood, member, IEEE, “The First Takeoff of a Biologically inspired At-Scale Robotic In-sect”, IEEE Transactions on Robotics, vol. 24, no. 2, pp. 341 - 347, April 2008.

[17] S. Das, C. Gleason, S. Shen, S. Goddard, L. C. Pérez, “2-D Tracking Performance Evaluation Using the Cricket location-Support”, IEEE International Conference on Electro Information Technology, Lincoln, NE, pp. 1 – 6, May 2005.

[18] S. Nouyan, R. Groß, M. Bonani, F. Mondada, and M. Dorigo, ”Teamwork in Self-Organized Robot Colonies”, IEEE transactions on evolutionary computation, vol. 13, no. 4, pp. 695 – 711, August 2009.

[19] Y. Shi, R.C. Eberhart, “Empirical study of particle swarm”, Proceedings of the 1999 Congress on Evolutionary Computation, vol. 3, pp. 1945-1950, Washington, DC, July 1999.

[20] "degree of freedom." Encyclopædia Britannica. 2010. Encyclopædia Britannica Online. 13 Nov. 2010 .





Download 68.4 Kb.

Share with your friends:




The database is protected by copyright ©ininet.org 2024
send message

    Main page