SUB-TASK 4: Implementation Plan

Goal

The goal is to benchmark the simulator to see how large an experiment we should be able to run

Implementation

The benchmark simulations will be done by Vojta and Daniel of Prague University.

Benchmarks will be held using KIT robots in a simple flat environment with walls. We measure simulation time on certain system (say a 3Ghz Pentium), but also want to measure additional things like times of physics/graphics/collisions etc. Daniel is looking into efficiently measure the other things.

Berend will supply the more intricate organism forms, as well as a controller that simulates the computational effort of an evolutionary algorithm.

Vojta is investigating using logging of robot positions to create movies later, instead of using a global camera.

We have decided on doing the following benchmarks:

  • Simulations of 5, 10, 50, and 100 robots in swarm mode
  • Simulations of 5, 10, 50, and 100 robots in organism mode, where the organism is a snake, and the controller a sine generator with shifted phase
  • Simulations of 5, 10, 50, and 100 robots with camera enabled (On request of Anne)
  • Simulations of 5, 10, 50, and 100 robots with most simple distance sensor enabled
  • Simulations of 5, 10, 50 100 robots in organism mode, in which the organism is very intricate, for instance one which is asymmetrical and connected to itself at more than one point, perhaps an A, or perhaps a double square?.
  • Simulations of 20, 50, and 100 robots, with 50 - 80% of the robots are in different organisms (divided equally among organisms of size 4-10), where the organisms are an H and the controller is a sine, and 50% of the robots in swarm mode, which are moving around (semi) randomly.
    • For 20 robots: 2 organisms of size 5
    • For 50 robots: 5 organisms of size 5
    • For 100 robots: 5 organisms of size 10
    • 100 robots: 16 organisms of size 5
Unless otherwise stated, the content of this page is licensed under Creative Commons Attribution-NonCommercial-ShareAlike 3.0 License