ANFIS based navigation for HVAC robot with visual feedback /
Heat, Ventilation and Air Conditioning (HVAC) mobile robot is an integrated mechatronic system that comprises of mechanical, electronic and control system. In general, there exist a lot of problems which are contributed by each of the element embodied in the robot. However, in this project, only rob...
Saved in:
| Main Author: | |
|---|---|
| Format: | Thesis |
| Language: | English |
| Published: |
Kuala Lumpur :
Kulliyyah of Engineering, International Islamic University Malaysia,
2015
|
| Subjects: | |
| Online Access: | http://studentrepo.iium.edu.my/handle/123456789/4340 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Heat, Ventilation and Air Conditioning (HVAC) mobile robot is an integrated mechatronic system that comprises of mechanical, electronic and control system. In general, there exist a lot of problems which are contributed by each of the element embodied in the robot. However, in this project, only robot navigation is the main focus. Certainly, there exist a lot of navigation procedures and methods which are applied by available HVAC robots, be it in commercial or research institution, but they always come in hefty 'price', complicated system to build, high-end and expensive hardwares and decline in performance due to 'heavy' algorithm utilized. In addition to that, the system is not easily transferable from one platform to another, since they are mostly unique and need a lot of changes in hardware to fit with new system. Hence, our project does not only concentrates on applying navigational algorithm but how to juggle between these issues. The system is implemented using Adaptive Neural and Fuzzy Inference System (ANFIS) rather than complicated nonlinear controllers in which it will reduces controller's complexities. ANFIS is used due to its ability to integrate human experience into mathematical modeling to approximate nonlinear function while at the same time capturing both the benefit of fuzzy principle and neural network in a single framework. On hardware level, processing is being carried out separately in microcontrollers and a Central Processing Unit (CPU) of a laptop. Microcontrollers handle simple tasks (e.g. controlling actuators, acquisition of distance measurement, communicating with laptop, etc) while computationally intensive ANFIS navigation and image processing (blob analysis and thresholding) are done via laptop's GPU. This has not only enabled us to build a system using off- the-shelf components and reduce the cost greatly, but also lessen the computing load as 'resources' are split accordingly. Transferability is improved by using only two most applied sensors in literature; camera and ultrasonic sensor. In this project, navigation process is essentially managing 1) steering angle for front wheels of the mobile robot and 2) the angular velocity of the wheels. Performances on both criteria are evaluated by comparing these values against reference values where certain marginal values are expected at certain distance/position in order to successfully navigate the robot. To be able to perform accurately, effective range of detection for the ultrasonic sensor is 2cm to 300 cm and for camera is approximately 60cm. A duct opening (statistical information such as perimeter, centroid are known in prior) with the camera distance of 50cm is selected as a reference. Expected angular velocity and steering angle is calculated at that point and compared with the experimental values. Accuracy of angular velocity is evaluated at roughly 77.5% while steering angle is recorded at approximately 79.2%. Overall evaluations are based on comparison of control, mobility and navigational methods, structure and mechanism and algorithm complexities. |
|---|---|
| Physical Description: | xvi, 147 leaves : ill. ; 30cm. |
| Bibliography: | Includes bibliographical references (leaves 136-145) |