Param Sangani

PhD Candidate, Computer Science · Saint Louis University

Param Sangani

I am a PhD candidate at Saint Louis University advised by Dr. Hadi Ali-Akbarpour in the AIRLab. My research sits at the intersection of physics-based imaging and robot perception: I work on polarimetric imaging for material understanding, and on multimodal sensing pipelines that let robots decide not just what to do, but when it is safe to act.

Recently this has meant two threads of work: quantum–classical hybrid methods for classifying materials from their polarimetric response, and shared-autonomy control for assistive manipulators driven by synchronized EEG, EMG, and eye-tracking signals. I also TA graduate courses in algorithms and software engineering.

News

Publications

  1. NeurIPS 2026 · WorkshopPoster

    Physics-guided Rolling Polarization Estimation (PRoPE)

    Physical World Model Workshop, NeurIPS 2026

    A method that recovers linear polarization from only a few sequential rotating-polarizer measurements over a narrow angular range, cutting the acquisition latency that limits division-of-time capture in dynamic scenes. A physics-guided decomposition refines the angle-independent intensity first, then reconstructs the remaining Stokes components in a better-conditioned closed form before learned refinement. With three measurements at 15° intervals it reaches 29.3 dB DoLP PSNR and 15.3° AoLP error.

    OpenReviewNeurIPS

  2. arXiv 2026First author

    DensePol: Dense-Angle Polarization Dataset for Learning-Based Polarimetric Vision

    Preprint, 2026

    A high-redundancy RGB–polarization dataset built on division-of-time acquisition, capturing 180 full-resolution analyzer orientations at 1° intervals rather than the four interleaved orientations of a division-of-focal-plane sensor. The denser angular sampling avoids the interpolation and instantaneous-field-of-view errors that limit existing supervision. 2,018 paired RGB–polarization images, released with the per-angle measurements and fitting residuals retained.

    arXiv

  3. IROS 2026Accepted

    NeuroCommitSSM: Decision-Centric Shared Autonomy for Safe Assistive Manipulation via EEG–EMG–ET Commit Readiness

    IEEE/RSJ International Conference on Intelligent Robots and Systems

    A framework that predicts a continuous commit-readiness score from synchronized EEG, EMG, and eye-tracking, and gates robot execution through a HOLD–ASSIST–COMMIT supervisor with real-time feasibility checks. Validated on a Kinova Gen3 arm across five activities of daily living with 32 subjects.

    arXiv

  4. arXiv 2026Under review

    Quantum-Enhanced Similarity Measures for Polarimetric Materials Classification

    Preprint, 2026

    A quantum–classical hybrid pipeline that classifies materials from polarized light reflections by encoding learned voxel embeddings as quantum amplitude states and comparing them with a SWAP-test fidelity measure, benchmarked against optimal transport and Hungarian matching on 23 material classes.

    arXiv

Experience