Adam HaroonCross-domain robotic intelligence
A selection of my research projects, from factory floors to the stratosphere. Each one is an experiment in the same program: find the hidden failure, then guarantee against it. See the publications page for the complete record.
Trusting what autonomy does
Learning When to Act
Comm-efficient RL under a Lyapunov run-time assurance shield: 3.51× fewer control updates, same safety certificate.
NeurIPS 2026 · under review
High-Altitude Balloon Swarms
Distributed area coverage with multi-agent RL in stochastic stratospheric flow fields, with NRL.
IEEE Aerospace 2026 · published
Urban Air Mobility
Urban Nav simulation framework and graph-encoded MARL for zero-shot scalable conflict resolution over an Austin skyline.
IEEE Aerospace 2026 · published
Trusting what autonomy sees
VIRTUS-FPP
Physics-faithful virtual FPP sensors in NVIDIA Isaac Sim: trustworthy synthetic data for learned 3D perception.
IEEE Sensors Journal · in press
Shortcut Diagnosis & PhiCalNet
Diagnosing shape-prior shortcuts in long-range single-shot FPP, then repairing them with fixed differentiable physics.
Elsevier Optics · under review
Differentiable Self-Calibration
Self-calibrating single-shot FPP with uncertainty propagation, and pattern design for identifiability.
Elsevier Optics · under review
Connecting the pillars
LCLA: Vision-Language Navigation
Aligning VLM vision and language features to the latent space of a frozen privileged policy: perception that plugs into a controller that already knows how to act.
CoRL 2026 · under review
Safe Preference-Based RL
State-conditioned safe offline preference-based reinforcement learning: policies learned from VLM-labeled preferences without trading away safety.
ICLR 2027 · in preparation