Micro Air Vehicles (MAVs) are becoming ever smaller and lighter. This opens new areas of application, mainly because of the capability to access small spaces. A straightforward example of such an application is building inspection. In such a case there is in general no prior knowledge about the environment. Furthermore there will be no direct communication between MAV and operator due to signal blocking, and GPS positioning is not possible for the same reason. These restrictions require the MAV to perform several autonomous tasks: attitude control, obstacle avoidance, navigation etc.
Extremely light-weight MAVs are very limited in the amount of available onboard sensors and processing power. This project aims at developing an onboard system that provides the resources required for full autonomy. It should enable the MAV to perform the following tasks:
- determine safe directions of flight
- explore the environment efficiently and thoroughly
- find out how to enter new spaces to increase the exploration area
- find its way back to previously visited points
This system will be used for and tested on DelFly, a Flapping Wing MAV of less than 20 grams developed by TUDelft. The dimensions of this MAV demand a system that makes use of an extremely efficient combination of sensor types, data processing algorithms and control loops. This will lead to approaches that are more biologically inspired than current state-of-the-art avoidance and mapping methods. The main focus will be on using cameras and vision-based methods. The objective is to develop a vision-based system that can be carried onboard a light weight MAV (<20 grams) which enables fully autonomous operation in GPS-denied environments.
Publications