Robotics | Validation, Simulation & Control Systems
Robotics engineer with a background in automotive engineering, specialized in control systems, simulation and validation. From modelling to control system design to structured validation against real hardware. I build autonomous systems end to end: ROS, localization, motion planning. Recently, I finished the IFROS Erasmus Mundus Master.
February 2026 — August 2026
Delivered a classical geometric RGB-D perception pipeline for autonomous pallet pickup by industrial mobile robots, recovering the pallet pose from depth-sensor processing, and integrated it as a feature of Duna, the company's robot software product (see the demo).
2021 — 2023
Developed a tilt-compensation control system for articulated boom lifts: multi-domain modelling in Simcenter AMESim, co-simulation with Simcenter 3D, PID control design, and FMI export to Unity for VR validation. Ran physical stability test campaigns on real machines to derive acceptance criteria, and contributed to FMEA on the new architecture.
P01 · MSc Thesis · RGB-D Perception
A purely geometric, training-free perception pipeline that takes a single RGB-D point cloud from the robot's onboard Intel RealSense D435 and outputs the pallet's pose at the front-face centre in robot base frame, ready for docking and pickup. Implemented as a PCL-based ROS node with a ROS-agnostic C++ core, and evaluated in Gazebo simulation.
MSc thesis with Beta Robots. The pallet detector is my own work, it lives in the company's private codebase, so the source code isn't public.
Operating-envelope sweep: the robot was placed across a grid of positions and yaws (±20°) in front of the pallet, and each detection was compared with ground truth to find where the detector can be trusted.
An autonomous car park surveillance application on the AgileX Scout 2.0: the robot navigates through a car park, dynamically avoids obstacles, detects vehicles and their license plates with the onboard camera, and places them on a map of the environment. Runs on ROS Melodic on an NVIDIA Jetson Xavier with a Velodyne LiDAR, Intel RealSense RGB-D camera and Xsens IMU.
Team project (4 people, IFROS Lab, ELTE) — my focus: mapping. I built the 3D mapping module (depth back-projection, data association, OCR consensus) that places detected license plates on the OctoMap occupancy grid.
Feature-based EKF SLAM on a differential-drive TurtleBot fusing wheel odometry, IMU heading and ArUco landmark observations. Tested in simulation and on the real robot.
Pair project. My half: the correction step, measurement models and data association (Mahalanobis + chi-square gating).
Task-priority kinematic control for a 4-DOF manipulator on a mobile base, resolving multiple prioritized tasks through redundancy for pick-and-place operations, sequenced by a behavior tree.
P05 · Stäubli TX60 · Industrial Arm
Programmed a Stäubli TX60 robot for pick-and-place part assembly, focusing on repeatable manipulation sequences and dependable handling in a constrained workspace.
P06 · Stäubli TS60 · SCARA
Developed a SCARA-based sorting workflow to classify metallic and plastic cylinders into target locations using precise pick-and-place routines.
P07 · TurtleBot · Navigation
Online navigation combining an RRT global planner with a reactive controller and LiDAR obstacle avoidance on an occupancy map. Built in ROS, tested in Stonefish and on the real TurtleBot.
A from-scratch CNN for joint classification and bounding-box regression on a synthetic dataset, benchmarked against a fine-tuned YOLO baseline.
P09 · TurtleBot · Coverage
Implementation of an online surface coverage algorithm using a BackTracking algorithm to visit all reachable areas with clearance distance to walls.
Feel free to reach out anytime
