FlightControllerv1.0

vigil systems

v.i.g.i.l.: visual indoor guard & intrusion locator

an autonomous drone that patrols a building at night, so a guard doesn't have to walk it alone.

working name. the project was previously called port, then lumaway.

---

the problem

overnight security is risky and expensive. a guard patrolling a dark building alone can run into an intruder with no backup, and every overnight post is a recurring cost. fixed cameras and alarms can report that something happened, but they can't go and look.

vigil is built to fill that gap: a drone that checks the building for you, verifies what tripped an alarm, and gets a photo to the people who need it.

---

what vigil does

1. patrols on its own. each drone covers one floor. the building is mapped and divided into zones, and a state machine picks the next waypoint at random, weighted toward zones that haven't been visited recently. no fixed route means no predictable gaps.
2. sees in the dark. thermal imaging and lidar detect people, and the classifier separates a person from an animal or a misplaced object.
3. responds. on a likely intruder, the drone takes a photo, triggers the building's alarm system, and notifies the property manager.
4. stays supervised. a remote operator can watch and take over any drone at any time. guards keep their jobs, they just do them from somewhere safe.

indoor only. night only. version 1 uses sight only, with no audio. photos are taken on a trigger, not continuously, and there is no face matching.

---

why a drone

a drone can cover a whole floor quickly, reach places a fixed camera can't, and investigate an alarm before anyone is dispatched. one remote operator can supervise several drones.

---

the system

custom flight controller (hardware) a 4-layer custom pcb with two stm32 microcontrollers (stm32f405 for navigation, stm32f103 for i/o) running px4. onboard sensors: dual imus (icm-20689, icm-20602), a barometer (ms5611), and a magnetometer (ist8310), plus a power monitor and flash for black-box logging. designed in kicad and manufactured in-house.

companion computer (autonomy) an nvidia jetson nano runs patrol logic and person detection, and talks to the flight controller over mavlink.

indoor sensing thermal camera and lidar for detection, plus a gps-free position source for indoor flight.

response and remote operation photo capture, alarm-system integration, and an operator link for live view and manual override.

---

project goals

• stable flight on the custom flight controller
• gps-free indoor position hold
• reliable person detection in the dark, with animals and objects filtered out
• random patrol that covers every zone within a set time window
• photo capture and alarm trigger on detection
• end-to-end demo: patrol, detect, photograph, alert, remote takeover

---

status

in development. uw bothell electrical engineering capstone, 2026 to 2027.

• flight controller v1.0: built and assembled, bring-up in progress
• test airframe: assembly in progress
• sensing, patrol logic and alerts: planned

---

team

built by a four-person electrical engineering capstone team at the university of washington bothell, supervised by dr. hannachi.

• cris r. martinez hernandez
• abraham chen
• frank ma
• howard sin

---

built with

• stm32f405 / stm32f103 (flight controller mcus)
• px4 autopilot
• nvidia jetson nano (onboard compute)
• ros2 and mavlink
• opencv / python (perception)
• kicad (pcb design)
• platformio (firmware)

---

vigil systems: always watching, so nobody has to.