Multi-agent systems - Revision history

Xvegm00 at 10:04, 23 January 2019

2019-01-23T10:04:32Z

Xvegm00 at 10:04, 23 January 2019

2019-01-23T10:04:13Z

Xvegm00 at 10:25, 9 January 2019

2019-01-09T10:25:26Z

Xvegm00 at 10:24, 9 January 2019

2019-01-09T10:24:23Z

Xvegm00 at 10:04, 9 January 2019

2019-01-09T10:04:50Z

Xvegm00 at 09:59, 9 January 2019

2019-01-09T09:59:40Z

Xvegm00 at 09:58, 9 January 2019

2019-01-09T09:58:37Z

Xvegm00: /* Multiagent simulation in Netlogo */

2019-01-09T09:58:08Z

‎Multiagent simulation in Netlogo

Xvegm00: /* Introduction */

2019-01-08T21:39:14Z

‎Introduction

Xvegm00 at 21:31, 8 January 2019

2019-01-08T21:31:22Z

Show changes

@@ Line 164: / Line 164: @@
 This will cause the algae to choose the shortest paths properly.
 [[File:xvegm00-algae2.jpg]]
 Author: Bc. Martin Vegner [[User:Xvegm00|Xvegm00]] ([[User talk:Xvegm00|talk]]) 22:30, 8 January 2019 (CET)

@@ Line 164: / Line 164: @@
 This will cause the algae to choose the shortest paths properly.
-[[File:Algae2-xvegm00.jpg]]
+[[File:xvegm00-algae2.jpg]]
 Author: Bc. Martin Vegner [[User:Xvegm00|Xvegm00]] ([[User talk:Xvegm00|talk]]) 22:30, 8 January 2019 (CET)

@@ Line 24: / Line 24: @@
 ==Multiagent system==
-Creating a multiagent system or a simulation is a way of testing a hypothesis or developing autonomous robots to perform tasks. The system consists of many connected entities.<ref name="systems">Multiagent Robotic Systems <i>Jiming Liu, Jianbing Wu </i>, CRC Press 2002, 310 p. 9781420038835 [cit. 2019-01-8]</ref> This connection may origin from actual communication between agents and agents with environment; however, this connection may just stem from the fact, that they share the environment and interact with it. There may be signs of order in the system, just as there is in the world and nature. However, these are just consequences of the programming of entities and their interaction. The system is not globally controlled and the agents make decisions for themselves. Intelligent behavior is either caused by a sophisticated algorithm or emergent properties.
+Creating a multiagent system or a simulation is a way of testing a hypothesis or developing autonomous robots to perform tasks. The system consists of many connected entities.<ref name="systems2">Multiagent Robotic Systems <i>Jiming Liu, Jianbing Wu </i>, CRC Press 2002, 310 p. 9781420038835 [cit. 2019-01-8]</ref> This connection may origin from actual communication between agents and agents with environment; however, this connection may just stem from the fact, that they share the environment and interact with it. There may be signs of order in the system, just as there is in the world and nature. However, these are just consequences of the programming of entities and their interaction. The system is not globally controlled and the agents make decisions for themselves. Intelligent behavior is either caused by a sophisticated algorithm or emergent properties.
 There are a few characteristics of multiagent systems:
 *Each agent has incomplete data about the environment or lacks the ability to solve the task at hand on its own

@@ Line 1: / Line 1: @@
 ==Introduction==
-Many organisms in the world function without complex nervous systems, yet they perform feats that seem to be impossible without decisive thinking. For example, the algae Physarum Polycephalum (literally the „many-headed slime“) are a slime mold that inhabits shady, cool, moist areas, such as decaying leaves and logs. It consists of many single cell organisms that merge together in a symbiotic relationship. It has no brain, not even a single neuron and yet it shows signs of intelligence. It is capable of solving the travelling salesmen problem (TSP for short), as it connects bits of food with shortest paths of itself, to establish a network with efficient nutrient transport. It avoids obstacles or harmful substances, it remembers and even seems to be able to share information between its cells. While, there are many mysteries about Physarum yet to be solved, we already know, how it is capable of solving the travelling salesman problem. Each of Physarum cells is programmed in such a way, that they are progressively able to optimize its paths. One Physarum cell is unable to determine anything, yet many cells can perform wonders. This phenomenon is called emergence.<ref name="masfoundations">Multiagent Systems: Algorithmic, Game-Theoretic, and Logical Foundations <i>Yoav Shoham, Kevin Leyton-Brown</i>, Cambridge University Press 2009, 473 p. ISBN 0521899435 [cit. 2019-01-8]</ref>
+Many organisms in the world function without complex nervous systems, yet they perform feats that seem to be impossible without decisive thinking. For example, the algae Physarum Polycephalum (literally the „many-headed slime“) are a slime mold that inhabits shady, cool, moist areas, such as decaying leaves and logs. It consists of many single cell organisms that merge together in a symbiotic relationship. It has no brain, not even a single neuron and yet it shows signs of intelligence. It is capable of solving the travelling salesmen problem (TSP for short), as it connects bits of food with shortest paths of itself, to establish a network with efficient nutrient transport. It avoids obstacles or harmful substances, it remembers and even seems to be able to share information between its cells. While, there are many mysteries about Physarum yet to be solved, we already know, how it is capable of solving the travelling salesman problem. Each of Physarum cells is programmed in such a way, that they are progressively able to optimize its paths. One Physarum cell is unable to determine anything, yet many cells can perform wonders. This phenomenon is called emergence.<ref name="complexity">Multiagent Systems: Algorithmic, Game-Theoretic, and Logical Foundations <i>Yoav Shoham, Kevin Leyton-Brown</i>, Cambridge University Press 2009, 473 p. ISBN 0-19-515070-8 [cit. 2019-01-8]</ref>
 ==Emergence==
-Emergence is described as condition of an entity having properties that its parts do not have. This property emerges from the interaction between those parts.
+Emergence is described as condition of an entity having properties that its parts do not have. This property emerges from the interaction between those parts. <ref name="complexity2">FROM COMPLEXITY TO LIFE <i>Niels Henrik Gregersen </i>, Oxford University Press 2004, 256 p. ISBN 0521899435 [cit. 2019-01-8]</ref>
-There are two kinds of emergence, strong and weak. Weak emergence describes properties that could be derived from the properties of the single parts. This emergence can be easily reproduced in computer simulation. However, strong emergence describes entirely new properties that cannot be connected with the properties of the original parts. An example of strong emergence property is the human consciousness. There is no apparent way our neurons could produce a sense of consciousness, yet we obviously have it. It is an emergent property of the complex system that is our brain.
+There are two kinds of emergence, strong and weak.<ref name="masfoundations">FROM COMPLEXITY TO LIFE <i>Niels Henrik Gregersen </i>, Oxford University Press 2004, 180 p. ISBN 0-19-515070-8 [cit. 2019-01-8]</ref> Weak emergence describes properties that could be derived from the properties of the single parts. This emergence can be easily reproduced in computer simulation. However, strong emergence describes entirely new properties that cannot be connected with the properties of the original parts. An example of strong emergence property is the human consciousness. There is no apparent way our neurons could produce a sense of consciousness, yet we obviously have it. It is an emergent property of the complex system that is our brain.
 ==Autonomous agents==
-An agent is a simulated or artificial unit that simulates the behavior of an existing or theoretical lifeform. It can also be defined as “a singular entity in an environment, where it can perform or interact”. Agents perform tasks based on conditions they were programmed with. They don’t need to be aware of their surroundings or even of themselves. Their function is of automatons, they react to inputs, which can be sensory information or just internal information such as time, with predetermined reactions.
+An agent is a simulated or artificial unit that simulates the behavior of an existing or theoretical lifeform. It can also be defined as “a singular entity in an environment, where it can perform or interact”.<ref name="agents">Socially intelligent agents: creating relationships <i>Cathérine Tessier, Laurent Chaudron, Heinz-Jürgen Müller </i>, Springer 2002, 298 p. 9780306469855 [cit. 2019-01-8]</ref> Agents perform tasks based on conditions they were programmed with. They don’t need to be aware of their surroundings or even of themselves. Their function is of automatons, they react to inputs, which can be sensory information or just internal information such as time, with predetermined reactions.
 Agents have several characteristics:
@@ Line 24: / Line 24: @@
 ==Multiagent system==
-Creating a multiagent system or a simulation is a way of testing a hypothesis or developing autonomous robots to perform tasks. The system consists of many connected entities. This connection may origin from actual communication between agents and agents with environment; however, this connection may just stem from the fact, that they share the environment and interact with it. There may be signs of order in the system, just as there is in the world and nature. However, these are just consequences of the programming of entities and their interaction. The system is not globally controlled and the agents make decisions for themselves. Intelligent behavior is either caused by a sophisticated algorithm or emergent properties.
+Creating a multiagent system or a simulation is a way of testing a hypothesis or developing autonomous robots to perform tasks. The system consists of many connected entities.<ref name="systems">Multiagent Robotic Systems <i>Jiming Liu, Jianbing Wu </i>, CRC Press 2002, 310 p. 9781420038835 [cit. 2019-01-8]</ref> This connection may origin from actual communication between agents and agents with environment; however, this connection may just stem from the fact, that they share the environment and interact with it. There may be signs of order in the system, just as there is in the world and nature. However, these are just consequences of the programming of entities and their interaction. The system is not globally controlled and the agents make decisions for themselves. Intelligent behavior is either caused by a sophisticated algorithm or emergent properties.
 There are a few characteristics of multiagent systems:
 *Each agent has incomplete data about the environment or lacks the ability to solve the task at hand on its own
@@ Line 38: / Line 38: @@
 Agent systems are already widespread in today’s civilization. They are used whenever centralized systems with predetermined behavior aren’t sufficient. There can be two (not mutually exclusive) reasons. For one, it might be impossible, during the analysis of the applicable environment, to foresee the possible behavior of the environment. The second case is emergency situations. Centralized systems are prone to collapse in unknown situations. However, in an agent system, it does not matter that much, if few agents stop working, as the rest can still react. Due to this ability, agent systems can be used to control airspace and land transport alike. This includes even trains and ships. Today’s focus in development is autonomous vehicles, which are of course agents.
-An example of publicly functioning virtual agents are Google’s bots (also sometimes called spiders or crawlers). These are purely software agents that search the whole web and catalogue new pages as well as update the old ones. The purpose of their work is to index the internet to allow for fast function of our favorite search engine.
+An example of publicly functioning virtual agents are Google’s bots (also sometimes called spiders or crawlers). These are purely software agents that search the whole web and catalogue new pages as well as update the old ones. The purpose of their work is to index the internet to allow for fast function of our favorite search engine.<ref name="systems">Multiagent Robotic Systems <i>Jiming Liu, Jianbing Wu </i>, CRC Press 2002, 156 p. 9781420038835 [cit. 2019-01-8]</ref>
 ==Multiagent simulation in Netlogo==

@@ Line 56: / Line 56: @@
 Using random we can make the algae turn in a random direction by a set number of degrees. It doesn’t really matter what is the number of degrees, we just need randomness to the movement. However, from trying we find that 40° stimulates outward motion that mimics the radial spread of the algae and also causes a fractal pattern to the resulting algae area.
-[[File:Example.jpg]]
+[[File:Algae1-xvegm00.jpg]]
 Now we need to draw the actual algae. Physarum is composed by single cells merged together; however having hundreds of agents in the simulation could be expensive. It is better to draw the algae as patches with certain color. You might object; but the algae also optimize its paths, thus it removes itself from area! To allow optimization, we give the patches a variable popularity, which describes how often the patch gets visited. Patches which are seldom visited get erased.
@@ Line 162: / Line 162: @@
+   ]
    end
 Author: Bc. Martin Vegner [[User:Xvegm00|Xvegm00]] ([[User talk:Xvegm00|talk]]) 22:30, 8 January 2019 (CET)

← Older revision		Revision as of 09:59, 9 January 2019
Line 56:		Line 56:

	Using random we can make the algae turn in a random direction by a set number of degrees. It doesn’t really matter what is the number of degrees, we just need randomness to the movement. However, from trying we find that 40° stimulates outward motion that mimics the radial spread of the algae and also causes a fractal pattern to the resulting algae area.		Using random we can make the algae turn in a random direction by a set number of degrees. It doesn’t really matter what is the number of degrees, we just need randomness to the movement. However, from trying we find that 40° stimulates outward motion that mimics the radial spread of the algae and also causes a fractal pattern to the resulting algae area.
		+	[[File:Example.jpg]]
	Now we need to draw the actual algae. Physarum is composed by single cells merged together; however having hundreds of agents in the simulation could be expensive. It is better to draw the algae as patches with certain color. You might object; but the algae also optimize its paths, thus it removes itself from area! To allow optimization, we give the patches a variable popularity, which describes how often the patch gets visited. Patches which are seldom visited get erased.		Now we need to draw the actual algae. Physarum is composed by single cells merged together; however having hundreds of agents in the simulation could be expensive. It is better to draw the algae as patches with certain color. You might object; but the algae also optimize its paths, thus it removes itself from area! To allow optimization, we give the patches a variable popularity, which describes how often the patch gets visited. Patches which are seldom visited get erased.

@@ Line 49: / Line 49: @@
 First we need to simulate the spread of the algae. The cells have no vision, they can probably sense light levels and chemical properties of the environment. This however we cannot use. For the purposes of our simulation, the algae spread randomly.
-To make-step
+  To make-step
-set randy random 2
+  set randy random 2
-ifelse randy = 1 [rt 40][rt -40]
+  ifelse randy = 1 [rt 40][rt -40]
-fd 1
+  fd 1
-end
+  end
 Using random we can make the algae turn in a random direction by a set number of degrees. It doesn’t really matter what is the number of degrees, we just need randomness to the movement. However, from trying we find that 40° stimulates outward motion that mimics the radial spread of the algae and also causes a fractal pattern to the resulting algae area.
 Now we need to draw the actual algae. Physarum is composed by single cells merged together; however having hundreds of agents in the simulation could be expensive. It is better to draw the algae as patches with certain color. You might object; but the algae also optimize its paths, thus it removes itself from area! To allow optimization, we give the patches a variable popularity, which describes how often the patch gets visited. Patches which are seldom visited get erased.
-;the algae sets patches yellow on each step
+  ;the algae sets patches yellow on each step
-To make-step
+  To make-step
-let popularity-per-step 50
+  let popularity-per-step 50
-set randy random 2
+  set randy random 2
-ifelse randy = 1 [rt 40][rt -40]
+  ifelse randy = 1 [rt 40][rt -40]
-fd 1
+  fd 1
-set pcolor yellow
+  set pcolor yellow
-ask patch-here [
+  ask patch-here [
-set popularity popularity + popularity-per-step
+  set popularity popularity + popularity-per-step
-−
+  ]
-end
+  end
-;this function is called every tick to lower the popularity of all yellow patches
+  ;this function is called every tick to lower the popularity of all yellow patches
-to decay-popularity
+  to decay-popularity
    ask patches with [ not any? algae-here and  pcolor = yellow]  [
      set popularity popularity - (popularity-decay-rate / 10)
      if popularity < 1 [ set pcolor brown + 4]
+   ]
-end
+  end
 This gives us a nice fractal and wildly spreading algae.
@@ Line 82: / Line 82: @@
 Lets create a reporter class to tell our agent where to go.
-to-report best-way-to [ destination ]
+  to-report best-way-to [ destination ]
    ; of all the visible route patches, select the ones
    ; that would take me closer to my destination
-set collector-vision-dist  30
+  set collector-vision-dist  30
    let visible-patches patches in-radius collector-vision-dist
    let visible-routes visible-patches with [ pcolor = 45]
     let routes-that-take-me-closer visible-routes with
-[    distance destination < [ distance destination - 1 ] of myself  ]
+    [    distance destination < [ distance destination - 1 ] of myself  ]
      ifelse any? x [
      ; from those route patches, choose the one that is the closest to me
      report min-one-of routes-that-take-me-closer [ distance self ]
-  ] [
+    ] [
      ; if there are no nearby routes to my destination
      report destination
-−
+    ]
-end
+  end
+  ;Now we can move the collectors.
-Now we can move the collectors.
+   to move-collectors
+   let popularity-per-step 1000
    ask collectors [
-;standing on food is the best time to set the next goal
+ ;standing on food is the best time to set the next goal
        ifelse any? foods-on patch-here [
     set goal one-of other foods
-face best-way-to goal
+   face best-way-to goal
-jump 1
+  jump 1
-;each step we strongly increase the popularity of the patch
+  ;each step we strongly increase the popularity of the patch
-ask patch-here [
+  ask patch-here [
-set popularity popularity + popularity-per-step
+  set popularity popularity + popularity-per-step
-−
+  ]
+         ]
-End
+   End
 With enough ticks passed, the algae optimizes nicely and connects all of the food nodes. But wait, that’s not correct, is it? Physarum is able to solve the shortest path problem, that means not only connecting the nodes, but also making connections only when they are actually shorter than existing ones. To remedy this problem, we need some serious adjustments.
 Author: Bc. Martin Vegner [[User:Xvegm00|Xvegm00]] ([[User talk:Xvegm00|talk]]) 22:30, 8 January 2019 (CET)