Hello World!
I'm a third-year Ph.D. student in Computer Science at Arizona State University, where I do research at AAIR Lab under Prof. Siddharth Srivastava. My research is centered around three exciting areas: AI and robotics, safe autonomy, and human-robot interaction.
Previously, I received my D.Eng. in Electrical Engineering from Lamar University, studying under the guidance of Prof. Hassan Zargarzadeh. My research focused on task and motion planning of heterogenous multi-agent systems, wherein I introduced the innovative concept of the hunter-and-gatherer approach. During the same period, I completed my M.Sc. in Computer Science, also at Lamar University, supervised by Prof. Peggy Isreal Doerschuk. My research in this field revolved around advancing the experience replay technique in reinforcement learning.
I received my B.Sc. in Electrical Engineering from Qazvin Azad University in 2017. During my undergraduate studies, I was a research assistant at AMRL, where I had the privilege of leading a diverse team of students on an inspiring journey, devoted to the development of tele-operative and autonomous rescue robots.
My Spotlight Projects
RL algorithms don't scale in sparse reward settings. We addressed this by proposing CAT+RL method that learns CATs automatically and on-the-fly while an RL agent interacts with the world. In our empirical evaluations, CAT+RL improved the sample efficiency to the point where vanilla Q-learning can outperform SOTA RL methods.
We explored the situations in which a basic social navigation scenario can get complicated and unsafe due to the lack of proper communication. To address this issue, we invented a new cobot planning framework that jointly computes the robot's motion plans and communication signals, and safely and efficiently enriches human-robot interactions.
Task and motion planning for multiple robots or agents with heterogeneous capabilities is a fundamental and yet significantly challenging problem in the robot planning context. We addressed this problem by introducing the game-theoretic Hunter-and-Gatherer approach.
Maxi-sized Response Robots
Natural and human-made disasters are inherently inevitable, but an effective response to a disaster is tremendously essential to control the damages and save lives. Motivated by this, we designed and prototyped autonomous and tele-operative rescue robots that carry out reconnaissance and dexterity operations in unknown environments comprising unstructured obstacles.
Entropy-Based Experience replay in RL
We explored the situations in which a basic social navigation scenario can get complicated and unsafe due to the lack of proper communication. To address this issue, we invented a new cobot planning framework that jointly computes the robot's motion plans and communication signals, and safely and efficiently enriches human-robot interactions.