Self-Balancing Robot for Painting
Authors
Olesen, Victor Højriis ; Christensen, Nikolai Lund ; Lau, Martin Michael
Term
4. term
Education
Publication year
2024
Abstract
Dette speciale undersøger analyse og styring af en mobil hjulbaseret inverteret pendulplatform (MWIP) som grundlag for en selvbalancerende malerobot. Udgangspunktet er et industrielt behov fra Turf Tank for at fjerne bageste svinghjul i linjemarkeringsrobotter uden at gå på kompromis med nøjagtighed. Målet er at udvikle mekaniske og elektriske matematiske modeller og anvende dem til at designe robuste styreløsninger, der kan regulere pitch, yaw og hastighed under varierende last og ydre påvirkninger. En foranalyse af en eksisterende prototype identificerer begrænsninger i MCU’ens regnekraft og I/O samt støj i IMU-målinger på grund af montagen; IMU’en udskiftes, og platformen benytter to PMSM-navmotorer drevet af en ODrive 3.6. Den mekaniske del modelleres i tre dimensioner via Lagrange-mekanik under antagelse af plan, jævn overflade og uden sloshing i maletanken. Motor og inverter modelleres i det roterende dq-referenceramme med Clarke/Park-transformationer, og feltorienteret styring (FOC) anvendes til strøm- og momentregulering. Det samlede system linearisers omkring opret stilstand, hvorefter en fuldtilstands LQR-regulator udformes og evalueres i simulationer. Eksperimentelt valideres hal-sensorerne og FOC-strømsløjfen gennem både lineære og ikke-lineære simuleringer samt forsøg på prototypen; motorparametre identificeres via optimering for at matche model og målinger. Styringen implementeres på en Teensy MCU og afprøves under gentagelige laboratorieforhold. De viste afsnit rapporterer validering af sensorer og strømsløjfe, parameteridentifikation samt laboratorietest af den implementerede LQR-styring; detaljerede præstationsmålinger er ikke angivet i uddraget.
This thesis investigates the analysis and control of a mobile wheeled inverted pendulum (MWIP) as a platform for a self-balancing painting robot. The work is motivated by an industrial need from Turf Tank to remove rear caster wheels in line-marking robots without sacrificing accuracy. The objective is to develop mechanical and electrical mathematical models and use them to design robust control that regulates pitch, yaw, and speed under varying loads and external disturbances. A preliminary analysis of an existing prototype identified limitations in MCU processing and I/O as well as IMU noise due to mounting; the IMU was replaced, and the platform uses two PMSM hub motors driven by an ODrive 3.6. The mechanical dynamics are modeled in three dimensions via Lagrangian mechanics, assuming flat, even surfaces and neglecting paint sloshing. The motor and inverter are modeled in the rotating dq frame using Clarke/Park transformations, with field-oriented control (FOC) for current and torque regulation. The combined system is linearized around upright standstill, and a full-state LQR controller is designed and evaluated in simulations. Experiments validate the motor Hall sensors and the FOC current loop through both linear and nonlinear simulations and hardware tests; motor parameters are identified via optimization to match model and measurements. The controller is implemented on a Teensy MCU and tested under repeatable laboratory conditions. The excerpt reports validation of sensors and current control, parameter identification, and lab testing of the implemented LQR controller; detailed performance metrics are not provided in the shown section.
[This summary has been generated with the help of AI directly from the project (PDF)]
Documents