Fast Exploration Using Multirotors: Analysis, Planning, and Experimentation

Abstract

High speed flight with multirotor aerial vehicles is limited by constraints on size, sensing range, on-board computation, accelerations, and velocities. For robotic exploration and operation in unknown environments, guaranteeing collision-free operation and selecting informative sensing actions further increase complexity. To this end, this work presents three contributions. First, we analyze the rate of reduction of the entropy of the map for idealized scenarios considering constraints on dynamics and sensing for a multirotor vehicle. Second, we propose an action representation that accounts for platform dynamics and provides actions that are useful for rapid exploration based on the prior analysis. Third, we present a receding-horizon sampling-based planner that uses this action representation, maximizes information gain, and ensures safe operation at high speeds. Finally, we present extensive simulation experiments in a complex 3D environment that demonstrate the significance of action design, horizon length, and re-planning rate on exploration performance.