Below is a selection of some of my class projects from undergrad and grad school.
Nonlinear Control of a Quadcopter
CDS 233 | Nonlinear Control | Caltech | Spring 2022
In this project nonlinear control of a quadcopter in 3D space was studied. The performance of various controllers such as a feedback linearizing controller, a min-norm controller, and an adaptive sliding mode controller in the presence of modeling accuracies were tested.
Analysis of a Double Pendulum System
CDS 232 | Nonlinear Dynamics | Caltech | Winter 2022
In this project the dynamics of a double pendulum system were investigated. The chaotic system's sensitivity to perturbations and initial conditions were studied and stability analysis were performed.
Towards Dynamic RRT: Local Planning
ME 133B | Robotics | Caltech | Winter 2022
In this project we studied how to use the RRT algorithm to find a path between some starting point and some goal in presence of fixed and varying obstacles. During the movement of the agent unknown obstacles show up and the agent has to replan to get to the desired goal.
Collaborators: Gilbert Bahati, Sergio Esteban
Sliding Attitude and Altitude Control of Tiltrotor Bi-Copters
2.152 | Nonlinear Control | MIT | Spring 2020
In this project a sliding mode controller was designed to stabilize a bi-copter in the presence of modelling errors. The controller was also made adaptive and its it performance in the presence of changing inertial parameters was studied.
Adaptive Control of a Quadcopter
2.153 | Adaptive Control | MIT | Fall 2019
In this project we compared the performance between a linear controller and a nonlinear adaptive controller for a quadcopter. We studied the controller performances under various scenarios such as during a motor failure and in the presence of a changing mass. We also used a Parrot Drone to perform hardware experiments.
Collaborators: Bernhard Græsdal
2D Robot Localization By Using a Simple RGB Camera
6.869 | Advances in Computer Vision | MIT | Fall 2019
In this project we presented a method to perform 2D position and orientation estimation of a robot by using a RGB camera, a CNN, and recursive least squares regression. The CNN was trained using transfer learning where the network was provided a small dataset of images with known positions orientations which were obtained by driving around in the toy world.
Collaborators: Aksel Danielsen