Gaspar I. Melsion
Papers
2
Total Citations
24
H-Index
2
About
Gaspar I. Melsion is a researcher at the intersection of safe artificial intelligence and human-robot collaboration, with a primary focus on reconciling formal safety guarantees with practical, real-world deployment. His most influential work, "Human-Feedback Shield Synthesis for Perceived Safety in Deep Reinforcement Learning" (2021, 20 citations), addresses a critical bottleneck in deep RL: the tension between rigid, adversarial formal verification and the need for flexible, socially-aware behavior. Melsion’s key contribution is a shield synthesis method that incorporates human feedback, allowing agents to learn safe policies without being overly constrained—a significant step toward trustworthy autonomy in human-centric environments. In complementary work on "Leveraging Explainability for Comprehending Referring Expressions in the Real World" (2021, 4 citations), he tackles the challenge of ambiguous human requests in human-robot interaction, demonstrating how explainable AI can enable robots to ask intelligent follow-up questions. While his citation counts reflect an early-career stage, the conceptual novelty of his safety framework positions him as a rising voice in the growing field of human-aligned reinforcement learning, bridging formal methods with user-centered design.
Research Focus
Key Achievements
Top Papers
- 1
- 2