The project involves integrating nonlinear model predictive control (NMPC) with adaptive control barrier functions (CBFs) for reactive collision avoidance in agile robots.

The project involves implementing the MSCKF paper for VIO.

The project involves exploring different deep learning architectures for accurate and robust VIO.

This project involves developing visualization similar to Tesla's dashboard using different deep learning model for object detection and localization.

This project involves implementing NeRF to synthesize novel views by optimizing a deep neural network to model scene geometry and appearance from sparse input images.

This project involves developing a SfM pipeline for 3D reconstruction using feature matching, motion recovery, and bundle adjustment to estimate camera positions and 3D points.

This project involves developing simulation with differentiable physics to train neural network control policies for high-speed aerial navigation using only depth map.

This project involves developing a neural network-based heuristic for reducing collision-checking and clearance estimation time significantly compared to traditional geometric methods.

This project compares DDPG, PPO, TD3, & DQN for autonomous TurtleBot3 navigation.

This project involves developing a Non-Linear Model Predictive Control for Quadrotor for precise trajectory tracking and also keeping point of interest in camera's frame.

This project involves engineering a Reachability-Guided RRT algorithm for efficient motion planning for robot with dynamics constraints.

This project involves implementing a Prioritized Double Deep Q-Network (DQN) algorithm to optimize performance in the Breakout game

This project involves building a U-Net, a specialized CNN, for semantic image segmentation on a self-driving car dataset to predict pixel-level labels for precise object recognition. The U-Net architecture ensures high-accuracy segmentation, critical for autonomous vehicle navigation and safety.

Trigger Word Detection with Deep Learning involves building a model to detect the word "activate" in audio streams, enabling devices to wake up upon hearing this trigger. The project includes synthesizing a speech dataset, processing audio into spectrograms, and training a neural network with convolutional and GRU layers to make accurate predictions.

Face Recognition Using Siamese Network leverages deep learning to accurately identify and verify individuals by mapping face images for precise differentiation between similar and dissimilar faces. This project addresses key challenges in face recognition, enhancing convenience across various applications.

This project involves building a U-Net, a specialized CNN, for semantic image segmentation on a self-driving car dataset to predict pixel-level labels for precise object recognition. The U-Net architecture ensures high-accuracy segmentation, critical for autonomous vehicle navigation and safety.

In this project, I utilized MobileNetV2 for transfer learning to make binary classifier of a custom dataset of Alpaca/Not Alpaca, augmenting the data, and fine-tuning the model for improved accuracy. The process involved adapting a pre-trained model, freezing layers, and optimizing the classifier for specific tasks.

Developing 3-wheel omnidirectional mobile robots capable of Cooperative Navigation as linear formation for pairs of robots and triangular formation for groups of three robots.

MATLAB-based model to simulate the attitude kinematics and dynamics of a satellite under the influence of thruster-generated external moments.

Developed an automated surveillance drone for accident, fire, and anomaly detection. Sent data to control station via GIS (Geographic Information System), used ROS Noetic for communication.

Made a 3 DoF robotic manipulator to perform tasks like pick & place, etc. via Alexa Voice commands.

Employed MATLAB to conduct a comprehensive analysis of unbalanced forces and moments within a V8 cross-plane engine configuration.

Developed a TurtleBot3-based surveillance system for intruder detection and video feedback using OpenCV and ROS.

Creating a Python wrapper for Pluto nano drone, enabling remote control and autonomous flight, by communiating using MSP (MultiWii Serial Protocol).

Performed finite element analysis of a tetrachiral auxetic stent with varying parameters to evaluate its mechanical behavior under different load conditions.