Using Wi-Fi and other sensors to navigate Turtlebot in indoor Environment – Research & Development
PATENT UNDER PROCESS FOR THE PROTOCOL DEVELOPED AS PART OF THIS PROJECT MENTIONED BELOW IN THE INTRODUCTION
I worked on this project in a group of two with my friend.
Wi-Fi has been widely accepted as an alternative method when GNSS and GPS have difficulties in providing position estimation, especially in indoor environments. Due to the properties of Wi-Fi signals, they can be used for indoor localisation purposes. The method of Wi-Fi fingerprinting is the most commonly used indoor localisation technique.
This thesis explores and implements indoor localisation and navigation of the Turtlebot autonomous agent using Wi-Fi fingerprinting, SLAM mapping methods and traditional robotic sensors to analyse the potential for a commercial implementation. These techniques provide indoor localisation accuracy to within 2m which is reasonable for indoor commercial applications where a robot’s location is required to be known. Some of these applications involve indoor surveillance, vacuum cleaning, mobile nursing and rehabilitation robots just to name a few.
In addition, the potential of a novel protocol to provide remote access to an autonomous agent via the internet without directly dealing with IP addresses whilst utilizing email is explored and implemented. This protocol is tested on a ROS operated autonomous agent to analyse its performance and effectiveness. The average delay of this protocol being 1.8ms proves effective for its purpose. The protocol is able to replace existing hardware such was WLAN chips resulting in cost-effectiveness and ease of IP management.
The physical limitations of GPS and GNSS such as requiring a direct line of sight between the receiver and transmitter makes it unsuitable for indoor environments. Hence, alternative resources such as Wi-Fi have been widely used for indoor positioning and localization. Wi-Fi provides local wireless access to a fixed network that is cheap, widely deployed and whose coverage is increasing at a swift rate. Due to the large scale deployments of wireless access points, Wi-Fi technology has become a promising technique to provide indoor localisation. A large amount of attention has been provided to Wi-Fi based localization approaches based on RSSI fingerprints. This approach provides a relatively high level of accuracy in indoor environments given the multipath propagation of signals.
Successful localisation and navigation of robots indoors has already been achieved through the use of laser scanners, image processing, sonar and even the Earth’s geomagnetic field. Due to the increasing prevalence of Wi-Fi and considerable amount of work done on localisation using RSSI, This Project implements and explores Wi-Fi RSSI based localisation on an autonomous agent (Turtlebot).
Using Wi-Fi fingerprinting and SLAM mapping techniques, the Turtlebot is expected to navigate to a user provided goal whilst localising itself in the SLAM map and verifying its location using Wi-Fi whilst maintaining a reasonable level of accuracy (to within 3m). This would provide a good indication of the potential for using Wi-Fi localisation in commercial and industrial based robots where position information is critical.
As the world tends towards the “internet of things” where a commodity is eventually connected to the internet, the ability to remotely access such devices becomes possible. The internet of things requires devices be assigned IP addresses in-order to connect to the internet. As these devices move networks, their IP addresses change which makes remote access slightly complicated as keeping track of the IP address becomes challenging. In order to simplify this, a novel protocol which allows a remote connection to be established (without dealing with IP addresses) with an autonomous agent regardless of location was developed and implemented.
This project used 3 different Laser Scanners to research accuracy and time taken to build a MAP
- Microsoft Kinect
- SICK Laser Scanner
- Hokuyu Laser Scanner