Projects & Creations

Description

• Developed a ROS2 Fanuc driver using Ethernet IP protocol, implemented in C++ using the EIP Scanner module, enabling full programmatic control of the Fanuc CrX10ia cobot.
• Built a custom PyQt5 UI for the Fanuc CrX10ia – Lincoln R450 PowerWave system, and developed a TCP/IP driver for the Lincoln R450 following the ArcLink message structure in Python.
• Programmed background TP and Karel firmware programs to interface with ROS2 running programs.
• Currently integrating an AI agent into the live robotic welding loop — beginning with LLMs and VLMs for high-level reasoning, with plans for task-specific SLMs.

Technologies

Description

• Helped build an in-pipe welding robotic arm at the SFU Motion and Power Electronics Control Lab. The mechanical design originated with the project lead; I owned the control software, embedded systems, and seam-tracking algorithms, and co-led electrical wiring and assembly.
• Developed and tested control software using TMS320F240 DSP (dSPACE) during prototyping; transitioned to Raspberry Pi Compute Module 4 and NXP ARM S32K144 MCU for final implementation.
• Created a seam tracking algorithm integrating Python scripts with line laser sensors for real-time accuracy, applying KNN algorithms for point-cloud filtering and piecewise-linear fitting for edge detection.
• Derived the DH kinematics, analytical Jacobian and Lagrangian rigid-body dynamics (inertia tensors from SolidWorks), then designed PI and computed-torque controllers for DC and PMSM motors plus a jerk-minimised piecewise-polynomial trajectory system interpolated in real time on the MCU. Validated on dSPACE via Hardware-in-the-Loop before the portable build.
• Measured closed-loop accuracy: ±0.13 cm in height control and 0.4° in rotation, with seam tracking and multi-pass welding emulation validated experimentally. Built a FreeMaster HMI and a JSON-RPC browser dashboard (HTML/CSS/JS) for real-time control and MCU tuning.

Technologies

Description

• Served as FEA lead on a Level 3 Fitness-For-Service assessment (API 579-1 / ASME FFS-1 Part 8) of the 1912 riveted/welded firebox on the Spirit of Kamloops, Locomotive 2141 — a regulator-facing, independently reviewed engineering report.
• Ran seven elasto-plastic finite-element simulations (SimScale) across four geometric configurations — design-intent, as-measured distorted, and weld-overlay repair variants — under pressure-only and coupled transient thermo-mechanical load cases at 200 psi MAWP.
• Reconstructed the as-measured firebox geometry from a Revoscan 3D laser-scan point cloud (slicing, thickening to measured plate thickness, sweeping into a clean parametric CAD model), and built nonlinear material models with bilinear isotropic hardening and temperature-dependent properties at the 350 °C design metal temperature, including a separately calibrated E7018 weld-overlay sub-material.
• Captured real structural behaviour through symmetry planes, Winkler elastic supports with per-stay stiffness K = EA/L calibrated to missing stay lengths, frictionless sliding contacts for ball-end stays, and a six-zone convective heat-flux BC set blending convection and radiation from 1500 °C fire-side to nucleate-boiling water-side HTCs.
• Automated stress/strain post-processing in Python (PyVista/SciPy) — projecting analytic paths onto deformed meshes, computing Dijkstra geodesic worst-orientation paths, and running pointwise local-failure (εL) and stress-linearization checks per ASME VIII Div 2 Annex 5.A, cross-validated against native solver output to within ±2 ksi. Mesh-sensitivity studies at 7× node-count refinement were independently reproduced in Abaqus to within ±5%.
• Authored the internal technical report on firebox corner-bend geometric analysis, with an interactive MATLAB tool for plane slicing, PCA profile ordering, bend-angle/inscribed-arc extraction, and stay-pitch measurement.
• The analysis supported return-to-service at the full historical 200 psi MAWP with no derate, avoiding a repair estimated at $200,000+.
• Alongside the FFS work: conducted cobot welding experiments at BCIT with laser scan and point-cloud processing before and after welds to detect geometric distortion; developed frontend software tools for welding engineers (JavaScript) and managed the website and database for Fleet Management (Ruby on Rails).

Technologies

Description

• Designed and implemented a fully automated conveyor-based sorting system using Studio 5000 and a CompactLogix L36ERMS Safety PLC.
• Developed modular ladder logic architecture using multiple subroutines with structured JSR execution and a state-machine–based control strategy.
• Integrated photo-eye sensors, pushbuttons, indicator lamps, and pneumatic actuators with deterministic timing using TON timers and CTU counters.
• Programmed PowerFlex 525 VFD control via digital I/O for safe start/stop sequencing. Designed fault detection logic using timers and counters to detect missing bins, initialization timeouts, and E-stop conditions.

Technologies

Description

• Built a fully automated mobile robot using Hector SLAM for navigation, with a YOLOv3 model for detecting stop signs.
• Implemented an OpenCV line-tracing algorithm and a CNN network for detecting cat and dog images via ArUco markers, enabling the robot to determine where to drop red and blue balls to accomplish its mission.

Technologies

Description

• Research project analysing the stress resistance of fiberglass-reinforced versus steel-reinforced concrete under heavy loading, using ANSYS finite-element simulation.
• Compared deflection and stress-distribution behaviour between the two reinforcement schemes to assess viability of fiberglass rebar for concrete floor reinforcement.

Technologies

Description

• Wrote a register-level ATmega328p library from the datasheet up to maximise hardware utilisation, bypassing the Arduino abstraction layer.
• Implemented a PID speed controller on top of it and tuned it against the physical vehicle.

Technologies

Description

• Technical review of K.C. Kong's work on Rotary Series Elastic Actuator control for human-robot interaction, with accompanying Simulink simulations reproducing the control behaviour.
• Explored compliance and force control as they apply to safe physical human-robot interaction.

Technologies

Description

• Proposed and prototyped a device capable of generating artificial 3D touch sensation from a 2D picture, aiming to solve the navigation problem for visually impaired or blind users.
• The solution extracts 3D dense information from a 2D image using deep learning and translates it into 3D tactile sensation applied to the user's skin via electromechanical devices.
• Combined mobile device, embedded system, and deep learning into an end-to-end wearable navigation assistant.

Technologies