[ { "id": "UR7e-Reinforcement-Learning-Suite", "title": "Robot learning, from simulation to deployment", "domain": "Robotics", "tags": "SAC / HER · ROS 2 · perception · EKF", "question": "Build a trajectory-following architecture that connects policy learning, camera observations and kinematic control.", "method": "Four separate implementations cover simple and full simulation, partial real deployment, and target reaching with obstacle avoidance.", "evidence": "The public source contains an original 520-episode simulation training log, policy/environment code, kinematic components and tests. Explore the exact log in the Run explorer.", "scope": "Real closed-loop deployment is incomplete: the laser point is lost during motion. Training curves do not establish held-out or physical-robot performance.", "authors": "Dev Kumar is the documented maintainer; consult the source for project contribution context.", "license": "MIT", "repository": "https://github.com/devkumar-projects/UR7e-Reinforcement-Learning-Suite", "commit": "fbba2e4ef768219b7f925c091ff027d025883500", "sources": [ { "label": "line-follower-simulation-complete/package/ur7e_line_follower/env.py", "url": "https://github.com/devkumar-projects/UR7e-Reinforcement-Learning-Suite/blob/fbba2e4ef768219b7f925c091ff027d025883500/line-follower-simulation-complete/package/ur7e_line_follower/env.py" }, { "label": "line-follower-simulation-complete/package/ur7e_line_follower/ekf.py", "url": "https://github.com/devkumar-projects/UR7e-Reinforcement-Learning-Suite/blob/fbba2e4ef768219b7f925c091ff027d025883500/line-follower-simulation-complete/package/ur7e_line_follower/ekf.py" }, { "label": "line-follower-real-implementation/README.md", "url": "https://github.com/devkumar-projects/UR7e-Reinforcement-Learning-Suite/blob/fbba2e4ef768219b7f925c091ff027d025883500/line-follower-real-implementation/README.md" } ], "notebooks": [ { "label": "UR7e Reinforcement Learning Training Log Analysis", "url": "https://www.kaggle.com/code/devkumarkaggle/ur7e-reinforcement-learning-training-log-analysis" }, { "label": "Generating Robot Drawing Trajectories on CPU", "url": "https://www.kaggle.com/code/devkumarkaggle/generating-robot-drawing-trajectories-on-cpu" }, { "label": "UR7e Data Quality and Failure Analysis", "url": "https://www.kaggle.com/code/devkumarkaggle/ur7e-data-quality-and-failure-analysis" }, { "label": "UR7e Laser Point Tracking with Kalman Filter", "url": "https://www.kaggle.com/code/devkumarkaggle/ur7e-laser-point-tracking-with-kalman-filter" }, { "label": "UR7e Simulated Robot Line-Following Training Log", "url": "https://www.kaggle.com/datasets/devkumarkaggle/ur7e-simulated-robot-line-following-training-log" } ], "reports": [] }, { "id": "Classical-Computer-Vision-Toolkit", "title": "Geometric perception: alignment, tracking, segmentation", "domain": "Perception", "tags": "ICP · KLT / PnP · circular Hough", "question": "Recover geometry from images and point clouds with interpretable classical methods.", "method": "Three components address rigid 3D alignment, optical-flow pose estimation and circular-feature segmentation.", "evidence": "The ICP component includes a notebook; the toolkit also includes KLT/PnP Python code and a Hough notebook. The linked ICP Lab is an independent browser implementation.", "scope": "Some original examples require external images or datasets. Synthetic notebook results do not measure real-material porosity or guarantee global registration.", "authors": "Dev Kumar is explicitly credited for the ICP component; attribution for the full toolkit follows its source documentation.", "license": "MIT", "repository": "https://github.com/devkumar-projects/Classical-Computer-Vision-Toolkit", "commit": "256ef38551d258ebfc1789bc38fef59a541775e9", "sources": [ { "label": "3d-point-cloud-registration-icp/ICP_3D_PointCloud_Registration.ipynb", "url": "https://github.com/devkumar-projects/Classical-Computer-Vision-Toolkit/blob/256ef38551d258ebfc1789bc38fef59a541775e9/3d-point-cloud-registration-icp/ICP_3D_PointCloud_Registration.ipynb" }, { "label": "klt-tracking-pnp-pose-estimation/KLT_PNP.py", "url": "https://github.com/devkumar-projects/Classical-Computer-Vision-Toolkit/blob/256ef38551d258ebfc1789bc38fef59a541775e9/klt-tracking-pnp-pose-estimation/KLT_PNP.py" }, { "label": "hough-circle-pore-segmentation/Hough_Circle_Pore_Segmentation.ipynb", "url": "https://github.com/devkumar-projects/Classical-Computer-Vision-Toolkit/blob/256ef38551d258ebfc1789bc38fef59a541775e9/hough-circle-pore-segmentation/Hough_Circle_Pore_Segmentation.ipynb" } ], "notebooks": [ { "label": "3D Point Cloud Registration with ICP", "url": "https://www.kaggle.com/code/devkumarkaggle/3d-point-cloud-registration-with-icp" }, { "label": "Detecting Synthetic Pores with Circular Hough", "url": "https://www.kaggle.com/code/devkumarkaggle/detecting-synthetic-pores-with-circular-hough" } ], "reports": [] }, { "id": "RST-controller-for-an-inverted-pendulum", "title": "Polynomial control of an unstable system", "domain": "Control", "tags": "Bézout identity · internal model · pole placement", "question": "Stabilize a reduced inverted-pendulum model and design sinusoidal reference tracking.", "method": "The MATLAB workflow synthesizes R, S and T and evaluates nominal and retuned designs; a Python figure explains pole placement.", "evidence": "The source includes MATLAB/Simulink files and a technical report. Control Lab below independently recomputes the algebra and frequency response in your browser.", "scope": "The interactive lab studies the nominal reduced linear transfer function. It does not reproduce the complete Simulink cart-pendulum model, saturation or hardware.", "authors": "Dev Kumar · academic supervision: Hervé Guillard.", "license": "MIT", "repository": "https://github.com/devkumar-projects/RST-controller-for-an-inverted-pendulum", "commit": "0a8a383810060c9c4d79e0ba39a400f11a3acb09", "sources": [ { "label": "matlab/rst_pendulum_synthesis.m", "url": "https://github.com/devkumar-projects/RST-controller-for-an-inverted-pendulum/blob/0a8a383810060c9c4d79e0ba39a400f11a3acb09/matlab/rst_pendulum_synthesis.m" }, { "label": "python/pole_placement_plot.py", "url": "https://github.com/devkumar-projects/RST-controller-for-an-inverted-pendulum/blob/0a8a383810060c9c4d79e0ba39a400f11a3acb09/python/pole_placement_plot.py" }, { "label": "README.md", "url": "https://github.com/devkumar-projects/RST-controller-for-an-inverted-pendulum/blob/0a8a383810060c9c4d79e0ba39a400f11a3acb09/README.md" } ], "notebooks": [ { "label": "RST Pole Placement: Pendulum in Python", "url": "https://www.kaggle.com/code/devkumarkaggle/rst-pole-placement-pendulum-in-python" } ], "reports": [ { "label": "RST Inverted Pendulum Report", "url": "https://github.com/devkumar-projects/RST-controller-for-an-inverted-pendulum/blob/0a8a383810060c9c4d79e0ba39a400f11a3acb09/docs/RST_Inverted_Pendulum_Report.pdf" } ] }, { "id": "TurtleBot3-ROS2-Navigation-Challenges", "title": "A mobile robot with a unified navigation state machine", "domain": "Robotics", "tags": "ROS 2 · LiDAR · HSV segmentation · odometry", "question": "Navigate a corridor, follow a coloured line and perform a ball-pushing task in simulation.", "method": "Separate perception and control nodes connect LiDAR, camera and odometry within a unified behavioural architecture.", "evidence": "Public ROS 2 source, launch configuration and a video demonstration document the pipeline. A CPU Kaggle example isolates the green-line mask on synthetic images.", "scope": "The original Gazebo course world is external and is not bundled. The synthetic notebook is a perception illustration, not a reproduction of the full robot.", "authors": "Dev Kumar and David Tovmassian.", "license": "Apache-2.0", "repository": "https://github.com/devkumar-projects/TurtleBot3-ROS2-Navigation-Challenges", "commit": "51278c9b8467263ed9ce5774036b8a5ebd14c9c5", "sources": [ { "label": "projet/line_follow_node.py", "url": "https://github.com/devkumar-projects/TurtleBot3-ROS2-Navigation-Challenges/blob/51278c9b8467263ed9ce5774036b8a5ebd14c9c5/projet/line_follow_node.py" }, { "label": "README.md", "url": "https://github.com/devkumar-projects/TurtleBot3-ROS2-Navigation-Challenges/blob/51278c9b8467263ed9ce5774036b8a5ebd14c9c5/README.md" }, { "label": "package.xml", "url": "https://github.com/devkumar-projects/TurtleBot3-ROS2-Navigation-Challenges/blob/51278c9b8467263ed9ce5774036b8a5ebd14c9c5/package.xml" } ], "notebooks": [ { "label": "TurtleBot3 Green Line Segmentation on CPU", "url": "https://www.kaggle.com/code/devkumarkaggle/turtlebot3-green-line-segmentation-on-cpu" } ], "reports": [] }, { "id": "ur5e-ReinforcementLearning-pick-and-place", "title": "Goal-conditioned pick-and-place", "domain": "Robotics", "tags": "SAC · sparse rewards · hindsight experience replay", "question": "Make a simulated manipulator learn from goal-distance feedback and sparse rewards.", "method": "A ROS 2 / Gazebo environment exposes goal-conditioned observations and a sparse reward for SAC with HER.", "evidence": "Public environment and training code are accompanied by a CPU notebook that verifies the reward threshold and demonstrates hindsight relabelling.", "scope": "The notebook uses synthetic positions. It neither trains a policy nor evaluates a physical robot. Maintainer identity alone is not a complete contribution list.", "authors": "Dev Kumar is the documented maintainer.", "license": "MIT", "repository": "https://github.com/devkumar-projects/ur5e-ReinforcementLearning-pick-and-place", "commit": "f332fc641d63f8b71d755efe2d1475f7bb48efec", "sources": [ { "label": "ur5e_rl_gazebo/env.py", "url": "https://github.com/devkumar-projects/ur5e-ReinforcementLearning-pick-and-place/blob/f332fc641d63f8b71d755efe2d1475f7bb48efec/ur5e_rl_gazebo/env.py" }, { "label": "README.md", "url": "https://github.com/devkumar-projects/ur5e-ReinforcementLearning-pick-and-place/blob/f332fc641d63f8b71d755efe2d1475f7bb48efec/README.md" } ], "notebooks": [ { "label": "Sparse Rewards and HER for UR5e Robots", "url": "https://www.kaggle.com/code/devkumarkaggle/sparse-rewards-and-her-for-ur5e-robots" } ], "reports": [] }, { "id": "Control-of-an-instable-system-using-digital-state-feedback", "title": "Twin-rotor identification and state-space control", "domain": "Control", "tags": "System identification · integral feedback · LQR", "question": "Control a coupled laboratory helicopter while balancing response speed and actuator effort.", "method": "Experimental identification feeds integral pole placement and multivariable LQR synthesis, with MATLAB/Simulink and dSPACE implementation.", "evidence": "The report distinguishes simulation traces and real-time experiments. Source scripts implement the controller designs and document the speed–effort trade-off.", "scope": "Numerical values are reported by the project; no independent rerun of dSPACE experiments is claimed here.", "authors": "Dev Kumar, Steevy Doumbe Moundy and Léa Couësme.", "license": "MIT", "repository": "https://github.com/devkumar-projects/Control-of-an-instable-system-using-digital-state-feedback", "commit": "7b2a966ac26a43f959ede9ce81aca5e787416679", "sources": [ { "label": "matlab/integral_state_feedback_pole_placement.m", "url": "https://github.com/devkumar-projects/Control-of-an-instable-system-using-digital-state-feedback/blob/7b2a966ac26a43f959ede9ce81aca5e787416679/matlab/integral_state_feedback_pole_placement.m" }, { "label": "matlab/lqr_integral_control.m", "url": "https://github.com/devkumar-projects/Control-of-an-instable-system-using-digital-state-feedback/blob/7b2a966ac26a43f959ede9ce81aca5e787416679/matlab/lqr_integral_control.m" }, { "label": "README.md", "url": "https://github.com/devkumar-projects/Control-of-an-instable-system-using-digital-state-feedback/blob/7b2a966ac26a43f959ede9ce81aca5e787416679/README.md" } ], "notebooks": [], "reports": [ { "label": "Helicopter Control Report", "url": "https://github.com/devkumar-projects/Control-of-an-instable-system-using-digital-state-feedback/blob/7b2a966ac26a43f959ede9ce81aca5e787416679/docs/Helicopter_Control_Report.pdf" } ] }, { "id": "HeMT-GaN-Transistor", "title": "GaN device modelling with explicit verification boundaries", "domain": "Engineering", "tags": "Semiconductor physics · Ecorce · deck auditing", "question": "Investigate an exploratory normally-off vertical GaN concept and make the available simulation artifacts inspectable.", "method": "The archive separates conceptual device reasoning from the recovered lateral GaN HEMT deck. Dependency-free Python tools inspect device geometry, contacts, materials and stimulus consistency.", "evidence": "The recovered .dev/.sti files, model manifest, Python inspector, tests and report are publicly inspectable.", "scope": "The recoverable runnable deck is lateral, not a validated vertical VertiGaN device. Ecorce and proprietary material libraries are excluded; no breakdown rating, manufacturing qualification or patentability is established.", "authors": "Dev Kumar · scientific supervision: Tanguy Phulpin · software support: Alain Michez.", "license": "MIT for repository-authored material; third-party exclusions in NOTICE.md", "repository": "https://github.com/devkumar-projects/HeMT-GaN-Transistor", "commit": "fefb15a0696f7f4c0545ac638b33a2c4449e3b4c", "sources": [ { "label": "tools/ecorce_deck.py", "url": "https://github.com/devkumar-projects/HeMT-GaN-Transistor/blob/fefb15a0696f7f4c0545ac638b33a2c4449e3b4c/tools/ecorce_deck.py" }, { "label": "models/ecorce/lateral-gan-hemt/model-manifest.json", "url": "https://github.com/devkumar-projects/HeMT-GaN-Transistor/blob/fefb15a0696f7f4c0545ac638b33a2c4449e3b4c/models/ecorce/lateral-gan-hemt/model-manifest.json" }, { "label": "NOTICE.md", "url": "https://github.com/devkumar-projects/HeMT-GaN-Transistor/blob/fefb15a0696f7f4c0545ac638b33a2c4449e3b4c/NOTICE.md" } ], "notebooks": [ { "label": "Auditing GaN Simulation Decks with Python", "url": "https://www.kaggle.com/code/devkumarkaggle/auditing-gan-simulation-decks-with-python" } ], "reports": [ { "label": "VertiGaN CNRS Research Internship Report", "url": "https://github.com/devkumar-projects/HeMT-GaN-Transistor/blob/fefb15a0696f7f4c0545ac638b33a2c4449e3b4c/report/VertiGaN_CNRS_Research_Internship_Report.pdf" } ] }, { "id": "-Design-of-a-Turbomachine", "title": "An integrated hydraulic turbine design chain", "domain": "Engineering", "tags": "Hydraulics · CAD · CFD · generator design · prototyping", "question": "Connect hydraulic sizing to geometry, flow simulation, electromechanical conversion and prototype testing.", "method": "A multidisciplinary workflow combines MATLAB/XFoil, CATIA, STAR-CCM+, FEMM/SIMBA and additive manufacturing.", "evidence": "The report documents a compact runner design, preliminary simulations and a wind-tunnel prototype analogue.", "scope": "The wind-tunnel analogue is distinct from full-scale hydraulic qualification. The separate 2 MW industrial scenario is not the reported 1.76 kW compact-runner CFD result. The inherited FEMM template has a separate attribution.", "authors": "Dev Kumar, Thien Ho, Seyf Daab and Wilhem Abboura.", "license": "MIT stated; inherited FEMM-template rights require separate verification", "repository": "https://github.com/devkumar-projects/-Design-of-a-Turbomachine", "commit": "6ea7c35a20d026a5e62b155061d2ab4f1e835496", "sources": [ { "label": "README.md", "url": "https://github.com/devkumar-projects/-Design-of-a-Turbomachine/blob/6ea7c35a20d026a5e62b155061d2ab4f1e835496/README.md" } ], "notebooks": [], "reports": [ { "label": "Compact Axial Hydraulic Turbine Report", "url": "https://github.com/devkumar-projects/-Design-of-a-Turbomachine/blob/6ea7c35a20d026a5e62b155061d2ab4f1e835496/docs/Compact_Axial_Hydraulic_Turbine_Report.pdf" } ] }, { "id": "-Spatial-Verification-of-a-Coordinate-Measuring-Machine", "title": "Robust geometric metrology and error identification", "domain": "Engineering", "tags": "Normal deviations · Huber IRLS · sphere fitting", "question": "Estimate geometric offsets while separating systematic error, form and direction-dependent effects.", "method": "A MATLAB pipeline fits four normal-deviation parameters, corrects the identified component and exports before/after diagnostics.", "evidence": "The public MATLAB implementation and report document planar tests and the limitations of the original angular identification. An independent synthetic Kaggle notebook explores robust fitting and identifiability.", "scope": "The original angular channels are not validated; the four-parameter correction is not a complete volumetric map or ISO acceptance test. The source component is licensed MIT; third-party material retains its own terms.", "authors": "Dev Kumar and Thien Ho · supervision: Thierry Coorevits.", "license": "MIT for repository-authored source; see LICENSE and NOTICE for scope", "repository": "https://github.com/devkumar-projects/-Spatial-Verification-of-a-Coordinate-Measuring-Machine", "commit": "1a3e68c66cb5fd97b11cb1949a25839b511caa9b", "sources": [ { "label": "matlab/Metrology_Correction.m", "url": "https://github.com/devkumar-projects/-Spatial-Verification-of-a-Coordinate-Measuring-Machine/blob/1a3e68c66cb5fd97b11cb1949a25839b511caa9b/matlab/Metrology_Correction.m" }, { "label": "README.md", "url": "https://github.com/devkumar-projects/-Spatial-Verification-of-a-Coordinate-Measuring-Machine/blob/1a3e68c66cb5fd97b11cb1949a25839b511caa9b/README.md" }, { "label": "Licence scope and third-party attribution", "url": "https://github.com/devkumar-projects/-Spatial-Verification-of-a-Coordinate-Measuring-Machine/blob/1a3e68c66cb5fd97b11cb1949a25839b511caa9b/NOTICE.md" } ], "notebooks": [ { "label": "Robust CMM Error Estimation on Synthetic Data", "url": "https://www.kaggle.com/code/devkumarkaggle/robust-cmm-error-estimation-on-synthetic-data" } ], "reports": [ { "label": "Spatial Verification of a CMM Report", "url": "https://github.com/devkumar-projects/-Spatial-Verification-of-a-Coordinate-Measuring-Machine/blob/1a3e68c66cb5fd97b11cb1949a25839b511caa9b/docs/Spatial_Verification_of_a_CMM_Report.pdf" } ] }, { "id": "INTARSO-Shooting-Range-Target-Robot", "title": "Embedded control delivered as a physical system", "domain": "Robotics", "tags": "ESP32 · LiDAR · VESC · web HMI · electronics", "question": "Control a motorized target carriage with distance feedback and an operator interface.", "method": "Firmware, sensing, motor commands, custom electronics and a wireless HMI are integrated as an engineering system.", "evidence": "The repository includes firmware and design artifacts, with a video demonstrating the delivered system.", "scope": "A physical demonstration is not functional-safety certification. Code, client documents, photos and third-party components have distinct reuse considerations.", "authors": "Dev Kumar and Tedjeddine Sebiane · JEECE / AMJE collaboration.", "license": "MIT for identified source; assess other artifacts separately", "repository": "https://github.com/devkumar-projects/INTARSO-Shooting-Range-Target-Robot", "commit": "8053d3f144073fef3278299eb9b037bb3160da89", "sources": [ { "label": "README.md", "url": "https://github.com/devkumar-projects/INTARSO-Shooting-Range-Target-Robot/blob/8053d3f144073fef3278299eb9b037bb3160da89/README.md" } ], "notebooks": [], "reports": [] }, { "id": "ProcureLens-AI", "title": "Auditable price extraction and explainable ranking", "domain": "Software", "tags": "Python · structured extraction · deterministic features · reporting", "question": "Turn candidate price observations into a traceable, reviewable output.", "method": "Extraction, validation, ranking and report generation are separated. Offline fictional examples exercise the deterministic local path.", "evidence": "The repository contains a desktop/CLI application, tests, synthetic fixtures and provider interfaces. A Kaggle notebook demonstrates the heuristic without scraping or credentials.", "scope": "Default heuristic scores are not calibrated probabilities. Optional external providers are not used in this workbench. No customer data or production-accuracy claim is included.", "authors": "Dev Kumar.", "license": "MIT", "repository": "https://github.com/devkumar-projects/ProcureLens-AI", "commit": "f3cb4f8b381d8b3a2bf17f6414c5ef0922804cee", "sources": [ { "label": "src/openprice_scraper/ranker.py", "url": "https://github.com/devkumar-projects/ProcureLens-AI/blob/f3cb4f8b381d8b3a2bf17f6414c5ef0922804cee/src/openprice_scraper/ranker.py" }, { "label": "tests/test_ranker.py", "url": "https://github.com/devkumar-projects/ProcureLens-AI/blob/f3cb4f8b381d8b3a2bf17f6414c5ef0922804cee/tests/test_ranker.py" }, { "label": "README.md", "url": "https://github.com/devkumar-projects/ProcureLens-AI/blob/f3cb4f8b381d8b3a2bf17f6414c5ef0922804cee/README.md" } ], "notebooks": [ { "label": "Explainable Price Ranking with ProcureLens", "url": "https://www.kaggle.com/code/devkumarkaggle/explainable-price-ranking-with-procurelens" } ], "reports": [] }, { "id": "The-game-2048", "title": "A complete small Python application", "domain": "Software", "tags": "Python · Tkinter · SQLite · application tests", "question": "Implement a familiar game as a small maintainable desktop application.", "method": "Game state, interface and persistence are separated in a pedagogical application.", "evidence": "Public source and tests demonstrate foundational software construction.", "scope": "This is educational application work; it is not presented as a research contribution or AI benchmark.", "authors": "Dev Kumar, Michel Attoum and Celine Fernandes, as recorded in AUTHORS.md and LICENSE.", "license": "MIT", "repository": "https://github.com/devkumar-projects/The-game-2048", "commit": "a3f251f0d1f3d86bf6dadf6593111183b0750e30", "sources": [ { "label": "AUTHORS.md", "url": "https://github.com/devkumar-projects/The-game-2048/blob/a3f251f0d1f3d86bf6dadf6593111183b0750e30/AUTHORS.md" }, { "label": "README.md", "url": "https://github.com/devkumar-projects/The-game-2048/blob/a3f251f0d1f3d86bf6dadf6593111183b0750e30/README.md" } ], "notebooks": [], "reports": [] } ]