← selected work

Lead Architect · Engineering Capstone

Robotic Service Dog

2025–2026

An autonomous quadruped that follows and monitors Alzheimer's patients.

  • ROS2
  • C++
  • Python
  • OpenCV
  • Inverse Kinematics
  • visual servoing
  • Docker
platform
8-DOF quadruped
control loop
visual servoing

/ the problem

Assistive robotics for dementia care needs reliable person-following and health monitoring on a legged platform — many interacting subsystems that must stay in sync in real time.

/ approach

  • Distributed ROS2 stack integrating perception, planning, and control
  • Closed-loop visual servoing: OpenCV state estimates → velocity commands → adaptive human tracking
  • Developed a 3-DOF inverse-kinematics solver for the legs; froze the hip joint in the final build to run 2 active DOF per leg (8-DOF quadruped)
  • Custom ROS2 QoS profiles for deterministic multi-node communication
  • C++ camera + actuator drivers integrated with the full stack; Dockerized team dev environment

/ results

  • End-to-end person-following on a physical quadruped
  • Led the system architecture across the capstone team

/ architecture

Perception (OpenCV)
Planning
Control (IK, servo)

on a ROS2 graph with custom QoS.

/ media

Closed-loop visual servoing — live AprilTag detection (left) driving a real-time pose/orientation estimate (right)
Follow-controller running in simulation — the quadruped ('snoopy') tracking a target across RViz and Gazebo
A 3D-printed 3-DOF leg actuated by its servos — the final build freezes the hip joint to run 2 active DOF per leg

/ links