CV
Education, research experience, manuscripts, projects, teaching, and service. A PDF version is available via the download button.
Contact Information
| Name | Sourav Selvaraj |
| Professional Title | M.S. Robotics Engineering Student |
| sselvaraj@wpi.edu |
Professional Summary
Robotics researcher working on contact-rich control and learning for legged robots — MPC/WBC, force-controlled loco-manipulation, humanoid locomotion, and sim-to-real validation.
Experience
-
2025 - present Worcester, MA
Graduate Researcher — ReSCoT-Q, Reference-Guided Surface-Frame Control
ALMaS Research Group, Worcester Polytechnic Institute
Advisor: Dr. Mahdi Agheli
-
2025 - 2026 Worcester, MA
Graduate Researcher — GaitNet, Learning-Based Acyclic Footstep Planning
ALMaS Research Group, Worcester Polytechnic Institute
Advisor: Dr. Mahdi Agheli
-
2024 - 2025 Tiruchirappalli, India
-
2023 - 2023 Chennai, India
Research Intern — Bio-mimetic Autonomous Underwater Vehicle (BAUV)
IIT Madras
Advisor: Dr. Asokan Thondiyath
Education
-
2025 - 2027 Worcester, MA
-
2021 - 2025 Tiruchirappalli, India
Bachelor of Technology
National Institute of Technology, Tiruchirappalli
Instrumentation and Control Engineering
Publications
-
2026 ReSCoT-Q: Reference-Guided Surface-Frame Control for Quadruped Robots
Manuscript in preparation
S. Selvaraj and M. Agheli. First-author manuscript in preparation, targeting IEEE Robotics and Automation Letters (RA-L).
-
2026 GaitNet: Learning-Based Acyclic Footstep Planning for Dynamic Quadruped Locomotion
Manuscript in preparation
O. Sullivan, S. Selvaraj, and M. Agheli. Manuscript in preparation, targeting IEEE ICRA 2027.
Projects
-
Emergent Ice-Skating Locomotion for Humanoids via Blade-Contact RL (Jul 2026 – present)
Unitree G1 trained to ice-skate on passive knife-edge blades with PPO in MuJoCo — no motion capture or imitation data; skating emerges from anisotropic blade–ice contact.
-
Human-Like Bimanual Lifting for Humanoids via Residual AMP–PPO (May 2026 – Jul 2026)
Unitree G1 trained to squat, grasp, and carry objects bimanually via a residual policy over a frozen locomotion actor, with an AMP discriminator as a capped style regularizer.
-
Conditional Flow Matching Policies for Robot Manipulation (Apr 2026 – May 2026)
Conditional Flow Matching visuomotor policy built from scratch in robomimic — 11× faster inference than Diffusion Policy at matched or better task success.
-
Acrobatic Quadrotor Flight using Geometric Control and L1 Adaptive Augmentation (Oct 2025 – Dec 2025)
SE(3) geometric controller with L1 adaptive augmentation on a Crazyflie 2.1 in ROS 2 simulation — 2.2 cm tracking RMSE, up to 94% disturbance rejection.
-
Autonomous Rover — ISRO Robotics Challenge 2024 (Dec 2022 – Aug 2024)
Led a 10-member team building a 6-wheeled rocker-bogie autonomous rover with ORB-SLAM terrain mapping, kinodynamic path planning, and custom control electronics.
Teaching Experience
- Teaching Assistant — Legged Robotics (Graduate), Worcester Polytechnic Institute, Summer 2026. Instructor: Dr. Mahdi Agheli.
- Laboratory Curriculum Developer, Robotics Engineering Department, Worcester Polytechnic Institute, 2026 – present.
Academic Service & Leadership
- Reviewer, IEEE Transactions on Automation Science and Engineering (T-ASE), Jan – Feb 2026.
- Treasurer & Team Lead, Robotics and Machine Intelligence Club, NIT Trichy, Aug 2024 – May 2025.