Here we describe how a 10-gram Flapping-Wing Micro Aerial Vehicle (FWMAV) was able to perform an automatic trajectory tracking task based on a vector field method. In this study, the desired heading was provided by a vector field which was computed depending on the desired trajectory. The FWMAV's heading was changed by a rear steering mechanism. This rear mechanism simultaneously (i) tenses one wing and relaxes the opposite wing, and (ii) moves the rudder in the same direction as the wing is relaxed. Due to the complex dynamics, system identification methods were used to identify simple linear models using a set of dedicated free flight tests. This yaw and roll simple models help to adjust the yaw controller and the inner loop roll controller. The experimental results obtained here show that a time-independent vector field-based strategy is robust to various initial position and/or speed conditions. The task of tracking circular and 8-shaped trajectories was accomplished successfully over tens of meters.

Both neurophysiological and psychophysical experiments have pointed out the crucial role of recurrent and feedback connections to process context-dependent information in the early visual cortex. While numerous models have accounted for feedback effects at either neural or representational level, none of them were able to bind those two levels of analysis. Is it possible to describe feedback effects at both levels using the same model? We answer this question by combining Predictive Coding (PC) and Sparse Coding (SC) into a hierarchical and convolutional framework applied to realistic problems. In the Sparse Deep Predictive Coding (SDPC) model, the SC component models the internal recurrent processing within each layer, and the PC component describes the interactions between layers using feedforward and feedback connections. Here, we train a 2-layered SDPC on two different databases of images, and we interpret it as a model of the early visual system (V1 & V2). We first demonstrate that once the training has converged, SDPC exhibits oriented and localized receptive fields in V1 and more complex features in V2. Second, we analyze the effects of feedback on the neural organization beyond the classical receptive field of V1 neurons using interaction maps. These maps are similar to association fields and reflect the Gestalt principle of good continuation. We demonstrate that feedback signals reorganize interaction maps and modulate neural activity to promote contour integration. Third, we demonstrate at the representational level that the SDPC feedback connections are able to overcome noise in input images. Therefore, the SDPC captures the association field principle at the neural level which results in a better reconstruction of blurred images at the representational level.

Hierarchical sparse coding (HSC) is a powerful model to efficiently represent multidimensional, structured data such as images. The simplest solution to solve this computationally hard problem is to decompose it into independent layer-wise subproblems. However, neuroscientific evidence would suggest interconnecting these subproblems as in predictive coding (PC) theory, which adds top-down connections between consecutive layers. In this study, we introduce a new model, 2-layer sparse predictive coding (2L-SPC), to assess the impact of this interlayer feedback connection. In particular, the 2L-SPC is compared with a hierarchical Lasso (Hi-La) network made out of a sequence of independent Lasso layers. The 2L-SPC and a 2-layer Hi-La networks are trained on four different databases and with different sparsity parameters on each layer. First, we show that the overall prediction error generated by 2L-SPC is lower thanks to the feedback mechanism as it transfers prediction error between layers. Second, we demonstrate that the inference stage of the 2L-SPC is faster to converge and generates a refined representation in the second layer compared to the Hi-La model. Third, we show that the 2L-SPC top-down connection accelerates the learning process of the HSC problem. Finally, the analysis of the emerging dictionaries shows that the 2L-SPC features are more generic and present a larger spatial extension.

In this paper, we present (i) a novel bio-inspired 1-D OF sensor which is robust to high-dynamic-range lighting conditions and independent of the visual patterns encountered, and (ii) a low-cost car-like robot called BioCarBot, which estimates its velocity and steering angle by means of an Extended Kalman Filer (EKF) using only the OF measurements delivered by two downward-facing sensors of this kind. Indoor experiments were carried out, in which the robot was driven in the closed-loop mode, using a proportional integral (PI) controller based on the velocity and steering angle estimates. The results presented here show that our novel OF sensor can deliver a wide range of high-frequency (333Hz) OF measurements (from 1 to 10 rad/s ) with a relatively high resolution (up to 0.05 rad/s) in a 5-decade high-dynamic range of light levels. Neither the refresh rate nor the resolution of the OF sensors presented here depended on either the visual patterns or the lighting conditions, and could be theoretically set at whatever value required.

The observer presented in this paper, which was based on the use of three minimalistic bio-inspired Visual Motion Sensors (VMS) detecting Optic Flow (OF) cues, states was intended as a backup solution in the case of Inertial Measurement Unit (IMU) failure. Contrary to most previous Guidance Navigation and Control (GNC) solutions for planetary landing, which have involved a sensor suite including an IMU, an innovative strategy is presented here for estimating states without any need for inertial measurements, based solely on information about the relative velocity of the images of the surrounding environment. A Linear Parameter Varying (LPV) observer designed on a LPV system linearized around a reference trajectory, estimates: the ventral OF, the expansion OF and the local pitch angle. A previously developed observer was applied here to a larger class of nonlinear systems by making an ingenious change of variable. Simulations performed on a lunar landing scenario yielded satisfactory performance and showed the robustness of the OF based observer to initial uncertainties and measurement noise.

The 6-pixel low-speed Visual Motion Sensor (VMS) inspired by insects' visual systems presented here performs local 1-D angular speed measurements ranging from 1.5°/s to 25°/s and weighs only 2.8 g. The entire optic flow processing system, including the spatial and temporal filtering stages, has been updated with respect to the original design. This new lightweight sensor was tested under free-flying outdoor conditions over various fields onboard a 80 kg unmanned helicopter called ReSSAC. The visual disturbances encountered included helicopter vibrations, uncontrolled illuminance, trees, roads, and houses. The optic flow measurements obtained were finely analyzed online and also offline, using the sensors of various kinds mounted onboard ReSSAC. The results show that the optic flow measured despite the complex disturbances encountered closely matched the approximate ground-truth optic flow.

The new robot called BeeRotor we have developed is a tandem rotorcraft that mimicks optic flow-based behaviors previously observed in flies and bees. This tethered miniature robot (80g), which is autonomous in terms of its computational power requirements, is equipped with a 13.5-g quasi-panoramic visual system consisting of 4 individual visual motion sensors responding to the optic flow generated by photographs of natural scenes, thanks to the bio-inspired “time of travel” scheme. Based on recent findings on insects' sensing abilities and control strategies, the BeeRotor robot was designed to use optic flow to perform complex tasks such as ground and ceiling following while also automatically driving its forward speed on the basis of the ventral or dorsal optic flow. In addition, the BeeRotor robot can perform tricky manoeuvers such as automatic ceiling docking by simply regulating its dorsal or ventral optic flow in high-roofed tunnel depicting natural scenes. Although it was built as a proof of concept, the BeeRotor robot is one step further towards achieving a fully-autonomous micro-helicopter which is capable of navigating mainly on the basis of the optic flow.

Autonomous landing on unknown extraterrestrial bodies requires fast, noise-resistant motion processing to elicit appropriate steering commands. Flying insects excellently master visual motion sensing techniques to cope with highly parallel data at a low energy cost, using dedicated motion processing circuits. Results obtained in neurophysiological, behavioural, and biorobotic studies on insect flight control were used to safely land a spacecraft on the Moon in a simulated environment. ESA's Advanced Concepts Team has identified autonomous lunar landing as a relevant situation for testing the potential applications of innovative bio-inspired visual guidance systems to space missions. Biomimetic optic flow-based strategies for controlling automatic landing were tested in a very realistic simulated Moon environment. Visual information was provided using the PANGU software program and used to regulate the optic flow generated during the landing of a two degrees of freedom spacecraft. The results of the simulation showed that a single elementary motion detector coupled to a regulator robustly controlled the autonomous descent and the approach of the simulated moonlander. “Low gate” located approximately 10 m above the ground was reached with acceptable vertical and horizontal speeds of 4 m/s and 5 m/s, respectively. It was also established that optic flow sensing methods can be used successfully to cope with temporary sensor blinding and poor lighting conditions.

We have developed a new vision based autopilot able to make a micro-air vehicle automatically navigate in steep relief. It uses onboard Optic Flow sensors inspired by the housefly's Elementary Motion Detectors (EMDs) that were previously investigated at our Laboratory with electrophysiological and micro-optical techniques. The paper investigates how the ground avoidance performances of the former OCTAVE robot could be enhanced to cope with steep relief. The idea is to combine frontal and ventral OF sensors and to merge feedback and feedforward loops. In the new robot, called OCTAVE(2), a feedback loop adjusts the lift so as to keep the ventral OF constant, while a feedforward loop based on a forward looking EMD sensor serves to anticipate the steep relief. We test the new autopilot on a 100-gram tethered rotorcraft that circles indoors over an environment composed of contrasting features randomly arranged on the floor. We show that OCTAVE(2) succeeds in following a relatively steep relief (maximum slope 17deg) while navigating close to the ground (groundheight in the order of lm). This risky task is performed thanks to a minimalist electronic visual system: the OF sensor suite is lightweight (4.3-grams including the lenses), and is therefore mounted onboard.

We have developed a visually based autopilot which is able to make a micro air vehicle (MAV) automatically take off, cruise and land, while reacting adequately to wind disturbances. We built a proof-of-concept, tethered rotorcraft that can travel indoors over an environment composed of contrasting features randomly arranged on the floor. Here we show the feasibility of a visuomotor control loop that acts upon the thrust so as to maintain the optic flow (OF) estimated in the downward direction to a reference value. The sensor involved in this OF regulator is an elementary motion detector (EMD). The functional structure of the EMD was inspired by that of the housefly, which was previously investigated at our laboratory by performing electrophysiological recordings while applying optical microstimuli to single photoreceptor cells of the compound eye. The vision based autopilot, which we have called OCTAVE (optic flow control system for aerospace vehicles) solves complex problems such as terrain following, controls risky maneuvers such as take off and landing and responds appropriately to wind disturbances. All these reputedly demanding tasks are performed with one and the same visuomotor control loop. The non-emissive sensor and simple processing system are particularly suitable for use with MAV, since the tolerated avionic payload of these micro-aircraft is only a few grams. OCTAVE autopilot could also contribute to relieve a remote operator from the lowly and difficult task of continuously piloting and guiding an UAV. It could also provide guiding assistance to pilots of manned aircraft.