Unity ML-Agents: Autonomous Systems Research

Research

About The Project

Reinforcement learning research projects for autonomous drone delivery and crossroad vehicle navigation

Key Achievements

—Participated in 2021 RLKR Drone Delivery Challenge developing autonomous drone navigation with Unity ML-Agents

—Implemented complex crossroad autonomous driving environment with multi-agent collision detection and interaction

—Designed realistic 8-lane crossroad simulation based on Gangnam Station area using BlenderGIS for accurate building models

—Developed checkpoint system with reward function optimization to prevent unintended learning patterns

—Applied Epsilon-greedy algorithm (ε=0.9) for balanced exploration-exploitation in multi-agent scenarios

—Configured Ray Perception Sensor 3D for agent detection and collision avoidance in dense traffic environments

—Presented research at Korean Society of Information Science and Technology conference and advanced to finals

—Demonstrated applicability of reinforcement learning to complex real-world autonomous navigation scenarios

Project Files

More Projects