PUBLICATIONS
Route Assignment for Autonomous Vehicles. This work examines the use of an evolutionary simulation to compute efficient routing for a large number of autonomous vehicles travelling through a network. Agents seek to competitively minimize their own travel time, while also displaying emergent cooperative behavior to utilize mulitple available routes. This work was presented at the ANTS Swarm Intelligence conference in 2016.
Coevolution with Indirect Competition and Cooperation for Traffic Routing. This is a workshop presentation relating to the above work presented at the ALIFE conference in 2016.
CURRENT RESEARCH
Effects of Cooperative and Competitive Coevolution on Complexity in a Linguistic Prediction Game (in preparation). We study the factors that lead to growth in agent complexity in a mult-species co-evolutionary simulation of language coordination. In particular, how to competitive and cooperative relationships between species impact the complexity of evolved strategies? Is pure competition the most productive path to complexity?
Modeling Evolutionary Games with Adversarial Neural Networks (in preparation). Drawing inspiration from GAN models, we model population distributions via a neural network which transforms a multi-dimensional normal distribution to a strategy distribution. This network is then trained adversarially against itself or another network to discover an equilibrium state of the underlying game.
Camouflage/detection dynamics with GANs (initial research). Attempting to use GANs to simulate the evolutionary dynamics of a predator/prey system in which prey attempts to hide within an environment using adaptive camouflage. Still in the early stage.