EECE5554 — Robotics Sensing and Navigation

Course lab repository for EECE5554 at Northeastern University. Each lab builds toward a complete sensor fusion stack using ROS 2 Humble.

Author: Adrian Ramirez (ramirez.adr@northeastern.edu)

---

Final Report (For the Class)

Magnetic Interference and Calibration: A Comparative Analysis of Magnetometers Malia Howe, Shayda Moezzi, Adrian Ramirez — EECE5554, April 2026

Capstone report for the course. Benchmarked four magnetometer calibration algorithms (MinMax, Magcal, Scalar EKF, and Levenberg-Marquardt) across an industrial-grade VN-100 and a consumer-grade iPhone 11 under controlled hard iron, soft iron, and combined distortion conditions at 7 cm and 30 cm separation distances.

Key findings:

• The iPhone 11 outperformed the VN-100 (13–19% vs 22–34% mean magnitude error), attributed to the VN-100's Vector Processing Engine violating algorithm independence assumptions
• Magcal and Levenberg-Marquardt produced the most consistent calibration output (~0.92 µT and ~0.94 µT std dev)
• MinMax was fastest but most sensitive to distortion intensity
• Scalar EKF achieved competitive accuracy but had the highest standard deviation and slowest runtime

Full report: EECE5554_Final_Report.pdf

---

Repository Structure

Directory	Topic
LAB0	ROS 2 Publisher / Subscriber basics
LAB1	GPS Driver (SparkFun NEO-M9N)
gnss (LAB2)	RTK GNSS Driver & Accuracy Analysis
LAB3	IMU Driver (VectorNav VN-100)
LAB4	GPS + IMU Driver Integration
LAB5	Camera Calibration & Image Mosaicking
Magnetometer_Comparison_Project	Magnetometer Calibration Algorithm Benchmarking

---

Prerequisites

• ROS 2 Humble
• Python 3
• colcon build tool
• Jupyter (for analysis notebooks)

source /opt/ros/humble/setup.bash

---

LAB0

Simple ROS 2 pub/sub demo.

A minimal talker/listener pair using std_msgs/String. Entry point for verifying the ROS 2 environment.

cd LAB0
colcon build
source install/setup.bash
ros2 run simple_chatter talker
ros2 run simple_chatter listener

---

LAB1

GPS Driver for the SparkFun NEO-M9N.

Parses NMEA GPGGA sentences over serial, converts to decimal degrees and UTM, and publishes a custom GpsMsg.

Custom message: gps_driver_msgs/GpsMsg

• latitude, longitude, altitude, hdop
• utm_easting, utm_northing, utm_zone, utm_letter
• utc_time

cd LAB1
colcon build
source install/setup.bash
ros2 launch gps_driver launch.py port:=/dev/ttyACM0

Launch argument: port (default /dev/ttyACM0)

Analysis notebook: LAB1/src/analysis.ipynb

---

LAB3

IMU Driver for the VectorNav VN-100.

Communicates over serial at 40 Hz. Publishes accelerometer, gyroscope, and magnetometer data as a combined custom message. Includes Allan deviation analysis for noise characterization.

Custom message: imu_msg/IMUmsg

• sensor_msgs/Imu (accel + gyro)
• sensor_msgs/MagneticField
• Raw device string

cd LAB3
colcon build
source install/setup.bash
ros2 launch imu_driver imu_launch.py port:=/dev/ttyUSB0

Record data:

bash LAB3/src/record_bag.sh

Analysis notebook: LAB3/src/analysis/analysis.ipynb

---

LAB4

GPS + IMU Driver Integration.

Brings up both drivers simultaneously via a single launch file. Collects synchronized GPS and IMU data for analysis, including trajectory comparison, position error, yaw from a complementary filter, and magnetometer data.

cd LAB4
colcon build
source install/setup.bash
ros2 launch lab4_bringup lab4.launch.py gps_port:=/dev/ttyACM0 imu_port:=/dev/ttyUSB0

Record data:

bash LAB4/src/record_bag.sh

Analysis notebook: LAB4/src/analysis/analysis.ipynb

Key outputs: trajectory comparison, position error, yaw from complementary filter vs. IMU.

---

LAB5

Camera Calibration & Panoramic Image Stitching.

Uses a checkerboard pattern to compute the intrinsic matrix K and lens distortion coefficients. Applies rectification, Harris corner detection, and homography-based mosaicking to construct panoramic images.

Tools: MATLAB (harris.m, mosaic.m)

Calibration data and results: LAB5/images/chessboards/

Mural stitching results: LAB5/images/Murals/

---

GNSS

RTK GNSS Driver for centimeter-level positioning.

Parses GNGGA sentences from an RTK-capable receiver. Publishes fix quality, HDOP, and full UTM coordinates. Datasets include open-sky, occluded, walking, and vehicle scenarios for accuracy comparison.

Custom message: rtk_driver_msgs/CustomRTK

• latitude, longitude, altitude, hdop, fix_quality
• utm_easting, utm_northing, utm_zone, utm_letter
• utc_time, raw GNGGA sentence

cd gnss
colcon build
source install/setup.bash
ros2 launch rtk_driver rtk_driver_launch.py

Analysis notebook: gnss/analysis.ipynb

---

Magnetometer Comparison Project

Magnetometer Calibration Algorithm Benchmarking (Course Final Project).

Benchmarks four calibration algorithms — MinMax, Magcal, Scalar EKF, and Levenberg-Marquardt — across two sensors: an industrial-grade VectorNav VN-100 and a consumer-grade iPhone 11 (via Sensor Logger). Data was collected under hard iron, soft iron, and combined distortion conditions at 7 cm and 30 cm separation distances.

Tools: MATLAB (masterAnalysis.m, functions/) and Python (magnetometer_analysis.ipynb)

Calibration algorithms:

• functions/runMinMax.m — MinMax (fastest, most sensitive to distortion intensity)
• functions/runMagcal.m — MATLAB Magcal (most consistent output)
• functions/runKalmanFilter.m — Scalar EKF (competitive accuracy, highest std dev and slowest runtime)
• Levenberg-Marquardt (implemented in masterAnalysis.m)

Data: Magnetometer_Comparison_Project/data/ — CSV files for VN-100 (vn*) and iPhone (SLA*) across all distortion conditions.

Full report: EECE5554_Final_Report.pdf

---

Data Format

All rosbag2 data uses the .db3 format with metadata.yaml. Topics recorded:

Topic	Type
/gps	gps_driver_msgs/GpsMsg
/imu	imu_msg/IMUmsg
/rtk	rtk_driver_msgs/CustomRTK