Autonomous Robotic Arm

(June 2017 - March 2018)
[ Video ] [ Report ]

Analyzed forward Kinematics of 3 DOF SARA robotic arm using DH parameters and designed the CAD model on SolidWorks studied the strength of the structure using FEA. Tuned the PID controller of the arm to reach the desired location and LabVIEW was used as backend software.

Language: Embedded C | Platform: SolidWorks, LabVIEW

Gesture Controled Wheelchair

(July 2019 - June 2019)
[ Report ]

Interfaced Arduino Uno with GY61 accelerometer sensor which has ADXL-335 as a sensing element, that can measure acceleration up to a minimum of +/- 3 g and gives the voltage linearly proportional to the acceleration. According to sensor values, L298N motor driver is used to control all four DC Motors mounted on wheelchair.

Language: C | Platform: Arduino

Gesture Controled Robotic Arm

(January 2017 - April 2017)
[ Report ]

Designed minimum embedded system board using EDA tools like Eagle, which get the X, Y and Z-Axis angles from an inertia sensor (MPU 6050) located on the user’s hand. According to the sensor value a SCARA based pick and place robotic arm is driven.

Language: Embedded C | Platform: Atmel Studios

Electromagnetic Pulse Generator

(August 2016 - October 2016)
[ Report ]

Designed and carried out the circuit simulation of various passive components using LTspice. I was capable to successfully generate 150V surge at primary coil using capacitors as an energy storage bank.

Platform: LTspice

Fire Alarm

(January 2014 - April 2014)

Germanium Diode was used as a heat-sensing element in the circuit. As the temperature near diode increases the diode reverse resistance drops increasing voltage at the non-inverting terminal of the operational amplifier which in turn on the external speaker and detect fire.

Platform: LTspice

TurtleBot Path Tracking using PID Controller

(May 2020 - August 2020) [ Video ] [ Report ]

Designed point-to-point and multi-point PID controller for turtlebot and simulated its performance on ROS gazebo. Deployed PID controller on actual turtlebot hardware and compared simulation results with actual hardware performance.

Language: C++ | Platform: ROS, Gazebo

Trajectory generation methods for serial manipulators

(February 2020 -May 2020) [ Video ] [ Report ]

Analyzed of two different 3-DoF serial manipulators, specificly cartesian (PPP) & articulated (RRR) with respect to different trajectory generation methods like Point-to-Point trajectory, continuous path like circle or sinusoidal wave trajectory. Compared energy consumed by the manipulator for different trajectory interpolation methods.

Platform: MATLAB

3D Object Detection using KITTI Dataset

(February 2020 - May 2020) [ Report ]

Performed 3D object detection on the KITTI dataset with the goal of reduced computation time and memory. Experimented with using SqueezeDet in the Frustum PointNet architecture for 2D detection instead of a fine-tuned Fast R-CNN

Language: Python | Platform: Google Cloud Platform

Indoor 3D Mapping using RGB-D Camera

(February 2020 - May 2020.) [ Video ]

The goal of the project was to implement Real-Time Appearance-Based Mapping of the indoor environment using RGB-D Camera on turtlebot3. Successfully generated robust 2D and 3D maps for an unknown environment using Li-DAR, IMU sensors and RGB-D camera on turtlebot3.

Language: C++ | Platform: ROS,Gazebo

Distracted Driver Detection

(September 2019 - December 2019) [ Report ]

Implementing an API that identifies the distracted driver and triggers the onboard alarm which helps to focus on driving. Designed and trained a different multi-class classification algorithm like SVM, Decision Trees, Random Forests and CNN’s capable of detecting 10 different human distractions based on driver’s activity with an average accuracy of ~ 92%

Language: Python

Detection and Recognition of Traffic Signals

(September 2019 - December 2019) [ Video ] [ Report ]

Extracted image feature’s using SHIFT, SURF, Hough Transform, and dynamic thresholding to detect traffic signal in a frame. Integrated Deep Neural Network based object detection method YOLOv3 for robust detection and SVM for status recognition.

Language: Python

djs-racing 04

(March 2017 - May 2018)
[ Video ] [ Specifications ]

Improvising on previous designs this season I designed a custom embedded board which serve as a Power Distribution Module for the car, which also has a feature to control the “Paddle Controlled High-Tech Servo Motor” which was used to actuate the clutch. Designed a Data Logger and Telemetry system using Race Capture Pro MK3 which collects data from various sensors on the car and use that data to improvise its performance of the car.

djs-racing 03.v2

(March 2016 - Feb 2017)
[ Video ] [ Specifications ] [ Report ]

Designed an industrial-grade Electrical Wiring Loom which was compatible with Custom programmable Performance Electronics PE3 ECU. Developed Electronic Solenoid Shifter system with ignition interrupts for fast gear shifting. Effectively developed a closed-loop Drag Reduction System for Aerodynamic package which gets the feedback from brake pressure sensor and steering angle position

Snake Game Application

[ Git Repo ]

Capstone Project which utilizes the core concepts from object-oriented programming, memory management, and concurrency. Once the game starts it creates objects, then the game continues to loop through each component and simultaneously grabs input from the user. There are three different levels in which the user can start with Easy, Medium, and Hard.

Concurrent Traffic Simulation

[ Git Repo ]

Built a multithreaded traffic simulator using a real urban map. Each vehicle runs on a separate thread and managed intersections using traffic lights to facilitate traffic flow and avoid collisions. Visualization can be seen on github repository.

Memory Management Chatbot

[ Git Repo ]

Implemented Chatbot using modern C++ memory management techniques such as smart pointers and move semantics. The knowledge base was loaded from a text file, a knowledge graph representation was created in computer memory. After a user query has been sent to the chatbot, the Levenshtein distance is used to identify the most probable answer.

Process Monitor

[ Git Repo ]

The system monitor shows what’s the current process running on your computer. Processes id, CPU and memory usage, etc. Object-oriented C++ was the backbone to build a Linux system monitor similar to the widely used htop application.

Route Planner

[ Git Repo ]

Used OpenStreetMap data and the IO2D visualization library to build a route planner that finds a path between two points on a real-world map using the A-star algorithm.