Control Barrier Functions for Adaptive Stabilization of Fully Unknown Underactuated Nonlinear Systems 2027
IEEE Conference on Decision and Control (CDC, under review)
Who Am I?
Hi I'm Kasra Sinaei I am pursuing my graduate studies in the field of Control, Robotics and Physical AI at the Pennsylvania State Univrsity. I am conducting research on Safety Control via Barrier Functions and I develop methods to integrate Safety with model-free learning algorithms and emerging Vision Language Action Models (VLA). My undergrad research was focused on Legged Locomotion and humanoid gait control.
I have studied and researched Control Theory, Optimization, Non-linear Systems, Physical AI and Robotics. I am a proficient Robotic software developer and I have advanced programming skills in CPP, Python, MATLAB and C#.
Control Theory
Robotics
Physical AI
Learning Controllers
Peer Reviewed Publications
Conference presentations
Paper Citations
Internship & Co-op
Papers
Strategic and Reactive Stream Partitioning for Path-Efficient LiDAR-Based Obstacle Avoidance 2026
IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)
Safe Vision Language Action Models Via Barrier Enhanced Flow Matching 2026
IEEE Robotics and Automation Letters (RA-L, under review)
Safe Adaptive Control with Vanishing Conservativeness for Robotic Systems with Unknown Dynamics Via Barrier Functions 2026
IEEE Robotics and Automation Letters (RA-L)
Motion Planning for Mobile Manipulators Navigating Doorways via Model Predictive Path Integral 2026
Modeling Estimation and Control Conference (MECC), ASME
Safe Robust Control of Nonlinear Systems with Uncertain Regressor and Parameter Vector 2025
9th IEEE Conference on Control Technology and Applications (CCTA)
Safety-Critical Position Control of Robots: A Model-Free Approach 2025
American Control Conference, IEEE (ACC)
Online Bipedal Locomotion Adaptation for Stepping on Obstacles Using a Novel Foot Sensor 2022
IEEE-RAS 21st International Conference on Humanoid Robots (Humanoids)
Bipedal Locomotion Optimization by Exploitation of the Full Dynamics in DCM Trajectory Planning 2021
9th RSI International Conference on Robotics and Mechatronics (ICRoM)
PID Controller Tuning with Deep Reinforcement Learning Policy Gradient Methods 2021
29th Intermational Conference of Iranian Society of Mechanical Engineers
Education
Expecting graduation on Spring 2027.
Current GPA: 3.9
Advisor: Dr. Donald E. Ebeigbe
Graduated on May 2024.
Current GPA: 3.84
Advisor: Dr. Donald E. Ebeigbe
Graduated on February of 2022.
Cumulitive GPA: 17/20
Advisor: Dr. Ehsan Hosseinian
Work Experience
Chewy - AI Research Scientist Intern 2026
I am interning at Chewy Robotics Lab, developing semantic safety filters for mobile manipulators powered by physical AI policies. I leverage 3D scene graphs to retrieve semantic information from the environment and deploy Vision-Language-Action models for long-horizon warehouse tasks.
Alphaz - Robotic Software Engineer Intern 2025
I joined Alphaz as an intern in 2025. I have been Working on whole-body-control and planning of a humanoid robot with two 6-DoF arm and a differential-drive mobile base. I also worked on the robot perception and object detection. A great opportunity to build robotic softwares with a combination of modern learning algorithms and classic control theories.
PSU Graduate Researcher 2022-Now
Graduate researcher at CARL (Control and Robotics Lab) located in the Electrical Engineering department of Penn State University. Main focus of our work is towards developing smart controllers and finding formal proof for safety and stability in different settings. We use Unitree Go1 robots and Quanser Q-Arm robot for our experiments.
CAST Undergrad Research Assistant 2019-2022
I was memeber of Dynamic and Control group at Center of Advanced Systems and Technology (CAST). The group was working on humanoid robot gait. Our team successfully designed and fabricated 4 (soon to be 5) full-size humanoid robots. I was working on gait pattern generation and enhancing dnamic walking of the robot. Check out Surena Humanoid website for latest updates on this project.
MEWE Summer Intern 2018, 2019
I joined an Engineering consultant group based in Tehran for two consecutive summers. I learned basics of engineering, drafting, mechanical design and project management.
Open Source Projects
Manipulator Teleoperation with Smartphone
Emerging Foundation Models and Generative AI policies in robotics demand efficient trajectory collection strategies. This smartphone app is powered by Google ARCore built in VIO and utilizes Webrtc and fast UDP communication to help broadcasting expert input to the robot controller. githubSafe Flow Matching for Vision Language Action Models
Combining the power of VLAs to deliver complex robotic tasks with the robustness and formal safety proofs of classic control theory unleash a new power of large behavioral models. In this research I developed a novel modifiaction of the ocnventional Flow Matching to enforce safety for SmolVLA and Pi-0. Check this blog for more info.Safe Parameter Learning for Adaptive Controllers
Main objective of my PhD research is to integrate safety guarantees via classical control theories into the state of the art learning policies and controllers. In this project I addressed one of the biggest challenges of deplying adaptive controllers. Before the controller paramter converges to the nominal values it is very difficualt to maintain safety of the system. So I developed a Robust Adaptive controller that manipulates weight update to achieve safety during the learning phase. For more details visit this link.Door Opening with Mobile manipulators
During my internship at AlphaZ Robotics, I worked on a project involving mobile manipulators and door opening tasks. The goal was to develop a robust planning and control strategy that allows the robot to traverse various doors in different environments. We developed an MPPI planner to tackle this problem, for more details visit this link.Safety-critical Control
In this research I am studying Control Barrier Functions (CBF) and their applicatins in Robotics. I am trying to extend the CBF capabilities to better handle extreme scenarios that include leaving the safe set and retrieving safety. I have also worked on model-free safety-critical control of complex kinematic/kinetic systems. Check this blog for more details.Safe Robust Regressor-based Control
My PhD studies focus on developing safety filter for different type of controllers. Regressor-based controllers are an effective approach for ensuring safety in control systems. In this work I added safety layer to a well-known class of robust controllers used and performed an extensive analysis over their performance. Check results of my research in this website.Race Line Tracking and Autonomous Racing: MPC Vs. Adaptive Control
I am modeling and simulating a race car on a 2D race track for the course project of ME597:MPC and EE582:Adaptive control. This is an ongoing work. See the report on this repo.Surena Humanoid Robot
As a member of Dynamics and Control team at CAST I was working on gait pattern generation for a full full-size humanoid robot (SURENA V) using Divergent Component of Motion (DCM). We also worked on improving robot's gait stability and robustness with impedance controllers to enable the robot better traverse rough terrains. More InfoKalman Filterning and System Identification
This project was part of AERSP597 course. I studied and campared different methods for fusing sensors data to estimate position of a differnetial drive robot. Linear KF, Extended KF and Unscented KF were studied in this project. Project ReportWheeled Biped Robot
This is my undergrad thesis project. I examined dynamics of wheeled legged robots, optimally designed a simulated robot and implemented a novel control startegy that does not deped on robot's full dynamic equations. I also examined robot's jumping in the simulation using the proposed controller. More InfoPattern Recognition
During the Pattern recognition and Machine Learning course we studied music genre classification using the methods we learned in the class. First we gathered an adequate data set consisting of persian songs and labeled them, then we implemented several supervised and unsupervised learning algorithms such as SVM, KNN, MLP, K-means as well as dimension reduction technics. Project RepoReinforcement Learning and Classic Control
In this project I used Policy Gradient Methods for PID controller self tuning in physics engine simulation. I trained a PPO agent and Actor-critic agent to automatically adjust controller gains. Results of this work are presented in a conference that includes simulation of a simple inverted pendulum on wheel.I am a space geek!
I have been passionate about astronomy and space since I was a child. My father bought me a small refracting telescope when I was just 9 years old. I used my Apochromatic telescope to stare at the gas giants of our solar system and the craters on the moon. My interest in amateur astronomy stems from this childhood experience. As I am learning imaging and editing technics I share some of my astrophotography results in this page.
Contact
Electrical Engineering West, 013-C4,
State College, PA16802