MPC control for the BlueROV2 - Theory and Implementation
Authors
Lipenitis, Andris ; Einarsson, Emil Már
Term
4. Term
Publication year
2020
Submitted on
2020-10-19
Pages
112
Abstract
MPC-regulatorer (Model Predictive Control) bruges ofte i procesanlæg, men har traditionelt krævet meget beregning og har derfor været mindre udbredte på små indlejrede systemer. MPC bruger en model til at forudsige systemets adfærd og vælge styresignaler. Efterhånden som indlejrede computere er blevet kraftigere, kan MPC anvendes i flere typer styringsopgaver. I dette speciale tilpasser vi både software og hardware på den kommercielle, open-source platform BlueROV2, så den bedre kan køre en MPC og et Kalman-filter. Et Kalman-filter kombinerer sensordata for at estimere fartøjets tilstand. Koden er skrevet i Python med en objektbaseret struktur, og vi sammenligner den originale opsætning med vores ændringer. For at implementere MPC modelleres det fjernstyrede undervandsfartøj (ROV) med 6 frihedsgrader (DoF), og modellen valideres. Til tilstandsfeedback anvender vi et Kalman-filter, der fletter målinger fra en inertimåleenhed (IMU) og Underwater Global Positioning System (UGPS). Til sidst sammenligner vi MPC-tilgangen med en Proportional–Integral–Derivative (PID) regulator og en Linear Quadratic Regulator (LQR) på BlueROV2.
Model Predictive Control (MPC) is widely used in process plants but has traditionally required heavy computation, making it less common on small embedded systems. MPC uses a model to predict system behavior and choose control inputs. As embedded computers become more powerful, MPC can be applied to a broader set of control tasks. This thesis adapts both the software and hardware of the commercially available, open-source BlueROV2 platform to better support running an MPC and a Kalman filter. A Kalman filter combines sensor data to estimate the vehicle’s state. The code is written in Python with an object-based design, and we compare the original setup with our modifications. To implement MPC, the remotely operated vehicle (ROV) is modeled with 6 degrees of freedom (DoF), and the model is validated. State feedback is provided by a Kalman filter that fuses measurements from an Inertial Measurement Unit (IMU) and an Underwater Global Positioning System (UGPS). Finally, we compare the MPC approach with a Proportional–Integral–Derivative (PID) controller and a Linear Quadratic Regulator (LQR) on BlueROV2.
[This abstract was generated with the help of AI]
Keywords
MPC ; ROV ; Kalman filter ; Sensor fusion ; Python
Documents