UNSCENTED KALMAN FILTER WITH NONLINEAR OBSERVER FOR AUTOMATED GUIDED VEHICLE LOCALIZATION
Automated Guided Vehicle (AGV) is a type of vehicle used to transport objects from one place to another automatically. AGV can be used in container ports to increase efficiency. One of the factors to enable AGV perform well is having an accurate and global localization. To obtain global localization...
Saved in:
| Main Author: | |
|---|---|
| Format: | Theses |
| Language: | Indonesia |
| Online Access: | https://digilib.itb.ac.id/gdl/view/65326 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Institution: | Institut Teknologi Bandung |
| Language: | Indonesia |
| Summary: | Automated Guided Vehicle (AGV) is a type of vehicle used to transport objects from one place to another automatically. AGV can be used in container ports to increase efficiency. One of the factors to enable AGV perform well is having an accurate and global localization. To obtain global localization, AGV requires a Global Navigation Satellite System (GNSS) with a low sampling frequency. Sensor data in vehicles such as velocity sensors and steering angle sensors have a high sampling frequency and can determine locations by dead reckoning but measurement errors accumulate. Those sensors can be combined to improve measurement accuracy with a high sampling frequency. This technique is called sensor fusion. One of the sensor fusion methods is the Unscented Kalman Filter (UKF) which can handle non-linear systems and remain stable. The performance of the UKF algorithm can be improved by using a non-linear observer. This study focuses on improving the performance of UKF by adding a non-linear observer to the sensor fusion algorithm for the AGV localization. In algorithm design, non-linear observer is designed based on car-like vehicle dynamics using Lyapunov technique to obtain stable non-linear observer. The estimated state variables from the non-linear observers were then used to correct the UKF prediction results and continued with UKF estimation. The whole algorithm is made in the form of a simulation and implemented on the AGV prototype. Non-linear observation parameters and UKF were optimized using Particle Swarm Optimization (PSO). Sensor measurements are obtained with GNSS differential, velocity sensor, and steering angle sensor. The implementation results show that the developed algorithm has good results with a measurement error in RMSE of 0.0045 m for position and 0.00063 rad for orientation while the UKF algorithm has a measurement error in RMSE of 0.0078 m for position and 0.00123 rad for orientation.
|
|---|