Pedram Rabiee

Pedram Rabiee

I am a PhD candidate at the University of Kentucky.

My research interests span the fields of Safe Autonomy, Data-Driven Safety-Critical Control, Model-Based Safe Reinforcement Learning, Motion Planning, Predictive Control, and Learning-Based Control.

I am actively engaged in research on Safe Exploration in Reinforcement Learning, Trajectory Optimization, Control Barrier Functions, Risk-Aware Control for Safety-Critical Systems, and Nonlinear System Analysis and Control. Through my work, I aim to contribute to the development of advanced techniques that ensure the safe operation of autonomous systems in dynamic and uncertain environments.

CV Scholar Github

Publications

A Closed-Form Control for Safety Under Input Constraints Using a Composition of Control Barrier Functions

Pedram Rabiee, Jesse B. Hoagg

IEEE Open Journal of Control Systems, 2024 Under Review

Composing multiple state and input constraints of differing relative degrees into a single control barrier function (CBF) using a composite soft-minimum approach. Achieves safe control with a closed-form solution, eliminating online optimization for exceptional efficiency.

Project Page PDF

Safe Exploration in Reinforcement Learning: Training Backup Control Barrier Functions with Zero Training Time Safety Violations

Pedram Rabiee, Amirsaeid Safari

L4DC, 2024 Under Review

Learning a better backup policy with zero training time safety violations to expand the forward invariant subset of the safe set.

Project Page PDF arXiv Code

Composition of Control Barrier Functions With Differing Relative Degrees for Safety Under Input Constraints

Pedram Rabiee, Jesse B. Hoagg

ACC, 2024 Conference Paper

Composing multiple state and input constraints of differing relative degrees into a single control barrier function (CBF) using a composite soft-minimum approach.

Project Page PDF arXiv

Soft-Minimum and Soft-Maximum Barrier Functions for Safety with Actuation Constraints

Pedram Rabiee, Jesse B. Hoagg

Automatica, 2024 Journal Paper

Exploring in the finite-time backward reachable set of backup sets by introducing two real-time optimization control methods for guaranteed safety within actuator constraints: the first utilizes a softmin barrier function with a single backup control and backup set, while the second, employing an augmented softmax/softmin barrier function, enhances exploration using multiple backup controls.

Project Page PDF arXiv

Soft-Minimum Barrier Functions for Safety-Critical Control Subject to Actuation Constraints

Pedram Rabiee, Jesse B. Hoagg

ACC, 2023 Conference Paper

Exploring in the finite-time backward reachable set of a backup set using a softmin barrier function.

Project Page PDF arXiv

The Impact of Reference-Command Preview on Human-in-the-Loop Control Behavior

Pedram Rabiee, S. Alireza Seyyed Mousavi, Amelia J.S. Sheffler, Erik Hellström, Mrdjan Jankovic, Mario A. Santillo, T. Michael Seigler, Jesse B. Hoagg

IEEE Transactions on Human-Machine Systems, 2023 Under Review

Analyzing the impact of reference command preview on human subjects' performance in a dynamic system command-following task using subsystem identification to model control behavior and revealing insights into the trade-offs associated with preview duration.

Project Page PDF arXiv

Open-Source Projects

Safe RL

A unified framework integrating model-free, model-based, and data-based methods for safe reinforcement learning.

Code

Higher-Order Barrier Functions Composition Library

This library provides a set of classes and utilities for higher-order barrier functions (HOCBFs) compositions.

Code

Ground Robot Obstacle Reach Safe Environment

Gymnasium-style safe navigation environment for ground robots, featuring a composite barrier function derived from map and velocity constraints, ensuring robot safety during navigation with flexibility in dynamics and layouts.

Code

Human vs Agent Interaction Game in Unity

Human is interacting in a game with autonomous agents all trying to go to their designated goals.

Code