Abhiroop Ajith

Research Intern at Siemens · PhD Student at WPI

I am currently a Research Intern at Siemens in Berkeley, California, working on self-supervised video pretraining for robotics. I am also a third-year PhD student in Robotics Engineering at Worcester Polytechnic Institute, advised by Prof. Constantinos Chamzas and part of the ELPIS Lab, where my research focuses on learning world models and symbolic abstractions for long-horizon robot manipulation.

Before my PhD, I worked with Eugen Solowjow's team at Siemens in Berkeley, conducting research on industrial manipulation and manufacturing. Our work involved transitioning technologies from TRL 3 to TRL 7, supported by the ARM Institute via the Department of Defense and NIST. I contributed to projects including tactile manipulation for defect detection of aircraft (with Boeing, CMU, and GelSight) and textile manipulation (with Levi's and Bluewater Defense).

CV last updated: February 18, 2024
Portrait of Abhiroop Ajith

News

Sep 2026
Our paper "STRIPS-WM: Learning Grounded Propositional STRIPS-style World Models from Images" was accepted at CoRL 2026!
Jun 2026
Joined Siemens as a Research Intern, working on self-supervised video pretraining for robotics.
Sep 2025
The preprint of our paper "Learning Discrete Abstractions for Visual Rearrangement Tasks Using Vision-Guided Graph Coloring" is available.
Apr 2025
Received the SIGBED Student Travel Award ($2,000) and the Glenn Yee Travel Award ($450).
Mar 2025
Our paper "Robotic Automation in Apparel Manufacturing" was accepted at CASE 2025.
Jan 2025
Our paper "Multi-layer Motion Planning with Kinodynamic and Spatio-Temporal Constraints" was accepted at HSCC 2025.
Aug 2024
Started my PhD at Worcester Polytechnic Institute.

Research

STRIPS-WM: current and next state observations of a robot workspace encoded by student-teacher encoders
STRIPS-WM: Learning Grounded Propositional STRIPS-style World Models from Images
Abhiroop Ajith, Constantinos Chamzas
Conference on Robot Learning (CoRL), 2026
Learns propositional STRIPS-style world models grounded in images — binary predicates and operators with sparse preconditions and effects — enabling classical planning directly from novel start and goal images.
arXiv
Learning Discrete Abstractions for Visual Rearrangement Tasks Using Vision-Guided Graph Coloring
Abhiroop Ajith, Constantinos Chamzas
Under review
Casts state abstraction as vision-guided graph coloring, learning discrete symbolic states for visual rearrangement tasks.
arXiv
Multi-layer Motion Planning with Kinodynamic and Spatio-Temporal Constraints
Jeel Chatrola*, Abhiroop Ajith*, Kevin Leahy, Constantinos Chamzas (* equal contribution)
ACM Hybrid Systems: Computation and Control (HSCC), 2025
A multi-layer planner that satisfies signal temporal logic specifications under kinodynamic constraints.
arXiv/ Paper/ Code
Robotic Automation in Apparel Manufacturing: A Novel Approach to Fabric Handling and Sewing
Abhiroop Ajith*, Gokul Narayanan*, Jonathan Zornow, Carlos Calle, Auralis Herrero Lugo, Jose Luis Susa Rincon, Chengtao Wen, Eugen Solowjow (* equal contribution)
IEEE International Conference on Automation Science and Engineering (CASE), 2025
A robotic system for fabric handling and automated sewing, developed in collaboration with Levi's, Bluewater Defense, and Sewbo.
arXiv
Robotic Defect Inspection with Visual and Tactile Perception for Large-Scale Components
Arpit Agarwal, Abhiroop Ajith, Chengtao Wen, Veniamin Stryzheus, Brian Miller, Matthew Chen, Micah K. Johnson, Jose Luis Susa Rincon, Justinian Rosca, Wenzhen Yuan
IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2023
Combines visual and tactile perception for robotic defect inspection of large aircraft components, with Boeing, CMU, and GelSight.
arXiv/ Paper/ Dataset

Patents

Robotic visual tactile surface inspection system
WO2024181994A1 · Siemens Corporation · WIPO (PCT) · Published Sep 2024
Chengtao Wen, Justinian Rosca, Eugen Solowjow, Gokul Narayanan Sathya Narayanan, Jose Luis Susa Rincon, Abhiroop Ajith
Multimodal inspection of large-scale surfaces using vehicle-borne sensors (Method)
US20250388339A1 · Siemens Corporation, Boeing Co. · United States · Published Dec 2025
Veniamin V. Stryzheus, Brian T. Miller, Matthew J. Chen, Micah K. Johnson, Abhiroop Ajith, Chengtao Wen, Jose Luis Susa Rincon, Justinian Rosca
Multimodal inspection of large-scale surfaces using vehicle-borne sensors (System)
US20250388337A1 · Siemens Corporation, Boeing Co., Carnegie Mellon University · United States · Published Dec 2025
Veniamin V. Stryzheus, Veniamin Tereshchuk, Arpit Agarwal, Wenzhen Yuan, Micah K. Johnson, Abhiroop Ajith, Chengtao Wen, Justinian Rosca

Service