Automatic re-initialization and failure recovery for aggressive flight with a monocular vision-based quadrotor

Autonomous, vision-based quadrotor flight is widely regarded as a challenging perception and control problem since the accuracy of a flight maneuver is strongly influenced by the quality of the on-board state estimate. In addition, any vision-based state estimator can fail due to the lack of visual information in the scene or due to the loss of feature tracking after an aggressive maneuver. When this happens, the robot should automatically re-initialize the state estimate to maintain its autonomy and, thus, guarantee the safety for itself and the environment. In this paper, we present a system that enables a monocular-vision-based quadrotor to automatically recover from any unknown, initial attitude with significant velocity, such as after loss of visual tracking due to an aggressive maneuver. The recovery procedure consists of multiple stages, in which the quadrotor, first, stabilizes its attitude and altitude, then, re-initializes its visual state-estimation pipeline before stabilizing fully autonomously. To experimentally demonstrate the performance of our system, we aggressively throw the quadrotor in the air by hand and have it recover and stabilize all by itself. We chose this example as it simulates conditions similar to failure recovery during aggressive flight. Our system was able to recover successfully in several hundred throws in both indoor and outdoor environments.

I. Introduction

Autonomous Micro Aerial Vehicles (MAVs) will soon play a major role in remote inspection and search-and-rescue missions. In these applications, the MAVs will have to operate in unknown indoor and outdoor environments, which prevents them from relying on external positioning systems (e.g. GPS or motion-capture systems). A viable solution to maintain position tracking for lightweight MAVs is to use on-board cameras. Unfortunately, vision algorithms are prone to lose visual tracking during fast motions, e.g., when executing aggressive maneuvers, or under strong illumination changes that can occur when transitioning from dark to bright scenes. When visual tracking is lost, the vehicle typically has to descend and land in a partially open-loop maneuver, or a trained pilot has to take over control. To re-initialize the vision pipeline, manual procedures (by hand or remote control) are required by the operators, which renders re-initialization during flight very difficult or even impossible. Fig. 1:

Autonomous recovery after throwing the quadrotor by hand: (a) The quadrotor detects free fall and (b) starts to control its attitude to be horizontal. Once it is horizontal, (c) it first controls its vertical velocity and then, (d) its vertical position. The quadrotor uses its horizontal motion to initialize its visual-inertial state estimation and uses it (e) to first break its horizontal velocity and then (f) lock to the current position.