Research Projects

Automatic Shaving system with a Single Arm Panda Robot

This project aims at developing an automatic shaving system with a 7 DOF single arm robot. The tasks include detecting a human face with a realsense depth camera, picking up a shaver with the robot hand grippers, moving the arm to such a configuration that the shaver comes in contact with the face, sliding the shaver along the surface of the face with a certain amount of force to perform the shaving and finally placing back the shaver to the location from where it was picked up. The application shows how robotics can safely be used to perform household assistant tasks. The tasks and frameworks are:

• Pick-and-place, Hand-Eye calibration, Motion Planning, Manipulation,Obstacle Avoidance, Force Controller, Impedance Controller, Face Detection
• Programming Languages: Python and C++
• Frameworks: ROS, MoveIt, OpenCV, ViSP

Master Thesis

Data Driven Autonomous Driving using Bird's eye image and Convolutional Neural Networks

Classical autonomous driving uses images from front facing cameras to train a deep learning model to predict the steering angle of a car. This thesis proposes to use a bird's eye image instead to train the network for steering angle regression. The project was carried out in a simulator in Unity Engine

• Prepare a Surround View system (Bird's Eye view) of a car by implementing Computer Vision tasks: Homography, Perspective Transformation, Image Stitching
• Training a Deep Convolutional Neural Network (end-to-end learning) with the input Bird's Eye images to predict steering angle
• Frameworks: Tensorflow. Keras, OpenCV, Unity Game Engine
• Programming Languages: Python, C++

Background-Foreground Segmentation of images for Object Detection and Tracking

For detecting moving objects in images, background image consisting of static objects needs to be separated from the foreground image consisting of moving objects. Factors such as lightning changes in the environment, noises and disturbances cause a lot of challenges in the segmentation process.

• Review of existing Object Detection and Tracking algorithms and implementing some of them such as Gaussian Mixture Models, Motion Templates technique, Optical Flow
• Implementation of a novel Histogram and Mean based method for segmenting background image and detecting foreground moving objects
• Programming Languages: C++
• Frameworks: OpenCV, Rosbags
• Comparison of different methods in terms of efficiency and optimization of algorithms for better computational time and reduced memory
• Binary Classifier to label the outcomes and plot Receiver Operating Characteristic (ROC) curves

Robotics Project: Line Following, Obstacle Avoidance, Color Sorting, Goal Point Detection

To build an automated color sorter, line follower with obstacle avoidance, goal point detection using LEGO MINDSTORMS EV3 kit

• Obstacle avoidance with Ultrasonic sensors, line following and goal point detection with infrared sensors, implementing PID control strategy using Bricx Command Center for NXC programming
• Color sorting is carried out by modelling motors, encoders, sensors and actuators in SIMULINK and controlling the system with Simulink Stateflow models.

Position Control using ATMega128 Microcontroller

This project uses a Ball on an Inclined Plane model to control the position of the ball on the inclined plane by implementing a PID control algorithm

• Mathematical modelling of the system to determine the angular position of a servo motor arm
• Physical modelling of the plant, PID controller, low-pass filter, sampling and timer blocks, mux, etc. in Scilab and Xcos
• C Programming an ATMega128 MCU using Atmel Studio IDE for position and velocity control tasks
• Interrupt handling, timers, PWM (Pulse Width Modulation), Finite State Machines, UART serial interface, Analog-to-Digital converter

Bachelor Thesis

Design and Fabrication of a Pedal Powered Water Pump

This project aims at designing, fabricating and analysing a portable water pump powered by human pedal.

• Application areas: Small scale household and irrigation purposes, Remote locations where no electricity supply is available
• Design components: Pedal, assembly of Pulleys, Shafts and Belt Drives to run a Centrifugal Pump
• Advantages: Cost effective, affordable, eco friendly

Energy Harvesting using Piezoelectric Cantilever Beam

In this project, theoretical study of scavenging energy in the form of output voltage from mechanical vibrations using a piezoelectric cantilever beam is carried out

• Study of working of piezoelectric materials, types of cantilever beams
• Euler-Bernoulli beam is considered and governing equation of motion of the system is obtained using Lagrange's approach, Hamiltonian principle, Green's strain