Optimized Implementation of LOAM for Indoor Environments.

Introduction:

LOAM is a state of the art algorithm developed for Lidar SLAM. Many major developments have taken place after the influx of LOAM and the next major developement was the Lego_LOAM which was explicitly developed for Outdoor and Environments and AGVs. In our case we implements the SLAM on an AGV so this was preferred over original LOAM but we needed to make approriate adjustents for the sake of indoor environments.

State Esitmation using VIO

Introduction:

Here we are working on a Micro Air Vechicle, which has a camera and an IMU sensor mounted on it and the MAV is driven . The whole project is composed of three different parts where in the first part we estimate the pose of MAV using a mat of April Tags present on the ground. In the later part we estimate the linear and angular velocity of the MAV using Optical Flow. In the final part we use Extended Kalman Filter where IMU data is considered for prediction and measurement is done using the results of vision.

Object Detection For Visually Impaired

Introduction:

The goal of this project to build a system where object detection can be performed from the smartphone camera and the person can hear the details of the object pointed by the finger i.e, its name and distance from themself.

Tracking of Lagrangian River Floating Sensor using UAV based on Reinforcement Learning

Introduction:

A novel method of Reinforcement Learning, Deep Deterministic Policy Gradient(DDPG) is used to tackle the problem of landing an autonomous drone on a moving platform with the use just simple data i.e, GPS coordinates and imu data of drone and the base. Many methods have been emloyed before to tackle these kind of problems making use of raw pixels which need a lot of computation power to train and also less accurate. In our work we have just used low level control variables which made the training both less computative and more accurate.

Live Stock Monitoring Based from Drone Footage.

Introduction:

In this project I have developed an approach that detects and counts Livestock from footage of a drone. Latest CNN techiniques and image processing techniques have been used to accomplish the task.