[{"content":"This work integrates visuo-tactile sensing with haptic feedback for teleoperated robot manipulation. Both arms of a TIAGo humanoid robot are controlled through an HTC Vive VR setup combined with MANUS haptic feedback gloves; GelSight visuo-tactile sensors on the gripper measure contact, and the measured contact state is rendered back to the operator\u0026rsquo;s fingertips in real time as vibration signals.\nContact forces are estimated from the sensor images with two complementary methods: tracking the movement of markers embedded in the sensor surface, and a deep-learning estimator. On the feedback side, a ROS node streams per-finger force values to a modified MANUS SDK client driving the gloves' vibrotactile actuators. The work also introduces a novel setup for evaluating normal force, shear force, and slip (arXiv:2404.19585).\nI presented this work as the primary author at the IEEE ICRA@40 conference in Rotterdam (2024).\nTechnologies: computer vision, haptic motor control, virtual reality, teleoperation, ROS, Python\n","permalink":"https://noahbecker.tech/projects/icra40-haptic-teleoperation/","summary":"Teleoperated TIAGo humanoid with Vive VR, haptic gloves, and visuo-tactile sensors, presented as primary author at ICRA@40, Rotterdam.","title":"ICRA@40 — Visuo-Tactile Sensing with Haptic Feedback for Teleoperated Manipulation"},{"content":"My current research at the Intelligent Autonomous Systems group (TU Darmstadt) investigates how robots can feel objects slipping out of their grasp. I built a custom teststand that induces controlled slip with a linear actuator while recording four sensing modalities simultaneously: a GelSight Mini visuo-tactile camera, a six-axis force/torque sensor (ReSense HEX21), an ADXL355 accelerometer, and webcam-based ArUco marker tracking.\nOn the software side, the system runs on ROS2 Jazzy across a Raspberry Pi 5 (hardware control) and a laptop (high-bandwidth sensor processing). A branching neural network fuses CNN-encoded tactile images with force/torque and IMU time-series through an LSTM to regress slip velocity, the basis for slip-aware grip force control in teleoperation.\nTechnologies: Python, ROS2, PyTorch, GelSight, force/torque sensing, CAD, Docker\n","permalink":"https://noahbecker.tech/projects/visuo-tactile-slip-detection/","summary":"Master\u0026rsquo;s research at TU Darmstadt: a multi-sensor teststand and neural networks for detecting and regressing slip in robotic grippers.","title":"Visuo-Tactile Slip Detection for Teleoperated Manipulation"},{"content":"An earlier version of the haptic teleoperation work, accepted to the ViTac workshop (Robot Embodiment through Visuo-Tactile Perception) at ICRA 2024 in Yokohama, Japan (arXiv:2404.19585). Feedback in telerobotics is usually purely visual, which limits the operator\u0026rsquo;s dexterity; this work restores the missing sense of touch. GelSight visuo-tactile sensors on the robot\u0026rsquo;s gripper measure contact, and low-cost MANUS gloves render the estimated contact forces to the operator\u0026rsquo;s fingertips as vibration signals inside a VR teleoperation pipeline.\nThe goal of the project is a teleoperated solution aimed at improving the quality of life for immobile individuals by enabling independent remote task completion, for example shopping, through a robot avatar with tactile feedback.\nTechnologies: computer vision, haptics, virtual reality, teleoperation, ROS, Python\n","permalink":"https://noahbecker.tech/projects/vitac-teleoperation/","summary":"Earlier iteration of the haptic teleoperation system, accepted to the ViTac workshop at ICRA 2024, Yokohama.","title":"ICRA ViTac Workshop — Visuo-Tactile Sensing with Haptic Feedback"},{"content":"My bachelor thesis at FAU Erlangen-Nürnberg (\u0026ldquo;Optimizing Robotic Bin-Picking of Radioactive Debris using Supervised and Imitation Learning\u0026rdquo;) developed a neural network-based part selection system to improve robotic bin-picking for the semi-autonomous handling of radioactive waste.\nTraining data was generated through VR teleoperation and simulation in Unity: human operators demonstrated grasping motions in a VR rig (VIRERO), and the demonstrations were used for supervised and imitation learning (behavioral cloning, GAIL) alongside RL policies (PPO/SAC) built on Unity ML-Agents. The final system achieved a grasp accuracy of 73% (a 5 percentage-point improvement over the previous system) while adding only 1.2 ms of latency.\nThe underlying Unity simulation models the robot cell with URDF robot models and ROS integration (ROS#); trained policies were exported as ONNX models, with trainer configurations covering task variants from full bin-picking with the robot in the loop down to isolated move-to-grasp-pose subtasks.\nTechnologies: PyTorch, Vive VR, ROS, YOLO, Unity (C#), ML-Agents\n","permalink":"https://noahbecker.tech/projects/bachelor-thesis-bin-picking/","summary":"Neural network-based part selection for semi-autonomous bin-picking of radioactive waste, trained on VR teleoperation data. 73% grasp accuracy at +1.2 ms latency.","title":"Bachelor Thesis — Robotic Bin-Picking via Supervised and Imitation Learning"},{"content":"Designed and implemented a rendering pipeline featuring fragment-shader ray tracing, GLSL shading, view-frustum culling, screen mapping, and projection, along with a custom scene graph and animation system. The work spans four progressive C++/Qt applications: from basic triangle-mesh rendering through a shader-based solar system, skyboxes with normal/bump/displacement mapping and Perlin-noise terrain, up to shadow mapping and ray tracing.\nTechnologies: C++, OpenGL 3.3, GLSL, Qt, CMake\n","permalink":"https://noahbecker.tech/projects/opengl-rendering-pipeline/","summary":"Rendering pipeline with ray tracing, shading, culling, and a custom scene tree, built across four progressive Qt/OpenGL applications.","title":"Custom Rendering Pipeline in OpenGL"},{"content":"A framework based on AlphaZero capable of learning to play single- and two-player games (tic-tac-toe, Connect 4, and CartPole) using a unified model architecture, with self-play training and Monte-Carlo tree search. Both ResNet and MLP network backbones were explored and compared. Built as a team project for the course Deep Learning: Architectures \u0026amp; Methods at TU Darmstadt.\nThe cover shows a self-play game between two copies of the trained agent (600 MCTS simulations per move, ending in a hard-fought draw) and the search policy at the game\u0026rsquo;s most contested decision.\nTechnologies: PyTorch, AlphaZero, self-play RL\n","permalink":"https://noahbecker.tech/projects/alphazero-rl-agent/","summary":"AlphaZero-based framework that learns tic-tac-toe, Connect 4, and CartPole with a unified model architecture.","title":"Reinforcement Learning Agent in Games (AlphaZero)"},{"content":"Programmed a coffee machine using LabVIEW, covering sensor integration, actuator control, and robust software for touch-screen operation. Because hardware access was limited, I additionally developed a Python simulation of the machine so software development could continue off-device.\nTechnologies: LabVIEW, electrical engineering, microcontroller programming, Python\n","permalink":"https://noahbecker.tech/projects/labview-coffee-machine/","summary":"Sensor integration, actuator control, and a touch-screen UI for an automated coffee machine, plus a Python simulation of the hardware.","title":"LabVIEW Coffee Machine Automation"},{"content":"A C++17 project implementing greedy and local search optimization algorithms for the 2D rectangle bin packing problem: minimize the number of equal-sized square bins needed to pack a set of rectangles (rotations allowed). The greedy family covers Bottom-Left, Shelf, and MaxRects placement strategies combined with two rectangle-ordering strategies; the local search family provides geometry, permutation, and overlap neighborhoods with pluggable selectors and scoring heuristics. An interactive SDL2 + Dear ImGui GUI visualizes the algorithms step by step, complemented by a headless CLI for debugging.\nTechnologies: C++17, CMake, SDL2, Dear ImGui\n","permalink":"https://noahbecker.tech/projects/optimization-algorithms/","summary":"Greedy and local-search algorithms for 2D rectangle bin packing, with an interactive SDL2 + ImGui visualization GUI.","title":"Optimization Algorithms — 2D Rectangle Bin Packing"},{"content":"In this research project a pen was mounted to a Franka robot arm through a GelSight visuo-tactile sensor, and the force at the pen tip was estimated purely from the sensor\u0026rsquo;s images. I developed several computer vision approaches to infer the contact force (stereo correspondence, motion flow, and marker-based methods) as well as a neural network-based estimator.\nTechnologies: ROS, computer vision (stereo, motion flow, SIMD), PyTorch\n","permalink":"https://noahbecker.tech/projects/tool-force-estimation-robot-drawing/","summary":"A pen mounted on a Franka arm through a visuo-tactile sensor; computer-vision algorithms estimate the contact force at the pen tip.","title":"Tool Contact Force Estimation for Robot Drawing"},{"content":"An AI-based face detection door opener app, built for the Ambient Intelligence course at TU Darmstadt. The Android app scans faces on-device, encodes them into embeddings, and sends the encoding to a web server which authenticates the person and triggers the door control.\nTechnologies: PyTorch, Flask, Android Studio (Java)\n","permalink":"https://noahbecker.tech/projects/biometric-door-opener/","summary":"AI-based face recognition door opener: on-device face scanning and encoding, with server-side authentication and door control.","title":"Biometric Door Opener — Ambient Intelligence"},{"content":"Implemented a complete autonomy stack on a Turtlebot platform for autonomous object detection and pick-and-place tasks: kinematics, computer vision, path planning, SLAM, and high-level action planning.\nTechnologies: C++, ROS, computer vision\n","permalink":"https://noahbecker.tech/projects/robotics-101-turtlebot/","summary":"Full autonomy stack on a Turtlebot: kinematics, vision, path planning, SLAM, and high-level action planning.","title":"Robotics 101 — Mobile Robot Pick and Place"},{"content":"Using a GelSight Mini visuo-tactile sensor, this project reconstructs a unified 3D surface from multiple overlapping sensor presses. Each capture produces an RGB image and a neural-network-inferred depth map; individual captures are registered with SIFT keypoint matching and fused into a single colored point cloud, exportable as a triangulated mesh. Euro and cent coins served as test objects.\nBeyond the stitching pipeline itself, the project provides tools for live 3D visualization, snapshot capture, and inspecting SIFT correspondences between captures, along with a driver layer for the GelSight device and its depth reconstruction.\nTechnologies: computer vision, 3D point cloud analysis, image stitching, PyTorch\n","permalink":"https://noahbecker.tech/projects/gelsight-3d-scanning/","summary":"3D point clouds and surface stitching from GelSight Mini tactile captures: reconstructing coin surfaces by touch alone.","title":"3D Scanning and Measuring with Visuo-Tactile Sensors"},{"content":"For the Serious Games course I created a one-button game featuring procedurally generated levels, multiplayer, and AI voice-over. The entire game is playable with a single input: timing and rhythm replace complex controls.\nTechnologies: Godot, game design, procedural generation\n","permalink":"https://noahbecker.tech/projects/one-button-game/","summary":"A one-button game with procedurally generated levels, multiplayer, and AI voice-over.","title":"Serious Games — One-Button Game"},{"content":"As a working student at Siemens Energy I developed a bill-of-materials (BOM) web application integrated with Siemens databases, automated user management workflows with VBA, and participated in QA and project planning with an international team. (Internal industry project; no public code or screenshots.)\nTechnologies: JavaScript, HTML, CSS, VBA, Python, Java\n","permalink":"https://noahbecker.tech/projects/siemens-bom-webapp/","summary":"BOM web application integrated with Siemens databases, plus automated user-management workflows.","title":"Siemens — BOM Full-Stack Web Application"},{"content":"Programmed a team of humanoid NAO robots for a soccer challenge, implementing motion control, multi-agent collaboration, and high-level planning. The simulation bridges the NAOqi software-in-the-loop runtime with Gazebo through a custom plugin, enabling joint control and sensor feedback for the NAO inside the simulated world, alongside a soccer-field world model, a custom Kinect sensor model, and AprilTag-based ball localization.\nIn the penalty-kick scenario, a NAO autonomously navigates to the ball (detected via AprilTag tf transforms), positions itself, and executes the kick through the Gazebo physics engine, while a dedicated action-client node dispatches named NAOqi behaviors. Our team achieved first place in the class competition.\nTechnologies: ROS, Gazebo, C++, Python, YOLO, SLAM\n","permalink":"https://noahbecker.tech/projects/naobot-soccer/","summary":"A team of simulated NAO humanoid robots playing soccer; first place in the class competition.","title":"NaoBot Soccer Challenge"},{"content":"A Django web application for complex visualization of large-scale power plant data from Siemens, built in cooperation with the FAU Department for Data Management. The app combines time-series plots, outlier analysis, and topological data analysis (persistent homology via giotto-tda) with a CNN-based time-series classifier. Its purpose: re-match sensor data to the correct sensors where labels were lost, and detect anomalies.\nThe two analysis views render interactive box plots, 3D projections, persistence diagrams, and raw time plots in the browser; the CNN classifies time series from their Gramian Angular Field encodings. A helper module converts between the involved data formats (SQL dumps, ZIP archives, CSV, NumPy arrays).\nTechnologies: Django, Python, Pandas, TensorFlow, giotto-tda, JavaScript\n","permalink":"https://noahbecker.tech/projects/power-plant-data-viz/","summary":"Django app for exploring large-scale Siemens power plant sensor data: time-series plots, outlier analysis, and topological projections.","title":"Power Plant Data Visualization Web App"},{"content":"An autonomous blackjack robot built with a LEGO Mindstorms EV3 that deals, scans, and moderates a blackjack variant played with UNO cards. A wheel-based dispenser feeds single cards into a scanning tray, where a cartesian gantry, homed and calibrated via touch sensors, moves a brightness sensor across the card and streams the raw pixel data over a socket to a PC.\nOn the PC, the flattened brightness readings are reconstructed into a 2D image, cleaned up (hole filling, histogram-based thresholding, edge-artifact removal, rescaling), and recognized with a digits-only Tesseract/Tess4J OCR setup; the recognized card value is sent back to the EV3.\nThe EV3 side (leJOS, Java) acts as the dealer: it runs the blackjack game logic for a table of four players, takes the player count via the brick\u0026rsquo;s buttons, reads hit/stand decisions from a floor-mounted touch sensor, and ejects each dealt card to the right player with a lever mechanism. Built as part of the LEGO Mindstorms Praktikum.\nTechnologies: LEGO Mindstorms EV3, leJOS (Java), Tess4J/Tesseract OCR, sockets\n","permalink":"https://noahbecker.tech/projects/lego-blackjack-robot/","summary":"An autonomous blackjack dealer built from LEGO Mindstorms: card dispensing, brightness-sensor scanning, and OCR card recognition for four players.","title":"LEGO Mindstorms Blackjack Robot"},{"content":"Download as PDF\nEducation M.Sc. in Autonomous Systems and Robotics (2022–Present) Technical University of Darmstadt, Germany Preliminary grade: 1.4\nB.Sc. in Computer Science (2018–2022) Friedrich-Alexander-Universität Erlangen-Nürnberg, Germany\nExperience Research Assistant (2024–Present) Intelligent Autonomous Systems Group, TU Darmstadt Research into ML and analytical slip detection using visuo-tactile sensing in the context of teleoperation. Published and presented research at ICRA@40 and ICRA 2024.\nResearch and Teaching Assistant (2023–2024) Artificial Intelligence and Machine Learning Lab, TU Darmstadt Teaching assistant for \u0026ldquo;Introduction to Artificial Intelligence\u0026rdquo;; designed the course \u0026ldquo;Introduction to Machine Learning in Python\u0026rdquo;.\nWorking Student – Agile Software Developer (2019–2022) Siemens Energy, Erlangen Digital Operations Management: process automation, web development, software integration, user management, and technical support.\nFull-stack and ML-Ops (2020–2021) Department for Data Management in cooperation with Siemens Energy, Erlangen Development of a web application for the classification and visualization of power plant data in Django.\nResearch and Teaching Assistant (2019–2020) Department for Data Management, FAU Erlangen-Nürnberg Tutor for Process-Oriented Information Systems, Computer Science for Engineers, and Knowledge Discovery in Databases.\nIntern (2018) GreenYellow Madagascar, Antananarivo System administration and procurement for a solar panel installation company.\nVolunteer Work Team Supervisor (2023–Present) — Campus Connect Darmstadt Coordinated multiple teams and ensured smooth operations during weekly events.\nYouth Group Leader (2016–2024) — Parish Elia Erlangen Organized and led activities, provided mentorship, tutoring, and fostered community involvement among youth.\nTutor (2018–2022) — Studienkreis Erlangen Tutoring in math, English, and computer science for middle and high school students.\nSkills Advanced: Python (OpenCV, PyTorch, Pandas, NumPy), Java, VBA, SQL, LaTeX, Unity, ROS1, ROS2\nIntermediate: C, C++, C#, JavaScript, Android Studio, Bash, PowerShell, prompt engineering, Matlab, LabVIEW, OpenGL\nOther: Linux Debian, Docker, Raspberry Pi, Arduino, 3D printing, Fusion 360, regulator design, pixel and vector art design (Affinity), Microsoft Office\nLanguages: German (native), English (C2), French (B1)\n","permalink":"https://noahbecker.tech/cv/","summary":"Education, experience, and skills.","title":"Curriculum Vitae"},{"content":"M.Sc. Autonomous Systems and Robotics Technical University of Darmstadt\nRobotics \u0026amp; Robot Learning Learning Robots Learning Robots: Integrated Project – Part 1 Learning Robots: Integrated Project – Part 2 Intelligent Robot Manipulation: Part II Expert Lab in Robot Learning Master\u0026rsquo;s Thesis Autonomous Systems and Robotics Machine Learning \u0026amp; AI Deep Learning: Architectures \u0026amp; Methods Introduction to Artificial Intelligence Optimization Algorithms Ambient Intelligence Computer Vision \u0026amp; Graphics Computer Vision I Computer Vision II 3D Scanning \u0026amp; Motion Capture Computer Graphics I Serious Games Mechatronics \u0026amp; Control Mechatronic Systems I System Dynamics and Control Engineering I Actuator Materials and Principles Mechatronics and Assistance Systems in Automobiles Languages Advanced Course I Comprehension and Expression – Preparation for UNIcert II B.Sc. Computer Science Friedrich-Alexander-Universität Erlangen-Nürnberg\nMathematics \u0026amp; Theory Mathematics for Computer Science 1 Mathematics for Computer Science 2 Mathematics for Computer Science 3 Mathematics for Computer Science 4 Algorithms for Continuous Systems Computability and Formal Languages Foundations of Logic in Computer Science Theory of Programming Algorithms \u0026amp; Programming Algorithms and Data Structures Parallel and Functional Programming Software Development in Large Projects Systems Programming Computer Architecture \u0026amp; Systems Fundamentals of Computer Architecture and Organization Fundamentals of Computer Engineering Fundamentals of Circuit Design Computer Networks Cyber-Physical Systems Data \u0026amp; Information Systems Implementation of Database Systems Database Practice Conceptual Modeling eBusiness Technologies and Evolutionary Information Systems Robotics \u0026amp; Machine Learning Fundamentals of Robotics Machine Learning for Engineers I — Introduction to Methods and Tools Lego Mindstorms Lab Bachelor\u0026rsquo;s Thesis: Optimizing Robotic Bin Picking of Radioactive Debris using Supervised and Imitation Learning Applied \u0026amp; Interdisciplinary International Supply Chain Management Seminar Energy Informatics ","permalink":"https://noahbecker.tech/transcript/","summary":"Master\u0026rsquo;s and bachelor\u0026rsquo;s coursework — Autonomous Systems and Robotics, TU Darmstadt; Computer Science, FAU Erlangen-Nürnberg.","title":"Transcript"}]