Singapore Space Challenge

Introduction

In October, the whole team participated in the Singapore Space Challenge (SSC) 2015 in the Unmanned Aerial Systems category. Teams had to conceptualize a system to perform inspection operations on aircraft using UAVs. UAVs are already used in many other industries and also recreation purposes. UAVs can be applied to the Maintenance, Repair and Overhaul tasks on aircraft allow for timely access to parts of an aircraft normally difficult to inspect. Furthermore, the mission specifications state that:

The aerial system can comprise of more than one UAV. Systems comprising of more than 2 UAVs will be favorably graded, provided this does not impact the practicality and viability of the concept and solutions.

We started planning our project based on 4 fundamental objectives: Safety, speed, reliability and ease of operation. Some issues with the mission parameters became immediately clear:

• There would be no GPS signals in a closed hangar where the inspection will likely take place.
• Hangars have much infrastructure and equipment inside that need to be avoided.

Since the mission require detection of defects on the surface of the aircraft as well as knowledge of the surrounding environment to avoid obstacles, we opted to heavily leverage the power of vision processing in our system.

The Dragonfly emerges

The Dragonfly is a quadcopter that uses the usual sensors like accelerometers, gyroscopes for stability as well as 2 vision based sensors. The positioning process is split into 2 primary stages, primary flight stabilization and vision based postioning.

• Primary flight stabilization utilizes the onboard accelerometers and gyroscopes to detect attitude, sending commands to the motors to keep the Dragonfly stable.
• Onboard stereo cameras take continuous video which is processed by the onboard computer, an Intel Edison, to provide data about objects in the environment. This data is used for collision avoidance and keeping the Dragonfly on track when navigating to a destination.

Primary flight stabilization requires fast and immediate response to small changes in the Dragonfly's attitude, therefore it needs to be processed onboard. Position data also requires a fast response to avoid obstacles, therefore still needs to be processed onboard. Mapping data is not vital to flight stabilization, therefore is offloaded to the ground station for greater power efficiency onboard.

Inspection of aircraft surface

The high resolution camera is mounted on a pitch axis gimbal so that it can tilt up and down when flying below or above the aircraft. Both the stereo camera data and the high res camera data is streamed to the ground station to be crossmatched and processed. On the ground station, a 3 dimensional map of the aircraft is constructed from this data.

The ground station then decides where the Dragonfly needs to go next, and sends the position data to the onboard computer. The constructed map is then compared to a saved, perfect model of the plane, therefore highlighting anomalies.

Swarm of Dragonflies?

Using multiple Dragonflies can speed up the scan and improve the quality of the final map. Every Dragonfly in use sends its vision data to the ground computer, where the data is matched. Using a swarm algorithm, the ground computer can split the work between the Dragonflies for a fast operation.

After Action Review

We also researched other more novel methods for scanning like microwave and laser scanning. However none of these came to fruition due to feasibility issues. We were referring to a paper about using microwave scanners to detect cracks in aircraft skin but it required the scanner to be on the surface it is scanning. Given more time (better planning), we could have came up with a solution to the surface problem, thus truly innovating.

In hindsight, we should have planned out our work better in the beginning to reduce the rush midway through to redo unfeasible ideas. The swarm portion of our system could also have been more integrated as it seems to be and is more of an afterthought.

We came in first runner up in our category, second to only another team from our school. Still I am proud of what we achieved and we would definitely recommend anyone to participate in the competition.