Drone Over the Mountains

Shervin Ghasemlou

Machine Learning Scientist

I am a graduate research assistant in the Department of Computer Science and Engineering at University of South Carolina, where I am a member of the SCARR lab. My main research interests include: Machine Learning, SLAM, Machine Vision, Automatic Robot Design, and Algorithms. 

You can find my C.V. here.


Goals and Interests

  • The broad goal of my research is to design and implement highly scalable algorithms that enable complex systems, particularly the ones that utilize machine learning and machine vision, to operate autonomously, robustly, and inexpensively.

  • My main interests include, but are not limited to Machine/Robot Learning(Deep Learning, interpretable ML models), Robot Design(Minimality in design), Machine/Robot Vision(visual based SLAM), Planning(Path/Motion Planning), and Algorithms in general.



2015 -2019 (expected)

University of South Carolina

Dissertation Title: Algorithmic Robot Design: Label Maps, Procrustean Graphs, and the Boundary
of Non-destructiveness
Supervisor: Jason O’Kane



Thesis Title: Homecoming: a multi-robot exploration method for conjunct environments with a systematic return procedure
Supervisor: Ali Mohades

Amirkabir University of Technology



Project Title: The Problem of Finding Scalar Generators of a Natural Number
Supervisor: Majid Namazi

Urmia University


My current research is focused on Machine Learning, Vision based SLAM and Multi-robot Systems. To See my publications please refer to my Google Scholar page.

Simultaneous Localization and Mapping (SLAM)

Aqua is an under water robot. Here it collects some visual data sets for a project of underwater mapping. See here.

Screenshot from 2019-07-13 23_35_02_edit

Machine Learning based Robot Design

Should your robot abort its hundreds of million of dollars mission to Mars because one of it's sensors or actuators has gone bad? See this paper.

Multi-Robot Exploration

Sometimes the team of robots exploring an environment need to get back to the base station, for a variety of reasons, including charging and delivery. This should be done preferrably with an equilibrium in task allocation(see this paper).

Screenshot from 2019-07-13 23_44_32_edited_edited.png