Evolutionary Computing Systems Lab · ECSL · UNR
Genetic Algorithms · Game AI · Adaptive Training · Autonomous Navigation · Cybersecurity
Current Projects
Featured · ONR-Funded · N00014-22-1-2122
We designed and built a ship driving simulation trainer using knowledge tracing to adapt scenario difficulty in real time. Scenarios include multi-ship encounters set in Singapore and San Diego harbors covering all major COLREGs encounter types: head-on, crossing giveway, crossing standon, and overtaking. A comparative study of adaptive vs. non-adaptive training showed adaptive training is statistically significantly more effective (p < 0.0001), with 73% of students preferring it. The system is deployed in collaboration with the US Navy. Supported by the Office of Naval Research, grant N00014-22-1-2122.
Read the 2025 Springer paper →
Autonomous Navigation · COLREGs
VORRT-COLREGs combines velocity obstacles (VO) with rapidly-exploring random trees (RRT) to plan COLREGs-compliant trajectories for autonomous surface vessels in dynamic multi-ship environments. VO handles real-time local collision avoidance; RRT provides the global waypoint structure. Published in IEEE Robotics and Automation Letters, 2022.
IEEE paper →
Serious Games · Cybersecurity
TAISER is a multiplayer Unity-based cybersecurity training game where human and AI players take WHITEHAT or BLACKHAT roles defending or attacking a city network infrastructure. A CyberAI teammate provides real-time firewall rule recommendations, helping players learn to classify and respond to network packets. Supports mixed human/AI teams and adaptive difficulty.
Serious Games · Naval Training
A simulation-based game for training Naval Surface Warfare Officers to rapidly identify target angle — the bearing of a ship relative to an observer. The game presents scenarios with various vessel types and lighting conditions, requiring players to classify the encounter before selecting a COLREGs-compliant action. Presented at IEEE CoG 2024.
IEEE CoG 2024 paper →
Game AI · Evolutionary Computing
Using potential field-based co-evolutionary algorithms to develop sophisticated unit micro for StarCraft II. We have evolved Terran Marines, Marauders, and Medivacs against Zerg units. Recent work investigates NEAT-based neuroevolution and multi-objective co-evolutionary approaches for generalizable, emergent combat tactics that transfer across scenarios.
Related publications →
Evolutionary Computing · Networks
Genetic algorithms evolve dynamic topologies for UAV-hosted mesh networks, optimizing communication coverage and reliability under changing conditions. The evolved networks adapt to node failures and changing mission requirements. Presented at IEEE Congress on Evolutionary Computation, 2020.
See publication →Past Projects
The foundational ECSL research area. CIGAR improves GA performance by injecting solutions from past similar problems into a running GA's population. Applied to games, scheduling, circuit design, and optimization. Published in IEEE Transactions on Evolutionary Computation (2004, 2005).
Unity 3D simulation training and control system (STACS) for robotic bridge inspection. Developed in collaboration with the ARA Laboratory on the INSPIRE project. Includes ROS integration and VR-enabled operator interface for controlling robot inspection teams.
Interactive GAs where human preferences guide the evolutionary search, applied to user interface layout design. Developed the Sycophant API for research in context-aware adaptive interfaces. Published at GECCO, CEC, and IEEE IUI.
Ph.D. students interested in these research areas: contact me · Full lab: ecsl.cse.unr.edu