Marisa Patel

Project Overview

• Addresses the challenges and physical effort involved in manually fastening and unfastening vehicle lug nuts, which is a repetitive task prone to human error and inconsistency in torque application.
• Designed to be portable and accessible for everyday users, while being less expensive than industrial tire-changing machines.
• The finished system detects lug nut positions using a camera, moves a drill to each position automatically, and fastens/unfastens each bolt with minimal human input.
• Main components include a Raspberry Pi 4, ESP32-CAM, Raspberry Pi Pico 2, NEMA-17 stepper motors, a CoreXY gantry, a linear actuator, and a DC drill motor.
• Software stack includes Python, OpenCV, and MicroPython.

Engineering Approach

• The system was divided into four modules: Computer Vision and Detection, Mechanical CoreXY, CoreXY Motion Control, and the Drill subsystem.
• Computer Vision and Detection:
• The vision module runs on a Raspberry Pi 4, which generates its own Wi-Fi hotspot to connect to the ESP32-CAM. The camera streams live video so the user can position the tire hub before triggering an image capture.
• Bolt detection was implemented in Python using OpenCV. This process converts the image to grayscale, applies Gaussian blur to reduce noise, uses binary thresholding to isolate dark bolt faces, and filters contours by area and circularity to identify each bolt.
• Bolt centroids are calculated using image moments and converted from pixel coordinates to real-world millimeter positions using a pixel-to-mm scaling factor derived from the known bolt diameter.
• CoreXY and Motion Control:
• The CoreXY mechanical system consists of two NEMA-17 stepper motors driving GT2 timing belt pulleys along linear rods and LM8LUU bearings, providing low-friction planar motion across the workspace.
• The CoreXY architecture was chosen over a traditional gantry for its reduced moving mass and higher speed, where both NEMA-17 motors are controlled by Adafruit TMC2209 drivers that work simultaneously to produce X and Y motion.
• Coordinates from the Computer Vision and Detection module are written to a text file and passed to a Raspberry Pi Pico 2 programmed in MicroPython, which converts each position into synchronized NEMA-17 stepper motor commands to drive the CoreXY carriage to each bolt.
• Drill Actuation:
• Once the carriage reaches each bolt, a linear actuator controlled by a TB6612FNG driver extends in the Z-direction to engage the drill, while a BTS7960 driver controls the DC drill motor for clockwise or counterclockwise rotation to fasten or unfasten the lug nut.
• Power is supplied by a 12V LiPo battery with a buck-boost converter stepping the voltage down to 6V for safe operation across components.
• Structural components and the drill carriage were designed in Onshape and fabricated using 3D printing.

Outcome

• The system successfully performed end-to-end operation, from image capture and bolt detection to CoreXY positioning and drill actuation with minimal manual intervention between stages.
• Bolt detection was validated through iterative testing across different lighting conditions, with detection parameters tuned to reliably identify all lug nut positions in the image.
• The CoreXY system demonstrated accurate and repeatable positioning, consistently reaching intended lug nut coordinates at a stable operating speed of approximately 6 mm/s.
• The linear actuator and drill motor reliably engaged and disengaged each bolt across repeated test cycles with consistent alignment.
• The project validated the feasibility of a low-cost, portable alternative to industrial tire-changing equipment using accessible components and open-source software.

Project Overview

• The AEV Autonomous Navigation System project centered on a scale model autonomous electric vehicle developed to integrate embedded electronics, sensing, and control software into a functioning self driving platform.
• Core hardware included a NVIDIA Jetson Nano embedded computer, a VESC motor controller with brushless DC drive and servo steering, a LiDAR laser scanner, an inertial measurement unit, a depth camera, and a joystick for manual driving.
• Software was developed in the Robot Operating System framework using Python and C++, alongside MATLAB for supporting analysis, running on Linux across a workspace of custom and modified ROS packages.
• The project combined driver assist features with fully autonomous navigation, culminating in a vehicle capable of avoiding collisions, following walls, and navigating open environments without human input.

Engineering Approach

• Configured the operating system and hardware interfaces to reliably recognize the sensor and controller hardware, then built and tested ROS nodes for the LiDAR, camera, and joystick before integrating them into a shared workspace.
• Programmed and calibrated the motor controller through firmware and modified driver code to publish and accept motor and servo commands, establishing the relationship between commanded velocity and steering angle and actual vehicle motion using a racecar simulator and bicycle model kinematics.
• Derived and implemented odometry equations to estimate vehicle position and heading from wheel speed, fused with inertial orientation data and refined through LiDAR based scan matching, and validated the resulting localization against Simultaneous Localization and Mapping generated maps, achieving 5cm of accuracy.
• Developed a driver assist collision avoidance algorithm that applied a repelling force proportional to obstacle proximity from LiDAR scans, blending it with the operator's manual joystick commands to correct velocity and steering in real time.
• Designed a self driving wall following controller combining feedback linearization with a proportional derivative feedback law to regulate vehicle distance from corridor walls detected through LiDAR.
• Extended the navigation algorithm by formulating and solving a quadratic optimization problem to compute virtual separating line barriers from LiDAR returns, allowing the vehicle to identify the largest open gap ahead, center itself between obstacles, and reverse course automatically when facing a dead end.

Outcome

• Delivered a fully integrated autonomous vehicle system tested in both simulated and physical environments, capable of operator assisted driving, wall following, and optimization based obstacle avoidance.
• Demonstrated localization accuracy of 5cm, enabling reliable estimation of the vehicle position relative to its surroundings for both driver assist and autonomous modes.

Project Overview

• Addresses the need for an accessible, low cost way to capture spatial distance data of an indoor environment and reconstruct it into a usable 3D digital representation.
• Designed as a compact, portable scanning tool that can be aimed at a hallway, room, or object and produce an accurate 3D model of its surroundings without expensive lidar equipment.
• The finished system rotates a time of flight sensor through a full 360 degree sweep, captures distance readings at each angular step, and outputs a completed 3D point cloud and connected mesh visualization of the scanned space.
• Main components include a TI MSP432E401Y microcontroller, a VL53L1X time of flight sensor, a MOT 28BYJ 48 stepper motor with a ULN2003 driver board, and a PC running Python.
• Software stack includes embedded C for the microcontroller and Python 3.9 with the math, serial, numpy, and Open3D libraries for data processing and visualization.

Engineering Approach

• The VL53L1X sensor emits a 940nm invisible laser pulse that reflects off the nearest surface, and the sensor measures the travel time of that pulse, multiplies it by the speed of light, and divides the result by two to determine the distance to the object.
• I2C communication protocol linked the time of flight sensor to the microcontroller, following the standard two wire serial interface.
• The MOT 28BYJ 48 stepper motor, driven through a ULN2003 driver board, rotated the sensor in fixed 22.5 degree increments, completing 16 measurements per full 360 degree rotation.
• An onboard pushbutton configured with interrupts controlled the start and stop of each scan, while onboard LEDs provided real time status indication as each measurement was taken.
• Once a full rotation completed, the system paused for the user to manually displace the entire setup along the x axis before beginning the next scan, allowing multiple slices to be captured along a hallway.
• Completed distance and angle data were sent from the microcontroller to the PC through UART communication at a baud rate of 115200.
• Two Python scripts handled the software side, with the first script converting each distance and stepper angle reading into xyz coordinates using trigonometric functions, and the second script reading the resulting xyz file to generate and render the final point cloud and 3D mesh using the Open3D library.
• Testing and development were carried out on a laptop running Python 3.9 IDLE, with Keil, an embedded development environment, used to write, build, and load the firmware onto the microcontroller.

Outcome

• The system successfully performed end to end operation, from triggering a scan to capturing, transmitting, and rendering the distance data as a 3D visualization.
• A full 3D point cloud and mesh representation of a hallway in McMaster's Engineering Technology Building was generated, confirming the sensor, motor, and processing pipeline worked together accurately across multiple scans.
• The conversion pipeline accurately transformed raw distance and angle readings into correct xyz coordinates, validated through manual calculation checks against the Python output.

Project Overview

• Built for a client living with Ehlers Danlos Syndrome, a connective tissue condition that makes gripping and applying force through the hand joints painful and difficult during everyday tasks such as cutting food.
• The problem centered on standard knives, which require bending the fingers around a handle and pushing downward, placing significant strain on the carpometacarpal and metacarpophalangeal joints.
• The finished device is a fold over slicer attached to a cutting board, allowing the client to prepare food with a flat hand placement and without the lateral cutting motion required by a standard knife.
• Main components include a wooden cutting board with a sliding cutout and drawer, a hinged slicer with a cushioned wrist strap, detachable 3D printed blades, and a modular pin and hinge assembly.
• The device was designed to be lightweight, ergonomic, and adaptive, with requirements identified through client interviews and hand measurements early in the design process.
• Key design constraints included using food safe materials throughout, keeping the user's hand flat during operation, ensuring the blades were detachable for safe storage, and keeping the overall weight under 5 pounds so the client could carry the device easily.

Engineering Approach

• Several initial prototypes were explored, including a slicer with a handle, a cutting board attachment, and a hand rest, before the strongest features were combined into a single fold over slicer design.
• The slicer and cutting board were modeled in Autodesk Inventor using measurements taken directly from the client's hand, then fabricated through a combination of woodworking and 3D printing.
• Wood formed the main structural components, PLA filament was 3D printed for the blades and slicer strap for its food safety, durability, and ease of reprinting at home, and metal hinges and nails secured the slicer to the board.
• Small divots were sanded into the slicer surface to match the client's hand shape and support his joints, while cotton lining under the strap and rounded edges reduced pressure points and discomfort.
• To operate the device, the user secures the cutting board, attaches the slicer through its pin and hinge system, inserts the blade assembly, then slides a hand under the strap and rotates downward as the blades cut the food resting on the board below.
• Rubber and plastic no slip stoppers on the underside of the board kept it stable on kitchen counters during use.
• All detachable components, including the blades, blade insert, and slicer assembly, are dishwasher safe, meeting the client's need for easy cleaning and maintenance.

Outcome

• The device successfully functioned as a working assistive cutting tool, allowing the client to prepare food using a flat hand position with reduced lateral movement and joint strain compared to a standard knife.
• Testing confirmed the design could adaptively cut a range of small food products.
• The prototype weighed approximately 2 kilograms, meeting the lightweight objective, and testing indicated a lower estimated injury rate than cutting with a regular knife.
• The final deliverable was a fully assembled, functional wooden prototype with a fold over slicer, drawer, and modular hardware, successfully meeting the project's ergonomic, lightweight, and adaptive objectives for the client's daily living needs.

Project Overview

• Addresses the low recycling rate in Canada, where sensor limitations in identifying material type and contamination make large scale sorting difficult, resulting in most plastic waste ending up in landfills.
• Designed for recycling facilities seeking a more reliable, automated method of sorting waste by material type.
• The finished system is a small scale robot that identifies waste dispensed onto a rotary table, loads it onto a hopper, navigates a fixed track, and deposits it into the correct designated bin.
• Main components include a Quanser robot equipped with ultraviolet, IR, colour, and LDR sensors, a robotic arm, a linear actuator driven hopper mechanism, and a Python program built in Quanser Interactive Labs.

Engineering Approach

• The hopper mechanism was designed around a linear actuator, with a rod connecting the actuator to the hopper's baseplate so that forward actuation pushes the rod and rotates the hopper to release the waste.
• Pins secured both ends of the rod, with small cylindrical features holding each pin in place to keep the mechanism stable during operation.
• The physical mechanism and its components, including the rod, slider, and pins, were modeled in Autodesk Inventor before fabrication.
• The full system was programmed in Python within Quanser Interactive Labs, where waste was dispensed at random onto a rotary table and identified by material type.
• A robotic arm was programmed to load up to three items onto the hopper, with the combined weight enforced not to exceed 90 grams before each loading cycle.
• Bin locations were defined using coordinate values within the virtual environment, which the program referenced to navigate the robot and position it correctly before dumping.
• An onboard IR sensor kept the robot following its fixed path, while an ultrasonic sensor measured distance to position the robot accurately in front of each bin prior to disposal.

Outcome

• The system functioned as a fully automated small scale sorting robot, correctly identifying waste by material type before directing it to the appropriate bin.
• The robotic arm and hopper mechanism successfully loaded, transported, and released waste into its designated bin within the enforced weight limit.
• The project illustrated a viable small scale model for automated sorting, supporting the broader goal of increasing recycling efficiency at facilities limited by manual sorting capacity.

Project Overview

• Addresses the need for a system that can safely transfer surgical tool containers into an autoclave for sterilization without manual handling.
• Designed for use in a simulated sterilization workflow, where containers of varying sizes need to be delivered to the correct designated location.
• The finished system is a robotic arm that identifies each container, transports it to its assigned bin, and either opens a drawer or places the container on top depending on its size.
• Main components include a Quanser robotic arm, a Python program built in Quanser Interactive Labs, and a container designed to hold a flat clamp tool, featuring a rod and cap securing mechanism and ventilation windows to support sterilization.

Engineering Approach

• The program was structured into modular functions, each responsible for a distinct task including bin location lookup, end effector movement, gripper control, and drawer operation.
• Each of the six container IDs, split evenly between small and large containers, was mapped to a specific coordinate location identified through trial and error in the virtual environment.
• The arm was programmed to return to a home position before moving to any specified location, ensuring consistent motion between pickups and drop offs.
• Separate logic controlled the gripper and the appropriate autoclave drawer based on a flag value, with the drawer only triggered for large containers.
• Every action taken by the arm was gated by a sensor emulator reading, checked against a threshold of 0.5, before the corresponding step was allowed to execute.
• The overall program tied these components together, randomly selecting a container, moving to pick it up, closing the gripper, navigating to the correct bin, opening the drawer if needed, and releasing the container before returning home.
• This cycle repeated until all six containers, spawned in random order without repetition, had been successfully transferred.

Outcome

• The program successfully identified, transferred, and deposited all six containers into their correct autoclave locations without manual intervention.
• Large containers correctly triggered their designated drawer to open, while small containers were consistently placed on top of the bin.
• The final deliverable was a fully functional robotic arm program paired with a container design that held the flat clamp securely and stably throughout the transfer and sterilization process.