Linji (Joey) Wang

I am a Ph.D. student in Computer Science at George Mason University, working under the supervision of Dr. Xuesu Xiao at the RobotiXX Lab. Our lab pioneers adaptive learning paradigms for autonomous robotics, and my research addresses a fundamental challenge in robotics: how can robots learn complex behaviors as efficiently as humans do, adapting from simple to complex tasks while maintaining robustness in real-world deployments?

My work centers on Curriculum Learning for Robotics - developing intelligent systems that learn progressively, much like how humans master skills through structured practice. I believe the future of robotics lies not in hand-crafted solutions, but in self-improving systems that can adapt their learning strategies based on performance feedback. This vision has led me to develop two complementary frameworks:

Grounded Adaptive Curriculum Learning (GACL) introduces a teacher-student paradigm where an informed teacher agent automatically generates training curricula by monitoring student performance in real-time. This approach bridges the gap between simulation and reality by grounding curriculum generation in actual robot capabilities. Reward Training Wheels complements this by providing adaptive auxiliary rewards that guide robots through challenging learning phases - like training wheels that automatically adjust and eventually disappear as proficiency increases.

My research has been validated across diverse robotic platforms:

• Quadrupedal robots navigating complex terrains
• Ground navigation vehicles in constrained environments
• Off-road autonomous vehicles handling unpredictable dynamics
• Current research: Extending to humanoid robots for manipulation tasks

This work, resulting in three papers accepted at IROS 2025 and one in IEEE RA-L, demonstrates 24.58% higher success rates and 50% improved sample efficiency compared to state-of-the-art methods. Through collaboration with my advisor Dr. Xuesu Xiao and renowned researchers including Peter Stone (UT Austin), I am advancing the frontiers of adaptive curriculum learning and decremental dynamics planning for robust robot autonomy in complex, dynamic environments.

Prior to my Ph.D., I earned my M.Sc. in Mechanical Engineering from Carnegie Mellon University (GPA: 3.94/4.0), where I developed expertise in 3D perception and AR-guided robotics. I also gained industry experience as a Software Development Engineer Intern at Amazon AWS, where I built production-grade systems that reduced performance analysis time from 8 hours to 15 minutes.

My vision is to create robots that learn like we do - starting simple, building competence, and ultimately mastering complex real-world tasks through intelligent curriculum design. This research has profound implications for service robotics, industrial automation, and human-robot collaboration.

Feel free to explore my publications, projects, and CV to learn more about my research and experience.