Safety is undoubtedly the most important requirement in moving robots applications, especially for micro aerial vehicles (MAVs). And both academia and industry have been working hard to equip drones with GPS-free self-localization, obstacle sensing, and autonomous navigation capabilities without prior information about the environments.

The fisheye cameras with ultra-wide filed-of-view (FOV) can provide more spherical coverage of the surrounding environment. A dual-fisheye omnidirectional visual-inertial navigation system (VINS) combine two fisheye cameras and an inertial measurement unit (IMU), cover the whole surroundings of the MAV. It is the minimum sensor suite lending the omnidirectional perception with lightweight and small footprint.

In this paper, we show that it is possible...[ Read more ]

Safety is undoubtedly the most important requirement in moving robots applications, especially for micro aerial vehicles (MAVs). And both academia and industry have been working hard to equip drones with GPS-free self-localization, obstacle sensing, and autonomous navigation capabilities without prior information about the environments.

The fisheye cameras with ultra-wide filed-of-view (FOV) can provide more spherical coverage of the surrounding environment. A dual-fisheye omnidirectional visual-inertial navigation system (VINS) combine two fisheye cameras and an inertial measurement unit (IMU), cover the whole surroundings of the MAV. It is the minimum sensor suite lending the omnidirectional perception with lightweight and small footprint.

In this paper, we show that it is possible to achieve reliable online autonomous navigation with omnidirectional perception using dual-fisheye VINS. Our system is a quadrotor equipped with two ultra-wide FOV fisheye cameras, which are rigidly mounted on two sides of a rod and facing opposite directions, a low-cost IMU, and heterogeneous onboard computing resources. The two fisheye cameras provide stereo observations with full 360-degree FOV in the horizontal direction and 60-degree FOV in the vertical direction, also provide full spherical monocular coverage of the surrounding situational awareness. Combine a highly accurate optimization-based dual-fisheye visual-inertial state estimator with online initialization and self-extrinsic calibration, three-dimensional map of the environments can be built. And an online trajectory planner that operates guarantees the safe navigation through cluttered environments in any direction. We provide experimental results to validate individual system modules as well as the overall performance in both indoor and outdoor environments.