Date of Award

5-2019

Document Type

Thesis open access

Department

Computer Science

First Advisor

Matthew A. Hibbs

Second Advisor

Xin A. Jiang

Abstract

The ubiquity of human-like characters in video games presents the challenge of implementing human-like behaviors. To address the pathfinding and behavior selection problems faced in a real project, we came up with two improved methods based upon mainstream solutions. To make pathfinding agent take into account more incentives than only a destination, We designed a new pathfinding algorithm named Cost Radiation A* (CRA*), based on the A* heuristic search algorithm. CRA* incorporates the agent's preference for other objects, represented as cost radiators in our scheme. We also want to enable non-player characters (NPCs) to learn in real-time in response to a player's actions. We adopt the behavior tree framework, and design a new composite node for it, named learner node, which enables developers to design learning behaviors. The learner node achieves basic reinforcement learning but is also open to more sophisticated use.

Download

Share

COinS