PROJECTS
2019
2018
2018
2018
2016
2015
2015
2014
Place Recognition and Factor Graph Localization for Mobile Robots using Google Indoor StreetView
​
Supervisors - Dr. Oscar De Silva, Dr. G.K.I.Mann, Dr.Raymond.G.Gosine
​
Designing an algorithm to correct the drifted trajectory of a Visual Inertial Navigation System (VINS). We use Factor Graph based optimization techniques instead of filter based methods. The Smoothing and Mapping Library by Georgia Institute of Technology, GTSAM is used for factor graph based optimization. Visual feedback from a monocular camera is used as the feedback apart from other measurements such as odometry or VINS data. The system is completely modular is implemented in ROS.
Place Recognition using Google Street View
​
Supervisors - Dr. Oscar De Silva, Dr.Andrew Vardy
​
Designing an algorithm to correct the drifted trajectory of a Visual Inertial Navigation System (VINS). We use Factor Graph based optimization techniques instead of filter based methods. The Smoothing and Mapping Library by Georgia Institute of Technology, GTSAM is used for factor graph based optimization. Visual feedback from a monocular camera is used as the feedback apart from other measurements such as odometry or VINS data.
Implementation of a Gaussian Filter in FPGA
​
Supervisors - Prof. Andrew House
​
Implementing a 3 x 3 Gaussian filter on a Altera DE1- SoC FPGA using VHDL. The Gaussian blur process was divided into four sub-tasks : multiplication, addition, division and truncation. VHDL entities were created for these sub-tasks and the controller. Inbuilt divider and RAM entities were used for division and data storage.
Place Recognition using Google Street View
​
Supervisors - Dr. Oscar De Silva, Dr. Mohommed Shehata
​
Implementing a topological place recognition systems using Google Street View and Visual Bag-of-Words(BoW). A method proposed in a research paper was used. The objective was to predict the node closest to a query image input to the system. Simulations generated satisfactory results.
Implementation of a vision based obstacle avoidance and safety system for a quadcopter
​
Supervisors - Dr. D.H.S.Mithripala
​
Group project involving robotics,mathematical modelling, Machine Learning (Clustering), computer vision, Visual Odometry, algorithms and hardware implementation. A vision based system that extracts and matches feature points between consecutive images obtained using an image sensor, estimates the distances to each feature point, clusters the point cloud in a 3-D space, identify obstacles within the point cloud, determine the closest obstacle and avoid it. The project included a re-configuring the PX4Flow image sensor to obtain the binary images and rotation matrix of the system. The system was implemented on Raspberry Pi 3 model B using Python and OpenCV. Re-configuring the PX4Flow image sensor was carried out using C.
Was awarded the Prof. E.F.Bartholomuez prize which is awarded for the best project involving Engineering Mathematics
Mathematical modelling,simulation and implementation of a vision based object following mobile robot
​
Supervisors - Dr.D.H.S.Mithripala
​
Individual project on robotics and mathematical modelling. A vision based robot that follows a blue colour object by identifying it purely with the aid of vision.Uses OpenCV to process images with the aid of a Raspberry Pi as the central processing unit and an Arduino as the controller.
Identification of an unknown material
​
Supervisors - Dr. C.K. Pathirana
​
A debris was assigned to our team.Objective was to predict the material ,material composition, mode of failure and the object to which the debris belonged to using known experimental procedures and analysis
Design of mechanism to board a wheelchair onto a train
​
Supervisors - Dr. P.B.B.Boyagoda
​
A mechanical system which allows a person on a wheel chair to embark and disembark himself onto and out of a train with minimal effort and assistance