bio-inspired artificial intelligence for collective systems

17
Bio-inspired Artificial Intelligence for Collective Systems Name :Achini Adikari Index No : 104002P Supervisor : Dr. H. Thilak Chaminda Faculty of Information Technology University of Moratuwa

Upload: achiniadikari

Post on 12-Jan-2017

156 views

Category:

Technology


3 download

TRANSCRIPT

Page 1: Bio-inspired Artificial Intelligence for Collective Systems

Bio-inspired Artificial Intelligence for Collective Systems

Name :Achini AdikariIndex No : 104002PSupervisor : Dr. H. Thilak Chaminda

Faculty of Information TechnologyUniversity of Moratuwa

Page 2: Bio-inspired Artificial Intelligence for Collective Systems

Introduction• Nature is the most organized dynamic system • Behaviors of these systems have optimal adaptations to any kind of

critical situation. • Systems developed in AI needs to do the correct thing at the correct

time• Collective systems in AI needs to have a balance between

components and adapt to complex scenarios• These could be influenced by Natural Collective Systems• Thus, Swarm AI concept was introduced.

Page 3: Bio-inspired Artificial Intelligence for Collective Systems

In collective AI systems there is a need to,• Take decisions which have beneficial effects for all the components• Use local information among sub components and systems• Adapt to catastrophic and complex situations

Background and Motivation

Collective systems in nature are,• Self organized• Naturally adaptable to complex situations• Have non linear interactions between each other• Chooses the best option over many

Page 4: Bio-inspired Artificial Intelligence for Collective Systems

Overview – Swarm Intelligence• Swarm Intelligence is the study of collective behaviors of systems of

nature, mainly insects and birds• Swarm AI is based on two main concepts which are self-

organization and Stigmergy.

There are four main swarm models, • Ant Colony Optimization• Ant Clustering Model• Particle Swarm Optimization model• Bird Flocking Model

Basic Structure of a Swarm Technique

Page 5: Bio-inspired Artificial Intelligence for Collective Systems

Ant Colony Optimization• The ant agent keeps a record of visited nodes and the time elapsed for arrival. • It will return following the same path and updates the digital pheromone value on the links that it passes by. • The pheromone level decides the speed of the transmission.

Ant Clustering Model • Agent (ant) action rule is that the agent moves randomly in the

grid. • They only recognize objects which are immediately in front of

them. • Picking up or dropping item is based on pickup probability and

drop probability

Page 6: Bio-inspired Artificial Intelligence for Collective Systems

Particle Swarm Optimization• Particles move through the solution space, and are evaluated according to some fitness criterion after each timestamp

Page 7: Bio-inspired Artificial Intelligence for Collective Systems

Bird Flocking Model

• Basic models of flocking behavior are controlled by three simple rules:– Separation - avoid crowding neighbors (short range repulsion)– Alignment - steer towards average heading of neighbors– Cohesion - steer towards average position of neighbors (long range

attraction)

Page 8: Bio-inspired Artificial Intelligence for Collective Systems

Researches related to Swarm AISwarm Intelligence for Networking Principles and applications of swarm intelligence for adaptive routing in telecommunications networks• Study about the concepts of Wireless and telecommunication networks using

swarm intelligent agents• They have studied many applications of the Swarm Intelligence paradigm,

considering routing algorithms for wired and wireless networks, best-effort and quality-of-service networks.

Multicast Routing for Mobile Ad-Hoc Networks using Swarm Intelligence• The study is done regarding group communication applications which demand

a large degree of coordination and have highly dynamic group membership changes

• Presented an alternate approach to solve the multicast routing problem in mobile ad hoc networks

Page 9: Bio-inspired Artificial Intelligence for Collective Systems

Swarm Intelligence for Data Mining• Two broad categories of Swarm AI, Effective Search and Data

organizing were studied.• The benchmarking experiments done in this research showed that

ant-based clustering performs better than other techniques:

Page 10: Bio-inspired Artificial Intelligence for Collective Systems

Applications of Swarm AI• Concept of Ant colony Optimization is used in Southwest Airlines . They

are implementing and studying more about this technique and has got impressive feedbacks

• The US Military uses swarm techniques to control unmanned vehicles. The need to find the optimal path and best alternatives this foundation is being used.

• Particle Swarm Optimization is used in the theory of social interaction to problem solving. Particles can be regarded as simple agents that fly through the search space and communicate the best solution that they have reached.

• NASA has developed systems to investigate planetary mapping and controlling micro satellites with the use of swarm technologies

Page 11: Bio-inspired Artificial Intelligence for Collective Systems

• Using the concept of Ant based Routing, routing packets, reinforcement of routing forward, backward and both directions have been researched in telecommunication networks

• Location of transmission infrastructure for wireless communication networks is also addressed using these techniques.

• Birds flocking model is heavily used in film industry, animations and as well in controlling unmanned air vehicles

• In film production, swarm techniques are used in rendering and to generate Complex interactive virtual environments, Break the Ice, Lord of the Rings

• Data Mining, data sensoring in router networks are also inspired by the collective behaviors of natural systems.

• Swarm techniques are used in cargo arrangement in airline companies, route scheduling in delivery companies and in power grid optimization control.

• Research state that swarm techniques could be used to control nanobots within the body to kill cancer tumors.

Page 12: Bio-inspired Artificial Intelligence for Collective Systems

DiscussionAdvantages:• The natural simplicity of Swarm AI agents and their communication

makes it easier to understand and results in a fast design process of a Swarm AI system

• Agents in Swarm AI systems are necessarily fast hence they are very efficient

• memory requirements are limited since these systems have simple reactive and utility based agents which do not store previous information

• Systems are robust and have adaptive nature with good performance.

Page 13: Bio-inspired Artificial Intelligence for Collective Systems

Drawbacks• Swarm AI systems are not applicable in instances where exact

results are required since they provide approximate solutions. • Expensive system methodologies• Increasing the number of processing units in an agent will have

complexity issues when communicating and coordinating with other sub systems

Page 14: Bio-inspired Artificial Intelligence for Collective Systems

Algorithm Special FeaturesAnt Colony Optimization It allows dynamic rerouting through shortest path if one

node is broken whereas other algorithms consider the path to be static

Inherent parallelism Positive Feedback leads to rapid discovery of good

Solutions

Particle Swarm Optimization This does not have any overlapping and mutation calculations

Based on theories and easy to calculate PSO does not have genetic operators such as crossover

and mutation Can be applied into both scientific research and engineering

use Cannot work out the problems of scattering and optimization

Bird Flocking Model Collision avoidance mechanisms Centralization and coordination between components

Page 15: Bio-inspired Artificial Intelligence for Collective Systems

Future work• One of the core focus areas is Data Mining and data clustering

where those can be inspired by swarm techniques.

• Prospects of having complex routing and telecommunication systems.

• Research is being done regarding astronomy for satellites which are auto mated.

• Robotics robots can be modeled to imitate the behavior of natural organisms.

• Binding Swarm AI techniques with other artificial intelligence models and algorithms, combination of many models will compensate loop holes of some algorithms and will make an efficient practice

Page 16: Bio-inspired Artificial Intelligence for Collective Systems

ReferenceBinitha S, S Silva Sathya, "A survey of Bio Inspires Optimization Algorithms" ISSN: 2231-2307, Volume-2, Issue-2, May 2012

Dr. Xiaohui CuiApplied Software Engineering Research GroupOak Ridge National Laboratory,Swarm Intelligence, Bio-inspired Emergent Intelligence Mano Jean-Pierre, Bourjot Christine, Lopardo Gabriel, Glize Pierre, Bio Inspired Mechanisms for Artificial Self-organized systems Falko Dressler and Ozgur B. AkanComputer Networks and Communication Systems, Dept. of Computer Sciences, University of Erlangen, Germany, Bio Inspired networking

Mrs. B.D. Shirodkar, Dr. S.S.Manvi, A.J.Umbarkar, Multicast Routing for Mobile Ad-Hoc Networks using Swarm Intelligence

David Martens, Bart Baesens · Tom FawcettSwarm Intelligence for Data Mining

Page 17: Bio-inspired Artificial Intelligence for Collective Systems

Thank You…