My research aims to develop the hardware and control algorithm for the flight control of an at-scale insect robot . The UW Robofly is a 75 mg flapping wing insect scale robot. A low weight camera of 26 mg is designed and fabricated to go on-board the Robofly and provide visual control of its motions. As a demonstration of the visual feedback, image classification is performed using the images captured by the camera and training a simple CNN. The CNN classifies flowers and predators and makes the insect robot move forward and backward respectively. The next step is to use optical flow values measured from the camera for visual servoing of the Robofly.
Worked on unsupervised clustering techniques for labeling paths in app code based graph data. Analyzed supervised classification models for validating the paths. Worked on sequence modeling with Recurrent neural networks.
Completed a projected titled 'Haptic Rendering of Planar Rigid Body Dynamics and Kinematics' with Novint Falcon Haptic device. The project was funded by Mitacs under the Globalink research internship program.
Guide: Dr. Daniela Constantinescu, Asst. Professor at Department of Mechanical Engineering,University of Victoria.
Completed a projected titled 'Comparison between white line following and interrupt position control AGV navigation' with firebird V robotic platform.
Guide: Dr. K. Panneerselvam, Asst. Professor at Department of Production Engineering,NITT.[project report].
Underwent In-plant training.[certificate]
The project tries to analyse three mobile robotic methods which can be used for friction-stir welding.The motion of the robots in the three methods are discussed and the various advantages and disadvantages of these methods are determined in the project.
Developed a linear control system for the 2-D flight control of an insect robot. The LTI system is designed and tested for controllability and observability. Open and closed loop control is implemented on the LTI and the non linear system and their behaviours are tested.
The food retrieving behaviour of the Saharan Silver Desert ant is mimicked and Simulated in MATLAB.
The project aims to design a LQR estimator/controller for the optimal control of flight of an insect robot.
The project aims to compare the standard Kalman filter and the Extended Kalman filter for estimating the states of an inverted pendulum. The results show that Extended Kalman filters work better for systems with non linear dynamics.
The project detects Harris corner points, tracks their positions in subsequent frames using optical flow algorithms and uses a logistic regression to learn if a particular point is from a moving object or the stationary background. It also clusters moving objects close to each other to identify moving regions.
The project aims to detect the normalized center coordinates of an object in an image. A CNN regression was used on a dataset containing phones on different floor patterns and an accuracy of 70 % on new test images.
1. Participated in e-Yantra robotics competition-2015 conducted by IIT-Bombay and developed a cargo alignment robot with firebird V robotic platform. [video 1][video2][certificate].
2. Participated in HOVER ONE organized in Pragyan 2013, & 2014 - The International TechnoManagement festival of NITT.[certificate]
3. First in the event ARESTO MOMENTUM [certificate]
4. First in the event GREEN AEROMODELLING [certificate]
5. Second in the event PROJECT X organized during Synergy 2013 - National Level Technical Symposium of Mechanical Engg Department, NITT. [certificate]
6. Attended Touch screen workshop conducted by EI systems and completed course on ImageProcessing and surface computing. [certificate]