• Journal of Korea Game Society
• /
• v.17 no.5
• /
• pp.123-132
• /
• 2017

The path-finding algorithm refers to an algorithm for navigating the route order from the current position to the destination in a virtual world in a game. The conventional path-finding algorithm performs graph search based on cost such as A-Star and Dijkstra. A-Star and Dijkstra require movable node and edge data in the world map, so it is difficult to apply online games with lots of map data. In this paper, we provide a Heuristic-based Genetic Algorithm Path-finding(HGAP) using Genetic Algorithm(GA). Genetic Algorithm is a path-finding algorithm applicable to game with variable environment and lots of map data. It seek solutions through mating, crossing, mutation and evolutionary operations without the map data. The proposed algorithm is based on Binary-Coded Genetic Algorithm and searches for a path by performing a heuristic operation that estimates a path to a destination to arrive at a destination more quickly.

• The Journal of Korea Robotics Society
• /
• v.14 no.3
• /
• pp.157-162
• /
• 2019

The video game Tetris is one of most popular game and it is well known that its game rule can be modelled as MDP (Markov Decision Process). This paper presents a DQN (Deep Q-Network) based game agent for Tetris game. To this end, the state is defined as the captured image of the Tetris game board and the reward is designed as a function of cleared lines by the game agent. The action is defined as left, right, rotate, drop, and their finite number of combinations. In addition to this, PER (Prioritized Experience Replay) is employed in order to enhance learning performance. To train the network more than 500000 episodes are used. The game agent employs the trained network to make a decision. The performance of the developed algorithm is validated via not only simulation but also real Tetris robot agent which is made of a camera, two Arduinos, 4 servo motors, and artificial fingers by 3D printing.

• Journal of Korea Multimedia Society
• /
• v.25 no.2
• /
• pp.432-439
• /
• 2022

This paper introduces deliberative behavior tree using utility theory. The proposed approach combine the strengths of behavior trees and utility theory to implement complex behavior of AI agents in an easier and more concise way. To achieve this goal, we devised and implemented three types of additional behavior tree nodes, which evaluate utility values of its own node or its subtree while traversing and selecting its child nodes based on the evaluated values. In order to validate our approach, we implemented a sample scenario using conventional behavior tree and our proposed deliberative tree respectively. And then we compared and analyzed the simulation results.

• Proceedings of the Korean Society of Computer Information Conference
• /
• 2023.01a
• /
• pp.25-26
• /
• 2023

본 논문에서는 OpenAI Gym 환경에서 제공하는 CartPole-v1에 대해 강화학습을 통해 에이전트를 학습시키고, 학습에 적용되는 활성화함수의 성능을 비교분석하였다. 본 논문에서 적용한 활성화함수는 Sigmoid, ReLU, ReakyReLU 그리고 softplus 함수이며, 각 활성화함수를 DQN(Deep Q-Networks) 강화학습에 적용했을 때 보상 값을 비교하였다. 실험결과 ReLU 활성화함수를 적용하였을 때의 보상이 가장 높은 것을 알 수 있었다.

• The KIPS Transactions:PartB
• /
• v.11B no.2
• /
• pp.207-212
• /
• 2004

This paper presents a reinforcement learning framework for stock trading systems. Trading system parameters are optimized by Q-learning algorithm and neural networks are adopted for value approximation. In this framework, cooperative multiple agents are used to efficiently integrate global trend prediction and local trading strategy for obtaining better trading performance. Agents Communicate With Others Sharing training episodes and learned policies, while keeping the overall scheme of conventional Q-learning. Experimental results on KOSPI 200 show that a trading system based on the proposed framework outperforms the market average and makes appreciable profits. Furthermore, in view of risk management, the system is superior to a system trained by supervised learning.

• KIPS Transactions on Computer and Communication Systems
• /
• v.3 no.12
• /
• pp.455-462
• /
• 2014

In recent years, various services in mobile environments due to the growth of mobile devices and related techniques like wireless networks. Furthermore, as the increasing communication traffic in cellular networks has become a new significant issue, many studies for device-to-device(D2D) communication and D2D-based cooperative services have been performed recently. In this paper, we design a smart agent system for D2D-based cooperative services and propose a dynamic service binding method based on service ontology. We classify roles of mobile devices for cooperative services by defining three types of smart agents, and construct a knowledge base in order to describe properties of 'service' unit. The proposed knowledge model, D2D cooperative service ontology, can enable a autonomous cooperative services between mobile devices by binding a requested service to the appropriate member device according to the real-time context in mobile environments.

• Journal of the Korea Society of Computer and Information
• /
• v.11 no.4 s.42
• /
• pp.271-281
• /
• 2006

Due to the development of Internet, number of online documents and the amount of web services are increasing dramatically. However, there are several procedures required, before you actually find what you were looking for. These procedures are necessary to Internet users, but it takes time to search. As a method to systematize and simplify this repetitive job, this paper suggests a DotQuery based interactive search agent. This agent enables a user to search, from his computer, a plenty of information through the DotQuery service. which includes natural languages. and it executes several procedures required instead. This agent also functions as a plug-in service within general web browsers such as Internet Explorer and decodes the DotQuery service. Then it analyzes the DotQuery from a user through its own program and acquires service results through multiple browsers of its own.

• Journal of the Korean Society for Aeronautical & Space Sciences
• /
• v.40 no.3
• /
• pp.215-221
• /
• 2012

In this paper, we introduce an optimization method to improve pursuit performance of a pursuer in a pursuit-evasion game (PEG). Pursuers build a probability map and employ a hybrid pursuit policy which combines the merits of local-max and global-max pursuit policies to search and capture evaders as soon as possible in a 2-dimensional space. We propose an episodic parameter optimization (EPO) algorithm to learn good values for the weighting parameters of a hybrid pursuit policy. The EPO algorithm is performed while many episodes of the PEG are run repeatedly and the reward of each episode is accumulated using reinforcement learning, and the candidate weighting parameter is selected in a way that maximizes the total averaged reward by using the golden section search method. We found the best pursuit policy in various situations which are the different number of evaders and the different size of spaces and analyzed results.

• Journal of the Korea Society for Simulation
• /
• v.24 no.4
• /
• pp.107-118
• /
• 2015

Many approaches to agent collaboration have been introduced in military war-games, and those approaches address methods for simulation entity (actor) collaboration within a team to achieve given goals. To meet fast-changing battlefield situations, an actor must be loosely coupled with their tasks and be able to take over the role of other actors if necessary to reflect role handovers occurring in real combat. Achieving these requirements allows the transfer of tasks assigned one actor to another actor in circumstances when that actor cannot execute its assigned role, such as when destroyed in action. Tight coupling between an actor and its tasks can prevent role handover in fast-changing situations. Unfortunately, existing approaches and war-game strictly assign tasks to actors during design, therefore they prevent the loose coupling. To overcome these shortcomings, our Role-based Command Hierarchy (ROCH) model dynamically assigns roles to actors based on their situation at runtime. In the model, "Role" separates actors from their tasks. In this paper, we implement the ROCH model as a component that uses a publish-subscribe pattern to handle the link between an actor and the roles of its subordinates (other actors).

• Journal of the Korea Society of Computer and Information
• /
• v.29 no.2
• /
• pp.61-66
• /
• 2024

This paper aims to design a simulation for path-finding of objects in a simulation or game environment using reinforcement learning techniques. The main feature of this study is that the objects in the simulation are trained to avoid obstacles at random locations generated on a given track and to automatically explore path to get items. To implement the simulation, ML Agents provided by Unity Game Engine were used, and a learning policy based on PPO (Proximal Policy Optimization) was established to form a reinforcement learning environment. Through the reinforcement learning-based simulation designed in this study, we were able to confirm that the object moves on the track by avoiding obstacles and exploring path to acquire items as it learns, by analyzing the simulation results and learning result graph.