Autonomous Navigation for Micro Aerial Vehicles

Abstract

In this thesis we have experimented in building a Micro Aerial Vehicle navigation system with completely open source components. The system consists of a DIY quadrotor with an opensource flightcontroller and a onboard raspberry pi which communicates with the ground station laptop over wifi. The sensors used are a monocular camera and an Inertial Measurement Unit (IMU).The flight controller communicates with the raspberry pi over usb. The monocular camera is connected to the raspberry pi, which communicates with the ground station laptop over wifi. Thus resource heavy algorithms like for example the Visual SLAM algorithm used could be shifted to the ground station. This was done to demonstrate the feasibility of building an autonomous navigation controller using only open source components. Also this is to facilitate further work on autonomous navigation and path planning using the developed system.
On the software side ROS (Robot Operating System) was used on both the on board raspberry pi and the ground station laptop. This facilitated the use a distributed computing paradigm with the processing heavy Visual SLAM being done on the ground station and the control commands being relayed to the flight controller through the raspberry pi.
Thus, a cheap open source MAV platform can be developed with open source software and hardware. Advanced state of the art SLAM and control algorithms can be tested and developed with little requirement of high end products. We using the above setup demonstrate the feasibility of using an single camera in conjunction with an IMU to localize the MAV in an unknown GPS constrained environment and use it to compute required controller and navigational commands for autonomous navigation.