Vision-based obstacle detection and avoidance

— We consider autonomously flying a miniature co-axial helicopter in various indoor environments. The architecture allows developers to write navigation algorithms for aerial robots without worrying about the underlying control features. Our work aims to provide a helicopter with a robust stabilization controller, and an obstacle detection and avoidance feature. We stabilize the helicopter using a combination of Optical Flow and Sonar Sensor data. Obstacle Avoidance allows us to detect and avoid obstacles using a support-vector machine on multiple segments of a live image streamed from the helicopter. We test the architecture using corridor-following and stair-following algorithms written as plug-ins into the system.