Robotics Expertise
Comprehensive Robotics Foundation
My expertise in robotics is built upon numerous academic courses and hands-on application with manipulators and humanoid robots. From core principles to joint implementation with AI.
Especially the joint integration of robotics with AI is what really excites me. This is also why I chose to write my master thesis and my first published paper in this direction.
Explore my work on Variable Stiffness for Robust Locomotion through Reinforcement Learning published at IFAC.
Robotics Domains
Locomotion
Manipulation
AI Control
ROS
Core Skills & Services
With a multidisciplinary background in electrical engineering, control systems, and artificial intelligence, I bring a full-stack approach to modern robotics. My work spans academic research, hands-on prototyping, and intelligent system design.
- Robotics Systems Development: Full-cycle development of robotic platforms (locomotion & manipulation), ROS (Robot Operating System) expertise (Kinetic & Noetic).
- AI-Driven Control & Learning: Implementation of Supervised/unsupervised techniques and Reinforcement learning for adaptive robot behavior.
- Cross-Disciplinary Thinking: My strong background in electrical engineering and mechanical engineering provides me with a solid foundation. I am comfortable moving between code, circuits, and real-world deployment.
Motivation & Experience
My journey into robotics has been driven by curiosity and a hands-on mindset. With a solid academic foundation in kinematics, dynamics, interaction control, and the linear algebra that supports them, I've been able to turn complex theory into practical, working systems. I've had the opportunity to implement a variable stiffness locomotion policy on the GO2 humanoid robot, and to assemble and integrate control systems for a 6DOF robotic arm.
These experiences have shaped my approach to robotics—not just as a technical field, but as a space where creative problem-solving and real-world application intersect. I really enjoy this field since multiple disciplines (like electrical, mechanical and software engineering) come together to create something that is more than the sum of its parts. Since I have background in electrical engineering, I deeply value the hardware side of robotics. With my experience in both manipulation, locomotion, applied AI and a cross-functional understanding I am able to see systems as a whole and design solutions that are both innovative and grounded.
AI & Robotics Integration
Broadly speaking, there are two domains in robotics:
- Classical control domain: which heavily relies on mathematical modeling and control theory and optimization based techniques. In which I have a solid foundation.
- AI domain: which is more data driven and relies on machine learning techniques.
The AI side of robotics is evolving quickly, opening up exciting possibilities—from autonomous vehicles to smart assistants and beyond. My work has focused on the integration of these two areas, combining control theory with the flexibility of AI. It's a space I'm deeply passionate about, and one where I bring both solid experience and an eagerness to explore emerging frontiers. My previous work includes manipulation as well as locomotion.
Project on Locomotion
Within my master thesis I worked on the integration of a variable stiffness locomotion policy on the GO2 humanoid robot. This work was published at IFAC and is available for download here. This approach applies reinforcement learning to train an AI agent to walk. Additionally to the previous approaches which learn to predict the joint positions, we trained the agent to learn the joint stiffnesses. Normally, the stiffness of the joints is tuned manually, which can be time consuming and not optimal. The approach aims to automate the adjustment of motor stiffness according to task requirements, eliminating the need for manual tuning.
Key Results
- Velocity tracking: Our method enabled more accurate velocity tracking than position based approaches (baselines).
- Push recovery: Our method outperformed the baselines showing robust learned gait.
- Energy efficiency: While showing competitive results with low stiffness policies it outperforms high stiffness policies.
Project in Manipulation
During my time at the Institute ICT at TU Vienna, I contributed to the development of a 6DOF robotic arm, working under the guidance of Shail Jadav. Building on the open manipulator platform, we extended its capabilities by adding two additional degrees of freedom, greatly increasing its flexibility and range of motion. The arm was equipped with a camera and gripper, enabling it to perform a wide variety of manipulation tasks.
What made this project unique was the AI-driven control approach. Instead of relying on traditional control theory, we explored learning-based methods to achieve autonomous manipulation. This opened the door to more adaptive and intelligent robot behavior.
Goal
The primary objective was to solve the Tower of Hanoi puzzle autonomously, combining motion learning, perception, and reasoning through AI.
Achieved Goals
- Kinesthetic Teaching: Physically guiding the robot to record pick and place movements. This trajectories should be storable and replayable.
- Learning from demonstrations: One recorded trajectory might solve one problem. But what if I want to pick an object from a different location? This is where Dynamic Movement Primitives comes in. We use DMPs to generalise the recorded trajectories to pick/place from to any location.
- Object detection and recognition: Using the camera the objects must be detected, segmented and locate the object in the world frame, so the robot is able to grasp it.
- Reasoning with LLMs: Once the basic motions and the location is determined all is left is the reasoning to solve the tower of hanoi problem. An LLM, based on the outputs of the other modules and the present state, is used to determine the next steps.
The strength of this project lies in its interdisciplinary approach, bringing together robotic manipulation, computer vision, and AI reasoning. Despite its cost-effective design, the platform closely resembles advanced robots such as the KUKA LBR iiwa, making the developed concepts highly transferable. Built entirely on the ROS middleware, the framework can be applied to a wide range of robotic systems.
The results demonstrate the potential of learning-based methods for controlling robotic arms — a field still in its early stages but already showing strong promise for the future of autonomous robotics. This work serves as a foundation for ongoing research and further development.