I am a Ph.D. candidate in the Department of Computer Science at the University of Texas at Austin. I work under the supervision of
Prof. Joydeep Biswas at the
Autonomous Mobile Robotics Laboratory (AMRL).
I am interested in safe navigation for mobile robots via accurate motion models and perception algorithms that are competency-aware.
I received my M.S. in Computer Science from the University of Massachusetts Amherst in spring 2018 and my B.S. in Electrical Engineering
with a major in Control Systems from the University of Tehran in spring 2015.
IV-SLAM: Introspective Vision for Simultaneous Localization and Mapping
Existing solutions to visual simultaneous localization and mapping (V-SLAM) assume that errors in feature extraction and matching are independent and identically distributed (i.i.d). This simplifying assumption makes V-SLAM algorithms prone to catastrophic tracking failures when sensed images include challenging conditions such as specular reflections, lens flare, or shadows of dynamic objects.
We present introspective vision for SLAM (IV-SLAM), a fundamentally different approach for addressing these challenges. IV-SLAM improves the tracking accuracy and robustness of V-SLAM by explicitly modeling the noise process of reprojection errors from visual features to be context-dependent, and hence non-i.i.d. We introduce an autonomously supervised approach for IV-SLAM to collect training data to learn such a context-aware noise model.
IVOA: Introspective Vision for Obstacle Avoidance
Vision, as an inexpensive yet information rich sensor, is commonly used for perception on autonomous mobile robots. However,
vision systems are prone to errors from various sources such as image saturation, blur, and texture-less scenes. In this project,
we develop an approach for self-supervised learning of a model that can predict failures of stereo vision-based obstacle avoidance systems.
The learned model predicts the probability of different types of failure (false positive and false negative) and pinpoints
the location of the error on the input image.
Friction-Based Kinematic Model for Skid-Steer Wheeled Mobile Robots
Skid-steer drive systems are widely used in mobile robot platforms. Such systems are subject to significant slippage and skidding
during normal operation due to their nature. The ability to predict and compensate for such slippages in the forward kinematics of
these types of robots is of great importance and provides the means for accurate control and safe navigation. In this work,
we propose a new kinematic model capable of slip prediction for skid-steer wheeled mobile robots (SSWMRs) leveraging the wheel-ground contact model.
Inverse Kinematics Based Human Mimicking System using Skeletal Tracking Technology
Mimicking is a fast and user-friendly way to teach humanoid robots human-like motions. This project presents a general and efficient,
inverse kinematics based human mimicking system to map human upper limb motions to robot’s joints safely and smoothly. Microsoft Kinect
sensor is used for natural perceiving of human motions.
The Jackal is used for research on campus-scale long-term autonomy at both UMass Amherst an UT Austin. It is equipped with a stereo vision
system and inertial sensors, and an Intel NUC PC for onboard computation. I have been in charge of developing this
robot in both software and hardware aspects and I use it extensively in my research projects. I have written various
algorithms ranging from low level kinematics to obstacle avoidance and path planning to get this robot autonomously navigate on the campus.
The UT AUTOmata race car is a fully autonomous, low power, and portable wheeled mobile robot with Ackerman drive system.
At Autonomous Mobile Robotics Laboratory (AMRL), we use this robot as a platform for multi-robot planning research
as well as for teaching robotics. It is equipped with a Jetson TX2, a Hokuyo UST-10LX Lidar, and an Orbbec Astra depth camera. I have
been in charge of developing and maintaining both the software and hardware for this robot.
At AMRL, we have built a team of soccer robots for the RoboCup Small Size League(SSL). We use
this platform to implement and stress test our research on actual robots and in a competitive environment.
As a member of the UMass Minutebots Team, I have developed software for motion model learning and state
estimation of the robots, and have been in charge of the hardware development.
Nao is a humanoid programmable robot that is vastly used in different areas of robotics
research. I have used this robot as a platform for implementing my research on
learning from demonstration (LfD). I developed algorithms for solving the inverse kinematics
problem for a humanoid robot that enabled the robot to mimick a human's motion as similar to the
instructor as possible while taking into account the constraints imposed by the robot's configuration.
University of Texas at Austin [expected graduation: Dec 2021]
Ph.D in Computer Science - Advisor: Prof. Joydeep Biswas
University of Massachusetts Amherst (2015 - 2018)
M.S. in Computer Science - Advisor: Prof. Joydeep Biswas