Gareth A. Harrison

Projects

13 projects, 2023 to now
TurtleBot in the test arena at Toyohashi, surrounded by cones and boxes carrying AprilTag landmarks

Toyohashi University of Technology · Mar 2026 - May 2026

TurtleBot Autonomous Navigation

Autonomous navigation and mapping for a TurtleBot in cluttered indoor environments, combining global path planning with visual servoing and an AprilTag landmark infrastructure to defeat odometry drift.

What I did
  • Architected a reactive state machine for autonomous perimeter exploration, using windowed LiDAR data for dynamic obstacle evasion and trap recovery.
  • Engineered a custom visual-servoing controller for final docking, reaching 0.4 mm error at a 15 cm staging range and bypassing the limits of the standard ROS navigation stack.
  • Ran intrinsic camera calibration for a Raspberry Pi Camera V2 and designed a tiered landmark system to mitigate long-term dead-reckoning drift.
  • ROS Noetic
  • Python
  • Visual Servoing
  • LiDAR SLAM
  • AprilTag

Repository →

3D-printed strawberry-picking arm prototype on the bench, reaching for strawberries beside a laptop

BINUS ASO School of Engineering · Sep 2025 - Jan 2026

Strawberry Picking Robotic Arm

Autonomous robotic arm for precision strawberry harvesting, taken from cardboard mockup to a fully 3D-printed working prototype with vision-driven actuation.

What I did
  • Engineered the manipulator structure as detailed 3D CAD models and assemblies in SolidWorks.
  • Trained an object detection model to localize strawberries and generate bounding boxes for the control system.
  • Quantized the model to INT8 TensorFlow Lite for Raspberry Pi 4B deployment, targeting 20-30 FPS.
  • YOLOv8
  • Computer Vision
  • SolidWorks
  • TensorFlow Lite
  • Raspberry Pi

Repository →

IndoHoaxDetector running on Hugging Face Spaces, classifying a pasted Indonesian article as HOAX

BINUS ASO School of Engineering · Sep 2025 - Jan 2026

IndoHoaxDetector

End-to-end NLP pipeline classifying Indonesian news as FAKTA or HOAX, benchmarked across five classifiers and deployed as a live Gradio application.

What I did
  • Built an Indonesian text normalization pipeline over 62,972 samples using regex and the Sastrawi stemmer, cutting vocabulary by 63.6% to 45,678 terms and average document length from 312 to 156 words.
  • Benchmarked Logistic Regression, Naive Bayes, Random Forest, SVM and IndoBERT under stratified 5-fold cross-validation with hyperparameter grid sweeps.
  • Audited error distributions and feature coefficients, then serialized the pipeline with joblib and deployed it to Hugging Face Spaces.
  • NLP
  • Sastrawi
  • IndoBERT
  • Scikit-learn
  • Gradio

Live demo →

Two AeroBASE members holding the HAYABUSA drone up outdoors

BINUS ASO School of Engineering (AeroBASE) · Oct 2024 - Oct 2025

HAYABUSA, KRTI 2025

Autonomous drone carrying three LiDAR sensors for obstacle avoidance and mapping, plus a dual-camera setup: front-facing for gate detection and navigation, bottom-facing for object detection and mission tasks.

What I did
  • Developed and integrated drone simulation environments in Gazebo using ROS for flight control and mission testing.
  • Built and tested object detection models for drone perception, improving target recognition accuracy in simulation.
  • ROS
  • Gazebo
  • Object Detection
  • LiDAR
The KARASU quadcopter on an outdoor court during a flight trial

BINUS ASO School of Engineering (AeroBASE) · Jun 2023 - Oct 2024

KARASU (カラス), KRTI 2024

First-generation UAV for aerial surveillance and image capture, and AeroBASE's debut entry in the Kontes Robot Terbang Indonesia. The team's first end-to-end effort across airframe design, software integration, and flight testing.

What I did
  • Debugged and optimized ROS packages controlling UAV navigation, sensor data, and communication modules.
  • Conducted multiple flight trials validating obstacle avoidance and autonomous decision-making, improving stability and responsiveness.
  • ROS
  • UAV
  • Flight Testing
Gazebo simulation of the KRTI 2025 drone on the ground plane with the entity tree open

BINUS ASO School of Engineering (AeroBASE)

KRTI 2025 Gazebo Simulation Assets

Drone model and simulation world for the AeroBASE KRTI 2025 entry, built for Gazebo Sim and ROS 2 Jazzy.

What I did
  • Modelled a hexacopter airframe as SDF/URDF with frame, propeller, and camera-mount meshes.
  • Built a Gazebo Sim world for visual and physics-based flight testing.
  • Gazebo Sim
  • ROS 2
  • SDF/URDF
  • Simulation

Repository →

Tomato leaf with early blight lesions outlined in red by the Plantesa detector, reading Brown 7.06%

BINUS ASO School of Engineering · Feb 2025 - Jun 2025

PLANTESA: IoT & ML Plant Incubation System

Smart plant incubation system combining environmental sensors, an ESP32 and Raspberry Pi 4B for real-time monitoring and automated irrigation, with a CNN detecting tomato leaf disease at an early stage.

What I did
  • Developed the full IoT system integrating temperature, humidity, NPK, water level, and ultrasonic sensors.
  • Implemented cloud data processing via Firebase to control irrigation and surface readings to a web front end.
  • Applied a CNN to webcam images for early tomato leaf disease detection.
  • IoT
  • CNN
  • Firebase
  • ESP32
  • Raspberry Pi

Repository →

Raw waveform of a happy-voice recording plotted in MATLAB

BINUS ASO School of Engineering · Sep 2024 - Jan 2025

Depression Detection via Voice Analysis

Non-invasive diagnostic system extracting MFCC features from speech and classifying them with an SVM, reaching 99.46% on angry, 96.8% on sad, and 89.47% on happy emotional speech.

What I did
  • Collected and curated 50+ voice samples across multiple emotional categories.
  • Preprocessed audio with noise reduction, segmentation, and normalization for consistency.
  • Extracted MFCC features capturing pitch, tone, and rhythm variation linked to depressive speech, then trained the SVM classifier.
  • Digital Signal Processing
  • SVM
  • MFCC
  • Audio Processing
Three line charts side by side: the IHSG composite index from 1995, the USD/IDR exchange rate from 2001, and the Antam gold price per gram from 2010

Nov 2025 - present

Indonesian Financial Market Datasets

Three open time-series datasets on Indonesian markets - the IHSG composite index, USD/IDR exchange rates, and Antam gold prices - published in parallel on Hugging Face and Kaggle and refreshed daily by an automated pipeline. Around 19,000 rows in total, all MIT-licensed.

What I did
  • Built a stateless daily ETL job that pulls the current CSVs from Hugging Face, appends new market rows from Yahoo Finance, and pushes identical files to both platforms, no-opping on market-closed days.
  • Routed the fetch through a SOCKS5 tunnel to a residential egress host after the data provider blocked the hosting VPS, with three retries before giving up.
  • Made the job fail loudly rather than degrade, after a silently partial run lost a full trading day, and added a stats collector tracking per-dataset downloads and views over time.
  • Data Engineering
  • Python
  • ETL
  • Hugging Face
  • Kaggle

Datasets →

Line chart of Antam gold price in IDR from 2010 to 2026

Gold Price Prediction Models

Comparison of regression and gradient-boosting models on Antam historical gold prices, from raw scrape through to evaluation.

What I did
  • Built the acquisition pipeline converting scraped Antam price pages into a clean CSV series.
  • Trained and compared linear regression against XGBoost on the resulting dataset.
  • Evaluated with ROC and precision-recall curves, a confusion matrix, and feature importance.
  • Python
  • XGBoost
  • Scikit-learn
  • Pandas

Repository →

Rebound hammer test on a cured cement block with plastic content

Total Panel Beton · Jun 2025 - Sep 2025

Plastic Shredder & Vibration Mixer

Equipment supporting material efficiency and sustainability in production: a shredder reducing plastic waste to reusable particles, and a vibration mixer improving composite uniformity.

What I did
  • Designed and fabricated shredder blades to process plastic waste for reuse in cement-based composites.
  • Repaired and optimized the vibration table for stable operation during mixing and testing.
  • Formulated and tested cement mixtures with varying plastic content for strength, durability, and production feasibility.
  • Carried out with Johan Kim and Stivan Delon Sahertian.
  • Mechanical Design
  • Fabrication
  • Materials Testing
SMART Parking System model with toy cars in the bays and the ultrasonic barrier post

BINUS ASO School of Engineering · Feb 2024 - Jul 2024

SMART Parking System

Automated parking system using PIR sensors, infrared sensors, servo motors, and LED indicators to monitor availability and regulate access, cutting search time, fuel use, and emissions. Aligned with the UN Sustainable Development Goals on industry, innovation, and sustainable cities.

What I did
  • Designed the electronic circuits in EasyEDA, integrating sensors and actuators.
  • Tested circuit performance across operating conditions to verify reliability.
  • Optimized and implemented the final design for stable operation.
  • Embedded Systems
  • IoT
  • EasyEDA
  • Circuit Design
Automatic fish feeder unit with its LCD lit, mounted beside the aquarium

BINUS ASO School of Engineering · Feb 2024 - Jul 2024

Automatic Fish Feeding System

Arduino-based feeder using ultrasonic sensors, servo motors, and a fan mechanism to distribute food evenly, with an RTC module for precise scheduling and an LCD showing remaining capacity.

What I did
  • Designed the circuits in EasyEDA, integrating ultrasonic sensors, servo motors, and an LCD display.
  • Programmed the Arduino to automate scheduled feeding via a Real-Time Clock module.
  • Assembled and tested the system for reliable distribution in aquarium conditions.
  • Arduino
  • Microcontrollers
  • EasyEDA
  • Automation

Repository →