A 10-gram Microflyer for Vision-based Indoor Navigation

Abstract

Abstract-We aim at developing ultralight autonomous microflyers capable of navigating within houses or small built environments. Our latest prototype is a fixed-wing aircraft weighing a mere 10 g, flying around 1.5 m/s and carrying the necessary electronics for airspeed regulation and collision avoidance. This microflyer is equipped with two tiny camera modules, two rate gyroscopes, an anemometer, a small microcontroller, and a Bluetooth radio module. In-flight tests are carried out in a new experimentation room specifically designed for easy changing of surrounding textures. I. CHALLENGES AND STATE OF THE ARTThere are currently no autonomous flying robots capable of navigating indoors, within enclosed environments such as offices or houses. Although they could be useful in many applications such as surveillance, hazardous environment exploration, search and rescue, etc., the challenges engineers are facing to develop such robots are numerous. In order to be able to fly at very low speed (below 2 m/s) such flying systems must be ultra-lightweight (usually well below 50 g), which implies tremendous constraints in terms of embedded computational power, sensor simplicity, and airframe architecture. Moreover, controlling such systems is quite different from controlling more conventional outdoor micro aerial vehicles, which can rely on high-precision inertial measurement units, global positioning systems, radars or other conventional distance sensors, and/or visual horizon detection systems [1]- [3].In this paper we present the latest prototype resulting from our research in the domain of indoor microflyers since 2001 [4]-[7]. This robot, called MC1, has an overall weight of 10 g including visual, inertial, and air flow sensors, which enable a certain degree of autonomy: automatic take-off, speed regulation, and collision avoidance. These capabilities have been demonstrated in a 7x6-m room equipped with randomly textured walls.To the best of our knowledge, the MC1 is the lightest motorized free-flying robot produced to date. Oh and collaborators have been working on automatic landing and collision avoidance with an indoor flying robot weighing around 30 g [8], [9]. However, these experiments were carried out in relatively large indoor environments such as basketball courts and only one vision sensor was embedded at one time allowing In summer 2004, we demonstrated the first visually-guided free-flying indoor aircraft. This was done with an earlier prototype (designated F2) weighing 30 g and flying in a 16x16-m room equipped with evenly-distributed black and white vertical stripes made of suspended fabrics. The experiment consisted of having the 80-cm-wingspanned aircraft autonomously steer like a fly, i.e., following straight trajectories when far from any walls and engaging a fast left or right turn when close to a wall (see [7] for details).With the 10-gram MC1 described in this paper, we made three significant steps forward since then:1) The overall weight of the robot has been reduced from 30 g to 10 g while the...

1kHz 2D silicon retina motion sensor platform

“…If the shift between 0 f and the new frame 1 f is sufficiently small (which is achieved by adjusting t  small enough), then the image shift or optic flow x  can be estimated by finding an approximation to 1 f , called 1 f , that is a linear weighted combination of the reference images R f and L f :…”

“…The unknown shift is found by minimizing the least-square error between 1 f and its approximation 1 f . The optimum shift is computed in a closed form in a single step, with the assumption that a rigid object moves a short distance (less than a pixel) between 2 frames.…”

Three-dimensional position and velocity regulation of a quad-rotorcraft using optical flow

“…Inspired also by behaviors observed in nature, the authors in [14] developed an optical flow-based autopilot to regulate speed and enable lateral obstacle avoidance. A similar strategy was proposed in [15], where a 10 g micro-UAV was built and equipped with sensors and controllers allowing obstacle avoidance using optical-flow data. While most of the previously mentioned methods consider vision-based tasks relying on multiple-camera systems [3][4][5][6][7]14] and sometimes on nonconventional sensors [11,12,15], our method explores the implementation of a monocular imaging system consisting of a conventional camera.…”

“…A similar strategy was proposed in [15], where a 10 g micro-UAV was built and equipped with sensors and controllers allowing obstacle avoidance using optical-flow data. While most of the previously mentioned methods consider vision-based tasks relying on multiple-camera systems [3][4][5][6][7]14] and sometimes on nonconventional sensors [11,12,15], our method explores the implementation of a monocular imaging system consisting of a conventional camera. In contrast to the solutions presented in [8,9,10], which deal with stabilizing the aerial vehicle in hover or in take-off and landing tasks, the research presented in this paper complements and improves these previous results by addressing the problem of stabilizing the three-dimensional position and translational velocity of a quad-rotorcraft during autonomous flights.…”

Optical flow-based controller for reactive and relative navigation dedicated to a four rotor rotorcraft

“…After a calibration, the cancelation of the rotational components is obtained by substracting the gyroscopic value from the OF measurement. They use this fusion system in [9] to avoid the obstacles presents on a microflyer path.In this paper, we propose a state machine approach for the relative and reactive navigation of a rotorcraft.…”