DESIGN OF LOCALIZATION SYSTEM BASED ON PARTICLE FILTER ALGORITHM FOR MOBILE SOCCER ROBOT USING ENCODERS, COMPASS, AND OMNIDIRECTIONAL VISION SENSOR
Robocup is an international robot competition that was founded in 1996 with the aim of promoting research on robotics and artificial intelligence through an interesting activity but with difficult challenges. One of the categories in robocup is Middle Size League (MSL), a mobile soccer robot competi...
Saved in:
| Main Author: | |
|---|---|
| Format: | Final Project |
| Language: | Indonesia |
| Online Access: | https://digilib.itb.ac.id/gdl/view/43251 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Institution: | Institut Teknologi Bandung |
| Language: | Indonesia |
| Summary: | Robocup is an international robot competition that was founded in 1996 with the aim of promoting research on robotics and artificial intelligence through an interesting activity but with difficult challenges. One of the categories in robocup is Middle Size League (MSL), a mobile soccer robot competition. Navigation capability is needed by each robot in order to be able to make movements and strategies automatically. Therefore, the robot must be able to do self-localization. The methods that can be used for localization generally consist of Local Localization-System and a Global Localization-System. Local Localization-System uses sensors based on inertial sensors in the form of encoders, therefore it is prone to accumulation of reading errors. Whereas Global Localization-System utilizes sensors based on absolute sensors so that it has a long enough sample time.
The particle filter algorithm is designed as a technique for combining both inertial and absolute sensor data to overcome the problems of Local Localization-System and Global Localization-System. This algorithm can be implemented on wheeled soccer robots that have nonlinear kinematics models and nonparametric uncertainty distributions. Based on the results of particle filter simulation, the parameter that provides the best performance is by using M = 645 particles and n = 60 detected line points. This value is used in real-time testing. The results of particle filter implementation show better accuracy performance with an average particle filter error of 0.16 meters compared to Local Localization-System with 0.56 meters. The particle filter algorithm also has a sample time of about 2 times faster than the sample time required by image sensor processing.
Keywords: mobile soccer robot, Local and Global Localization System, particle filter, particle numbers, detected lines numbers
|
|---|