Simulation of Competitive Multi-Agent Search on NK Fitness Landscapes (2012)
Author: Erkin Bahceci
This package contains a simulation tool for competitive multi-agent search on NK fitness landscapes. Using strategies specified for each agent, the system simulates the competitive interactions (i.e. knowledge sharing and landscape changes) that result. Using the NKVis software, two types of visualizations of this process provided: a rectangular visualization that displays the whole space and a spherical visualization that shows a local neighborhood around a central origin point.

This example video depicts a waveriding behavior where the agent takes advantage of the changing popularity of search areas. Once an area becomes popular, its fitness increases (i.e. it is perceived as more desirable and its fitness boosted). After a while, too many agents crowd this same area, and its fitness decreases. However, in the meantime a nearby area is becoming popular. The agent depicted in the video rides this boosting wave, i.e. searches in areas that are currently being boosted. The video uses a spherical visualization of the search space, drawn from the current point's perspective. The current point first appears on a high ridge (i.e. in a boosted area); that area sinks over time as it becomes crowded, and eventually the agent moves to a new point on top of another boosted area. The previous points are shown located mostly in valleys of the landscape. Thus, this simulation in an abstract NK landscape demonstrates how innovating high-tech firms can take advantage of the dynamics of competitive multi-agent search.

Executables for Ubuntu 12.04 and OS X 10.7:
Ubuntu 12.04 (64-bit)
Ubuntu 12.04 (32-bit)
OS X 10.7 (64-bit)

For instructions on how to run this visualization tool, see the README.txt file.

Erkin Bahceci Ph.D. Alumni erkin [at] cs utexas edu
Risto Miikkulainen Faculty risto [at] cs utexas edu
Riitta Katila Collaborator rkatila [at] stanford edu
Competitive Multi-Agent Search Erkin Bahceci PhD Thesis, Department of Computer Science, The University of Texas at Austin, Austin, TX 78712, Dec... 2014

Evolving Strategies for Competitive Multi-Agent Search Erkin Bahceci, Riitta Katila, and Risto Miikkulainen arXiv:2306.10640, 2023. 2023

NKVis This package contains a 3D visualization tool for NK fitness landscapes. Two types of visualizations are provided: a 2011