CBL - Campus del Baix Llobregat

Projecte llegit

Títol: Plataforma de Pruebas para el Control de un Cuadricóptero


Estudiants que han llegit aquest projecte:


Director/a: LÓPEZ LAPEÑA, ÓSCAR

Departament: EEL

Títol: Plataforma de Pruebas para el Control de un Cuadricóptero

Data inici oferta: 03-02-2026     Data finalització oferta: 03-10-2026



Estudis d'assignació del projecte:
    GR ENG SIST AEROESP
Tipus: Individual
 
Lloc de realització: EETAC
 
Paraules clau:
Plataforma de pruebas, cuadricóptero, control PID, latencia del lazo, ESP-NOW
 
Descripció del contingut i pla d'activitats:
El presente proyecto aborda el diseño y desarrollo de una plataforma de ensayos para el control de un cuadricóptero en dos grados de libertad: movimiento vertical y alabeo (roll). La solución propuesta se estructura en dos subsistemas principales. Por un lado, una estructura mecánica que restringe el movimiento del vehículo a las dinámicas mencionadas; por otro, un subsistema electrónico formado por los sensores inerciales del dron y dos microcontroladores ESP32 con conectividad Wi-Fi.

Se implementará, mediante el entorno Arduino, el software de comunicaciones entre ambos ESP32 con el objetivo de transmitir y reproducir tanto las medidas sensoriales como las señales de control asociadas al accionamiento de los motores. Esta arquitectura permitirá desacoplar la adquisición de datos y la generación de consignas, facilitando la integración de un microcontrolador externo encargado de recibir las señales de los sensores y calcular las acciones de control necesarias para garantizar la estabilización del sistema en los dos modos de movimiento considerados.
 
Overview (resum en anglès):
This work develops a test platform for the control of a quadcopter, intended for teaching use. The rig restricts the motion of the drone to two degrees of freedom, altitude and roll, by means of a pulley with a counterweight and a pendulum with mechanical stops, so that the control can be tuned in a safe and repeatable environment. The starting point was a rig built by the supervisor of this work, which ran but never managed to stabilise the drone.

The architecture splits the work between three boards. One is carried on board the drone and reads the sensors and drives the motors, another acts as a base station and reproduces those sensors and actuators on a connector, and the third is the board the student programs, which closes both control loops. The first two communicate over radio using the ESP-NOW protocol. The student therefore interacts with the sensors and the actuators as if they were wired directly to his own circuit, without cables restricting the motion of the drone.

The development started with a diagnosis of the loop delay, the reason why the rig did not stabilise. The travel of the information across the three boards went from 57.5 milliseconds on average, with excursions up to 81.4, to a steady 20.7 milliseconds. The improvement came from several changes that add up, moving the heavy work out of the function that serves the radio and into a dedicated task, sending the packet as soon as the four control channels have been measured, doubling the actuation frequency from 50 to 100 hertz and choosing the radio channel. The reliability of the ultrasonic sensor was addressed next, generating its echo in hardware on the base station and compensating the effect of the tilt of the drone on the measured altitude. On the mechanical side, a support that centres the pulley and a new pendulum that limits roll to 15 degrees were designed and printed.

The rig was then characterised in SolidWorks and the transfer functions relating the control action to the altitude and to the roll angle were obtained, together with the stability margins of every set of gains tested. The part of the friction of the rails that can be linearised enters the model as an equivalent viscous damping adopted as a hypothesis. Against the step tests the model sits at an intermediate point, since it reproduces the shape of the response of both motions but does not fully describe the real behaviour of the platform, so it serves as a starting point for reaching a stable controller rather than as a prediction tool.

The outcome is a platform that flies with both loops closed and holds altitude and roll with no intervention, validated through step tests on the references. The remaining limit is the dry friction of the rails, which sets a floor on the error that the controller cannot lower on its own. The platform is ready for classroom use, where the student programs his own controller and sees on a real drone how it responds to the gains he chooses.


© CBLTIC Campus del Baix Llobregat - UPC