Abdelrahman Abdelgawad

Abdelrahman Abdelgawad

PhD Student at Boston University working on certified neural control and multi-objective reinforcement learning for safety-critical robotic systems.

Posts by Collection

portfolio

publications

A Study on Broadcast Networks for Music Genre Classification

Published in IEEE International Joint Conference on Neural Networks (IJCNN), 2022

Investigated broadcast network architectures for music genre classification tasks.

Recommended citation: A. Heakl, A. Abdelgawad, V. Parque. "A Study on Broadcast Networks for Music Genre Classification." IEEE International Joint Conference on Neural Networks (IJCNN), 2022.
Download Paper

Model- and Data-Based Control of Self-Balancing Robots

Published in IFAC International Conference on Programmable Devices and Embedded Systems, 2024

Compared model-based (PID) and data-based (Fuzzy) controllers for a self-balancing robot with hardware validation.

Recommended citation: A. Abdelgawad, T. Shohdy, A. Nada. "Model- and Data-Based Control of Self-Balancing Robots." IFAC International Conference on Programmable Devices and Embedded Systems, 2024.
Download Paper

Closing the Intent-to-Behavior Gap via Fulfillment Priority Logic

Published in IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2025

We introduce Fulfillment Priority Logic (FPL), a logic transforming declarative policy behavior descriptions into utility functions through principled algebraic transformations. Our Balanced Policy Gradient algorithm leverages FPL specifications to achieve up to 500% better sample efficiency compared to Soft Actor-Critic.

Recommended citation: B. El Mabsout*, A. Abdelgawad*, R. Mancuso. "Closing the Intent-to-Behavior Gap via Fulfillment Priority Logic." IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2025. (*equal contribution)
Download Paper

Accelerating Lyapunov-Stable Neural Control using Fulfillment Priority Logic

Published in American Control Conference (ACC), 2026

We present a two-stage approach for learning stability-certified neural controllers that achieves up to ~95% reduction in training time compared to existing methods. Our method combines monotonic neural Lyapunov functions with Fulfillment Priority Logic (FPL) to efficiently initialize controllers before formal MILP-based verification.

Recommended citation: A. Abdelgawad, B. El Mabsout, Z. Wang, R. Mancuso, S. B. Andersson, R. Tron. "Accelerating Lyapunov-Stable Neural Control using Fulfillment Priority Logic." American Control Conference (ACC), 2026.

talks

teaching

RISE Mentor

Mentoring, Boston University, Research in Science & Engineering (RISE), 2025

Role: Research Mentor, Summer 2025