Swarm Development Tools. Ricardo Hoar

Transcription

1 Swarm Development Tools Ricardo Hoar

2 Swarms Emergent global behaviour from many parallel local interactions Relatively simple local rules can produce complex results Since this idea can be applied to many types of simulation generalized development tools have been created.

3 Swarm Development Tools/Libraries Starlogo is developed by Mitchel Resnick at MIT Media Lab Breve is written by Jon Klein, Chalmers University Repast is developed by Social Science Research Computing, University of Chicago. Swarm is originally developed at the Santa Fe Institute.

4 StarLogo, NetLogo NetLogo is a programmable modeling environment for simulating natural and social phenomena. Discrete (CA) environment allows for many types of simulation but limits to 2D (although 3D simulations can be created and viewed in 2D). The emergence of global phenomena from the interaction between agents is observed (Swarm). StarLogo is intended for ages 13 and up. (Relatively simple syntax)

5 StarLogo characters Turtles. can represent almost any type of object: an ant in a colony, a car in a traffic jam, an antibody in an immune system, a molecule in a gas. Each turtle has a position, a heading, a color, and a "pen" for drawing. You can add more specialized traits and properties. Patches. Patches are pieces of the world in which the turtles live. Like turtles, patches can execute StarLogo commands, and they can act on turtles and patches. Patches are arranged in a grid, with each patch corresponding to a square in the Graphics area. Observer. The observer "looks down" on the turtles and patches from a bird s eye perspective. The observer can create new turtles, and it can monitor the activity of the existing turtles and patches.

6 StarLogo Random Walker To setup ca //C.A. environment crt 1 //create 1 turtle ask turtles [set colour yellow ] //<--Makes turtle yellow End To do-random-walk ask turtles [ rt random 360 //<--Rotate a random angle in 360 forward 1 //<--Move forward 1 stamp blue //stamp the cell colour blue wait 0.1 //wait (so as not to run too quick) ] End

7 Breve Breve is an environment for developing realistic simulations of decentralized systems and artificial life. Users define the behaviours of agents in a 3D world and see how they interact. The user can navigate the 3D display interactively, viewing a scene from different angles. An interpreted object-oriented language (steve), an OpenGL display engine, collision detection and articulated body physical simulation are included. Unfortunately, the physical simulation is still an experimental feature.

8 Wanderer code Controller mycontrol. Control : mycontrol { + variables: walkershape (object). + to init: print "Setting up the simulation.". self point-camera at (0, 0, 0) from (0, 60, 0). walkershape = new Shape. walkershape init-with-sphere radius new RandomWalker. //<--Over 1000 halts the comp. + to get-walker-shape: return walkershape. } Mobile : RandomWalker { + to init: self register with-shape (controller get-walker-shape). self set-color to random[(1.0, 1.0, 1.0)]. + to iterate: self set-velocity to random[(40, 40, 40)] - (20, 20, 20).}

9 Repast RePast envisions a simulation as a state machine whose state is constituted by the collective states of all its components. It provides a library of classes for creating, running, displaying and collecting data from an agent based simulation. Requires Java to run. Admittedly Swarm like

10 Swarm Like Repast, proivides libraries that work with Java or objective C, providing the programmer with access to a real language to develop with. Pre defined functions and classes ease development by providing black boxes which can be called by the user. Eg. Can do diffusion in the environment by calling diffusion on such an object without having to implement the diffusion yourself. By having access to the java/c source code implementation details are not hidden and can be seen and modified.

11 Example Some_agent.randomWalk(); public void randomwalk () { moveagent (Globals.env.uniformIntRand.getIntegerWithMi n$withmax (-scatter, scatter), Globals.env.uniformIntRand.getIntegerWithMi n$withmax (-scatter, scatter)); }

12 Similarities Languages / libraries provide the basis of the simulation, allowing the developer to concentrate on agent programming rather than implementation details. Visualization and interface is provided to ease development. Provides the framework for swarm systems so ideas can be implemented and tested quickly.

13 Differences StarLogo, Swarm and Repast use a discrete environment Breve becomes unusably slow with over 1000 agents (no swarm) Breve has a physics engine but it is only in an experimental state. Repast and Swarm harness the power of a programming language providing more control, and are thus more difficult to learn for non-programmers. Breve uses a continous model. Other systems allow for full swarm simulation (slowing down with over 100, 000) Other systems do not have experimental physics. Breve and StarLogo have their own languages making code un-portable and unfamilar

14 Conclusions Netlogo provides an adequate and complete discrete model that is easy to learn and use, but that is limited by the simulations discreteness. Breve provides a 3d model but can not handle a large number of agents, effectively eliminating it a a swarm tool. It also eliminates access to the implementation details rendering a detailed analysis of the results impossible. This is the only 3D visualization discussed today. Repast and Swarm, building upon an existing language, provide the best building blocks for would be developers, but require Java or C programming experience. Additionally, the libraires provided work only in discrete 2D space.

15 References Starlogo Repast Swarm Breve Klein, J breve: a 3D simulation environment for the simulation of decentralized systems and artificial life. To appear in Proceedings of Artificial Life VIII, the 8th International Conference on the Simulation and Synthesis of Living Systems. Anderson, Phillip, Seaquest, CarlL R. using starlogo to introduce differential equations.