Research Work

Recent Research

Identifying the Arrangement of Bottles inside a Cooler (Industrial Project)

This project was implemented on a client requirement. We were given 120 images of bottle arrangement
inside a cooler belonging to four classes namely, good, bad, worst, and not rated. First, we augmented these
images to get 330 images per class. Imagenet trained EfficientNetB1 was used as the base model without
top which was identified as the best model from a preliminary study. We added an average pooling layer, a
flatten layer, a dense layer with a relu, a dropout layer and another dense layer with a softmax to extend the
base model. Finally, the model was trained for 20 epochs while the base model was frozen. The validation
loss and accuracy were 0.0413 and 0.9924 respectively. This model will be used with an Android application
for the client.

An Autonomously Guided Differential Drive Robot Base Using Asus Xtion Pro Live

I'm doing this research with collaborating eVision Microsystems Pvt Ltd. In this research, we are implementing a wheeled robot base that can autonomously navigate in an indoor environment. It has two hardware units. A separate processing unit and a movable. These two components comprised ROS kinetic, Asus Xtion Pro, Arduino and Raspberry Pi that communicated in between. The robot communicates with the central location PC through a Private Wi-Fi connection. It is very profitable and easily reprogrammable. Further, it can be easily customized to obtain the desired size. Finally, the robot will help to make users' storage keeping tasks easier.

Development of an RGB-D sensor-based Office Assistant Robot (OABot) System

This project, I'm doing for my M.Phil degree under the supervision of Dr. W. K. I. L. Wanniarachchi, attached to the Department of Physics at the University of Sri Jayewardenepura, Sri Lanka. In this research, we have selected free and open-source software and inexpensive sensors and materials to build a low cost but efficient robot. The OABot comprised of two units: one that processes sensor data and another which is the navigating robot having sensors and actuators. We have used Robot Operating System (ROS) indigo as the ground operating system on Ubuntu 14.04 LTE primary operating system. iRobot Create 2 was used as the basic robot, and a structure was built to carry a mini Laptop, PrimeSense 3D vision sensor and RGB camera. TurtleBot software packages, openCV software packages and Python libraries were used to implement this OABot. Then, we built an Android application to provide commands to the OABot by its users. So, the users are able to control the robot using either voice or virtual keys on the Android application.

Past Research

Preparation of a Ceramic Floor Tile Body Containing Sodium Carbonate as a Strengthening Agent, for Lanka Floor Tile PLC

This research, I did for my bachelor's degree under the supervision of Dr. C. L. Ranatunga, attached to the Department of Physics at the University of Sri Jayewardenepura, Sri Lanka. Through this research, we found a suitable ratio (wt%) of Sodium Carbonate with sodium Silicate as a strengthening agent of the low plastic ceramic body for LankaFloor Tiles PLC. It would be highly recommended since it appreciably reduces ruptures of unfired ceramic tiles in their dry state during handling and transporting, and enables the pre-firing steps of the tile factory to be rationalized and productivity to be increased.