Kumar Manas - Postdoctoral Researcher (Robotic Manipulation and AI)

News & Announcements

New 2026

PhD Thesis Successfully Defended!

I successfully defended my PhD thesis, "Automated Rule Formalization and Layered Knowledge Integration for Safety-Aware Autonomous Systems," at Freie Universität Berlin.

About Me

I completed my PhD in a joint program between Freie Universität Berlin and Continental Automotive, supervised by Prof. Adrian Paschke, with a thesis on automated rule formalization and layered knowledge integration for safety-aware autonomous systems. In November 2025, I joined the PEARL Lab (Interactive Robot Perception & Learning) at TU Darmstadt as a Postdoctoral Researcher, supervised by Prof. Georgia Chalvatzaki.

My current research focuses on robotic manipulation and how foundation models can be leveraged for generalizable robot learning, enabling agents to reason, act, and adapt across diverse manipulation tasks. In particular, I work on long-horizon manipulation and planning, and on memory design for Vision-Language-Action (VLA) models and world models — combining structured reasoning, perception, and control for safe and flexible robot behavior.

Current Research

Memory Design for VLA Models

Developing memory mechanisms for Vision-Language-Action (VLA) models that let robots retain and reuse task-relevant context across long-horizon manipulation episodes.

World Models for Long-Horizon Manipulation

Building world models that support planning and decision-making in long-horizon robotic manipulation tasks, bridging perception, memory, and action.

Foundation Models for Generalizable Robot Learning

Leveraging foundation models to enable robots to reason, act, and adapt across diverse manipulation tasks, at the PEARL Lab, TU Darmstadt.

Prior Research (PhD)

Trajectory Prediction with Rules and Uncertainty

Integrating traffic rules into the trajectory prediction module in an implicit manner, alongside improving uncertainty quantification in the prediction.