Autonomous Indoor Navigation through Quadcopter in GPS denied environment

The use of Unmanned Aerial Vehicles (or UAVs) has seen an increased use in the across the world. UAVs are preferred for missions that carry a high risk for human life or require endurance that is beyond human capabilities. UAVs can either be controlled remotely by human operators o

Project Title

Autonomous Indoor Navigation through Quadcopter in GPS denied environment

Project Area of Specialization

Artificial Intelligence

Project Summary

The use of Unmanned Aerial Vehicles (or UAVs) has seen an increased use in the across the world. UAVs are preferred for missions that carry a high risk for human life or require endurance that is beyond human capabilities.

UAVs can either be controlled remotely by human operators or they can have their own autonomous decision-making power using Artificial Intelligence (AI) algorithms. 

Using AI to control UAVs extends their mission profiles and range of operations. This project is a research initiative towards development of Autonomous UAVs.

In the first phase, we are using Machine Learning and Deep Learning algorithms to make UAVs aware of their location and make real time maps of their environment. All this is being done by utilizing data from a limited number of sensors and processing it using AI based algorithms.

In the next phase we plan to use this information for autonomous navigation of UAVs in GPS-denied environments. This will give UAVs the capability to survey unknown and hazardous environments without using support from GPS.

Project Objectives

The objectives achieved through this project will be that our Quadcopter will be able to to do simultaneous localization and mapping (SLAM) of the indoor environment with only single monocular camera. Normally SLAM is performed with the help of other sensors like lidar, Depth sensors like Microsoft Kinect etc. which have their own limitations, whereas doing SLAM with monocular camera is really challenging. Furthermore, quadcopter will be able to navigate its path through previously unknown environment autonomously. We will be doing this in indoor environment where there is normally weak GPS signals and high demand of accuracy in terms of localization. Thus, we will be catering for one of the biggest challenges in indoor environment.

Project Implementation Method

We will be achieving our goals using RGB-D SLAM which is actually for Microsoft Kinect which is a depth sensor. We are going to implement this using monocular camera which is going to be our achievement because it’s not done before. To run RGB-D SLAM, it requires RGB image and depth image. We are going to use deep learning algorithms to do depth prediction through single images and give it to RGB-D SLAM, which will provide us with map of the surrounding and localization of quadcopter. Using this two information we are going to rung navigation algorithms which will enable the quadcopter to navigate autonomously.

Benefits of the Project

• This project could help us in security and surveillance of indoor environments.
• It could also be of Intel gathering aid to security personnel.
• It could help us in search and rescue in disaster relief operation.

Technical Details of Final Deliverable

The Project deliverables are as follows:

1. Development of ground station for communication with the quadcopter.
2. Depth prediction through convolutional neural networks.
3. Implementation of SLAM.
4. Implementation of Navigaiton algorithm in simulator.
5. Implementation of Navigaiton algorithm in  lab environment in real time.

Final Deliverable of the Project

HW/SW integrated system

Type of Industry

Security

Technologies

Artificial Intelligence(AI)

Sustainable Development Goals

Quality Education

Required Resources

If you need this project, please contact me on contact@adikhanofficial.com