Reactive Navigation in Cluttered Indoor Environment for Autonomous MAVs

In recent years, Micro Aerial Vehicles (MAVs) have been a desired solution for indoor applications. However, most of today's approaches are restricted to piloted applications for lack of autonomy in unknown scenario. There, the MAV must independently handle both infrastructure and arbitrary obstacles. We present a navigation system for autonomous flight in cluttered indoor environments. First, we transfer the Elliptic Limit Cycle (ELC) approach for target-oriented obstacle avoidance to the dynamics of MAV flights. In terms of automated perceptive functionality, laser data clustering and parameter estimation on the fly are presented. Second, we develop refinements to simultaneously handle both isolated obstacles and connected infrastructure, which is a necessary condition for any indoor application. We evaluate the automated obstacle perception by means of real experimental data acquired in interior space. The guidance system's proof of concept is given based on realistic indoor environment software in the loop tests.