Christian Villarreal

Autonomous Navigation System with Vision, Machine Learning and Control

Final project of the "Intelligent Robotics" undergraduate course where Closed-Loop PID and Image-Based Visual Servoing (IBVS) Control, Deep Learning, Classical Computer Vision, and Embedded Systems learnings are applied to a differential-drive robot named PuzzleBot provided by Manchester Robotics Ltd. capable of navigating autonomously in a structured environment. This was made possible by the guidance and teachings of professors Dr. Luis Alberto Muñoz Ubando, Mario Guillermo Martínez Guerrero, and Arturo Eduardo Cerón López.

🎯 Key features:

• Line tracking with homography & Bird’s Eye View (BEV) transformation, powered by CLAHE, filtering, and morphological operations.
• Traffic light & sign detection with YOLOv10 + TensorRT at >100 FPS on Jetson Nano, fine-tuned for robust performance.
• Intersection handling via classical vision, Finite State Machine (FSM), and dynamic waypoint generation.
• Hybrid architecture: Jetson Nano (ROS 2 + high-level control) + ESP32 Hackerboard (microROS + low-level control).
• Modular ROS 2 implementation in Python/C++, with real-time visualization in RViz & OpenCV.

🎥 Demo highlights:

• Robot’s physical & electronic design.
• Distributed system architecture.
• Full autonomous run: line following, sign recognition, traffic light handling.
• Real-time visualization tools.

Multi-Robot Control and State Monitoring for Swarm Experiments

Project developed with the guidance and assistance of professor in control systems Dr. Luis Guerrero Bonilla at the Multi-Robotic Systems Laboratory (MRSL) at the Tecnológico de Monterrey campus from the city of Monterrey. implements a MATLAB-based interface for automatic discovery, connection, and control of multiple mobile robots (Puzzlebots) in a ROS2 environment. It connects to Vicon motion capture topics for real-time position feedback, establishes ROS2 publishers for wheel velocity commands, and organizes all communication into a MATLAB object for streamlined multi-robot operation.

🎯 Key features:

• Vicon topics associated with PuzzleBots are scanned for real message traffic while preventing wasted connections to inactive topics and subscribers and publishers are dynamically created to receive Vicon-based pose data and send wheel velocity commands.
• Any number of PuzzleBots are supported with the required communication channels, operating within a specified DDS Domain ID.
• This experiment tested the ability of PuzzleBots to autonomously reach an initial position and perform for a defined number of iterations navigation within a circle gradient.
• The use cases of this experiment include the control of a fleet of differential drive robots in formation or distributed tasks, navigation testing leveraging precise motion capture for closed-loop control experiments, and a robust base for implementing localization, path planning, and control algorithms.

Dexterous Manipulation of 9 D.O.F Robotic Hand

Final project of the Kinematics module for dexterous manipulation in a three-fingered robotic hand with 9 D.O.F.

• Applies the Fixed-Point Rolling-Sliding motion planning algorithm for dexterous manipulation in a three-fingered robotic hand with three degrees of freedom each.
• The algorithm considers kinematic constraints, grip stability, and compliant movements, enabling precise object manipulation.
• The implementation included mathematical analysis, the definition of kinematic equations, the simulation of the 9 D.O.F hand, validation done with examples, and an analysis of future improvements for performance optimization.
• The results demonstrate the viability of the algorithm, suggesting improvements for future applications in advanced robotics.

To learn more about the Fixed-Point Rolling-Sliding Algorithm implemented in this project, check out the following resource: https://ieeexplore.ieee.org/document/525326

Prototype Tractor Dual-Mode Waypoint Navigation System

Main project of the undergraduate course "Design of Advanced Embedded Systems" which implements Design and Analysis of Algorithms, Digital Signal Processing, Shared-Memory Architecture, and Communication Interfaces. This project was made in collaboration with John Deere as a training partner. This was also made possible by the guidance and teachings of professors Dr. Alfonso Ávila Ortega and Raúl Peña Ortega.

• The project consists of a dual-mode waypoint navigation system.
• In internal mode, the system utilizes encoder-based odometry and IMU orientation for waypoint following. Raw counts measured from wheel-mounted encoders, pre-processed by an Arduino Nano, and transmitted to the main MCU via CAN for distance conversion and pose data obtained from an MPU6050 IMU are fused.
• In external mode, it interfaces with John Deere's GPS for marker position-based navigation.
• The results demonstrate the viability of the algorithm, suggesting improvements for future applications in advanced robotics.
• The system involves a tricycle-style vehicle representative of a John Deere tractor prototype. It employs a NUCLEO-H745ZI-Q STM32 development board to execute motion control with synchronized front steering and differential rear drive.
• A shield-style PCB board was also designed to interface with the STM32 board to integrate various components and connection interfaces required for the tractor navigation system.

John Deere Tractor Automatic Transmission Controller Prototype

Main project for the undergraduate course "System Design on Chip" which applies STM32 Microcontrollers, Raspberry Pi Microprocessors, Computational Organization, Embedded Linux, and Real-Time Operating System (RTOS). This project was made in collaboration with John Deere as a training partner.

• The project aimed to simulate the automatic transmission controller of a driving John Deere tractor by integrating STM32 and Raspberry Pi into a prototype.
• An STM32 NUCLEO-F103RB microcontroller processed in a rate-monotonic scheduling (RMS) priority algorithm motor revolutions per minute and agricultural tractor velocity, marching and acceleration data to visualize in an LCD 16x2 screen and transfer to a Raspberry Pi 3 via UART in a Python developed graphical interface to show historical data.
• A potentiometer was used to manipulate the simulated tractor's acceleration and a matrix keypad to control braking and traction direction.
• The data is then transmitted to an LCD screen and to an interface made in Python inside the Raspberry Pi via UART.
• The program was divided into tasks whose required execution times were measured and priorities were determined based on their importance.

Signal Conditioning System and Analog Filters for Sensor Applications

Main project for the undergraduate course "Design of Electronic Circuits" under the guidance of the associated director of mechatronics and biomedical engineering at Tecnológico de Monterrey Dr. José Isabel Gómez-Quiñones and electronics and communications engineer Dr. Rodolfo Anaya-Zamora, which applies Circuit Design, Electronic Circuit Simulators like LTSpice to evaluate circuit models and designs, Analog Signal Processing, and Embedded Prototyiping to implement an electronic circuit prototype with expected characteristics and functionalities of the objective application.

• I designed and simulated a signal conditioning system in LTSpice and the real world. A voltage-to-voltage converter using two op-amps, resistors, and power supplies to linearize the non-linear output of an MQ-135 gas sensor was built.
• This sensor was connected to an Arduino UNO microcontroller to record min/max values when exposed to alcohol.
• Then, a 2nd-order Butterworth low-pass active filter was designed and validated through LTSpice before physical implementation to process the sensor signal, mapping the output voltage to a 0-100 scale for accurate readings.
• Finally, the system once built was tested with input frequencies of 50 Hz, 30 Hz, and 350 Hz, successfully filtering out the 350 Hz and beyond noise.
• Furthermore, I designed and simulated two different filters. First, a 3rd-order Butterworth high-pass filter in LTSpice at two different cutoff frequencies: 500 Hz and 15 kHz. Second, I developed a Chebyshev filter with specified parameters: Passband ripple (in dB), stopband attenuation, and cutoff frequencies for passband and stopband.

AquaNet IoT System for Pet Fish Care

Main project for the undergraduate course "Internet of Things Implementation" and was selected among many others to be presented in the Expo Ingenierías 2023 event held in Tecnológico de Monterrey in the city of Monterrey, Mexico.

• Fish care requires constant attention from pet fish owners: Monitoring pH levels, lighting, and food, among other things, are all aspects that owners must manage. It can be challenging or even stressful for some.
• An STM32 NUCLEO-F103RB microcontroller processed in a rate-monotonic scheduling (RMS) priority algorithm motor revolutions per minute and agricultural tractor velocity, marching and acceleration data to visualize in an LCD 16x2 screen and transfer to a Raspberry Pi 3 via UART in a Python developed graphical interface to show historical data.
• Issues such as constant owner intervention and the complexity of temperature, pH, and feeding regulations are part of the lives of fish owners.
• AquaNet is a comprehensive system designed to regulate and maintain optimal environmental conditions in pet fish aquariums.
• This system uses NodeMCU using C++ for sensor and actuator activation, React and Vite for a web application and an SQL database to record and visualize sensor and actuator data in real time, and Postman API for building and using APIs to communicate both instruments.
• Our project was praised by our hardware and software professors Claudia M. Solís Garza and Yesenia Sarahi Rivera Leyva as the best in the group, with highlights including the LED lights, 3D-printed containers, and the overall successful execution of a challenging water-related model.
• The functional website and prepared presentation were noted as strong points and our team work was reflected in the high-quality result.

RoBorregos Candidates 2022 Tournament Robot

This tournament had the objective to find outstanding students to join the RoBorregos robotics team based on their performance in solving the challenge, share robotics knowledge within the community, and achieve both personal and professional growth for the applicants.

• Zone A had the objective of finding the end of the maze by passing through as many colored squares as possible and correctly identifying whether they are blue, red, or green through displays, sounds, movements or any other means.
• An STM32 NUCLEO-F103RB microcontroller processed in a rate-monotonic scheduling (RMS) priority algorithm motor revolutions per minute and agricultural tractor velocity, marching and acceleration data to visualize in an LCD 16x2 screen and transfer to a Raspberry Pi 3 via UART in a Python developed graphical interface to show historical data.
• Zone B consisted of classifying an object based on its color by moving it to an unloading zone corresponding to its color.
• Zone C required detecting an infrared ball while avoiding crossing white lines to reach it.
• The final zone, the Ramp, required the robot to climb the ramp and completely position itself on the upper platform.