Wind-Tolerant Precision Landing of a Quadrotor on a Moving Platform

1. Designed a wind-resistant tracking controller for a drone, capable of compensating wind speeds up to 10m/s.

2. Implemented MPC for iterative optimal trajectory generation leading to smooth landing with maximum thrust efficiency.

Skills: Non-linear Control, Path Planning, Lyapunov Theory, MPC, ROS, Gazebo, C++ 

Report

Planning and Control of a Robotic Manipulator to Capture Space Debris

Designed and simulated a quintic polynomial trajectory planner and an inverse dynamics controller for a robotic manipulator to capture space debris, achieving position errors below 5% for a 25 kg debris mass at 10 m/s.

Skills: Motion Planning, Control Systems, MATLAB 

Term Paper

Omnidirectional Thrust Vectoring DRONE

Designed a cascaded PID controller for a modular over-actuated aerial vehicle built on quad-copters.

Skills: Control Systems, LQR, PID, Crazyflie cf2.0, Python, Simulink, Simscape, Solidworks

Design, Fabrication, and Controller Development of a 5-DoF Serial Manipulator 

Skills: Solidworks, MALTAB, PID, Serial Manipulator Kinematics

Modelled a 5-DoF serial manipulator (P4R) using Solidworks, fabricated using 3D printing, solved for forward and inverse kinematics using algebraic analysis and tuned motor PID controllers for the manipulator; used for precise pick & place operations. 

GitHub | Video

International AERIAL ROBotics competition 2017

1. Worked with an interdisciplinary team of 15 students to design and fabricate the drone hardware, and develop the software for the competition, resulting in our team winning the "Best Innovative Design Award" for our system. 

2. Implemented the mission planner state machine as a C++ ROS package to herd a swarm of ground robots towards a target location by physically interacting with them using a drone. 

3. Built a color and contour-based detector for a swarm of ground robots and tracked their motion using Lucas-Kanade optical flow and Extended Kalman Filter (EKF). 

4. Designed a novel localization algorithm that used given visual markers in the arena. This algorithm was computationally less expensive than optimization-based SLAM algorithms and relied on only a single camera.

Skills: SLAM, Motion-Planning, Object Tracking, State Estimation, Lucas-Kanade Optical Flow, ROS, C++

Report | GitHub

International micro air vehicle competition (indoor), 2019 

1. Designed a warehouse inventory management module for Parrot Bebop2 that detected QR codes using ZBar while traversing through the warehouse aisles and used tesseract-ocr to match the package QR code with the alphanumeric code on the respective bin. 

2. Built a shelf cavity detector using the rect-detect library and then created a disparity map with libELAS to find the corresponding cavity. 

Skills: Computer Vision, OCR, ROS, Python