The control of fully actuated hexa-rotors is highly dependent on their physical parameters, especially those describing the rotors' positions and orientations. Uncertainties in these parameters significantly affect the performance of classical control approaches, particularly in small drones, where manufacturing tolerances amplify these uncertainties. To address this challenge, we propose a novel adaptive and robust control strategy that compensates for parameter uncertainties and external disturbances without requiring precise prior knowledge of the rotor poses. Unlike existing methods, our approach explicitly incorporates motor dynamics into the control design, resulting in a more realistic and implementable framework. Using Lyapunov-based stability analysis, we demonstrate the global asymptotic stability of the proposed control system under parameter uncertainties and disturbances. Extensive simulations validate the efficacy of our method, showcasing superior tracking performance and robustness compared to conventional controllers. This work represents a significant step toward enabling fully actuated multi-rotor UAVs to perform in real-world scenarios with uncertain and dynamic environments.

Optic flow-based navigation, inspired by biological systems, differs from traditional vision-based techniques, reducing computational requirements and enabling a more reactive approach. So far, this optic flow-based approach suffers from certain limitations, including a reduced ability to make corridor exploration decisions, difficulty to detect frontal obstacle in advance and sensitivity to rotations. To overcome these challenges while preserving the benefits of direct optic flow regulation, we propose to use the spatial profile of the optical flow field to: (i) precisely control the UAV's orientation parallel to a corridor wall and (ii) head toward the aperture in a highly responsive manner. In addition, we showed that extracting cues from the spatial OF profile reconciles continuous trajectory and turns, without stopping the course of the aerial robot by specifying rules for visual inhibition during turns. We found that the specific optical flow signature created by the apertures in the environment facilitated decision-making and enabled the control architecture to maintain a certain pace for the UAV's navigation through the turns in corridors.

In honeybees, foragers use the waggle dance to communicate the direction and 'a distance' to a food source from the hive to other members of the colony. Behavioral studies indicate that dancing foragers estimate this 'distance' during outward flights (from the hive to the food source) based on visual cues, in particular optic flow, even over mountain slopes. While optic flow-based biologically plausible models for the honeybee visual odometer were previously presented, their robustness with respect to uneven terrain has not yet been investigated. In this study, we present a new model for the honeybee visual odometer, calledSuRf(forSurface-Reference Leveling model), which combines vertical oscillations with the constant reorientation of the honeybee's compound eye with respect to the slope overflown. As the simulated honeybee's compound eye is typically kept level with the surface below, the direction of perception of ventral and divergent optic flow tends to remain perpendicular to the surface, enhancing the robustness of their detection across terrain irregularities. Tested in open field simulations across uneven terrain and varied wind conditions, theSuRfmodel demonstrated greater accuracy compared to previous models evaluated under the same conditions, while keeping precision. The reliability of theSuRfmodel accounts for the observed 'mountain slopes' (uphill and downhill) behaviors in honeybees and offers promising applications for minimalistic aerial robots navigating uneven terrain and indoor spaces.

We present an insect-inspired visual teach-and-repeat framework demonstrated on Antflie, a 33-gram MAV equipped with an ultra-low-resolution camera (24 & times; 24 px) and a narrow 87 degrees field of view (FoV). During a one-shot outbound teach flight, the MAV performs periodic scans and classifies views as left or right of the route using inertial and optic-flow cues. These lateralized views are stored as compact visual memories in a Mushroom Body network with a footprint under 4 kB. In the repeat phase, the MAV flies the inbound route by retracing the outbound path, and autonomously lands at its home location using only visual familiarity through direct sensorimotor coupling, rather than map-based reasoning. Offline simulations show that the Route Lateralized (R-Lat) algorithm in Antflie matches the accuracy of a state-of-the-art insect visual compass (V-Comp) while running up to 20 & times; faster and supporting narrow FoVs. Real-world indoor experiments further demonstrate 24 autonomous inbound repeats totaling 110 meters of flight, with a 13-cm median lateral error and a mean landing error of 34 cm. These results highlight the feasibility of frugal, bio-inspired, vision-only navigation for MAVs operating under strict size, weight, power, and cost constraints, inspired by the navigation of Cataglyphis and Melophorus ants.

The implementation of collective motion, traditionally, disregard the limited sensing capabilities of an individual, to instead assuming an omniscient perception of the environment. This study implements a visual flocking model in a 'robot-in-the-loop' approach to reproduce these behaviors with a flock composed of 10 independent spherical robots. The model achieves robotic collective motion by only using panoramic visual information of each robot, such as retinal position, optical size and optic flow of the neighboring robots. We introduce a virtual anchor to confine the collective robotic movements so to avoid wall interactions. For the first time, a simple visual robot-in-the-loop approach succeed in reproducing several collective motion phases, in particular, swarming, and milling. Another milestone achieved with by this model is bridging the gap between simulation and physical experiments by demonstrating nearly identical behaviors in both environments with the same visual model. To conclude, we show that our minimal visual collective motion model is sufficient to recreate most collective behaviors on a robot-in-the-loop system that be implemented using several individuals, behaves as numerical simulations predict and is easily comparable to traditional models.

Real-time dense visual-inertial SLAM remains a major challenge for resource-constrained UAVs, especially when using 3D Gaussian Splatting (3DGS). While 3DGS enables high-quality 3D reconstruction, it suffers from high computational cost and weak tracking robustness. In this paper, we introduce VIGS-Fusion, an efficient 3DGS SLAM system that can operate fully onboard. Our method computes a robust and accurate pose from a single tightly-coupled optimization problem that incorporates RGB-D data, preintegrated IMU measurements, velocity, and IMU bias. By using a second-order approximation and a loss based on the observed image gradients for pose estimation, we significantly reduced computational time while maintaining high tracking performance. We also provide a real-world dataset collected with our compact quadrotor, capturing RGB-D and IMU measurements. Experimental results demonstrated accurate and robust tracking while achieving over a 30x reduction in computation time for the tracking compared to state-of-the-art 3DGS SLAM systems. For the first time, we demonstrated real-time 3D Gaussian Splatting SLAM running entirely onboard a 6-inch UAV. The code is available at: https://github.com/AbdoullahNdoye/VIGS-Fusion.git.

Ants achieve robust visual homing with minimal sensory input and only a few learning walks, inspiring biomimetic solutions for autonomous navigation. While Mushroom Body (MB) models have been used in robotic route following, they have not yet been applied to visual homing. We present the first real-world implementation of a lateralized MB architecture for visual homing onboard a compact autonomous car-like robot. We test whether the sign of the angular path integration (PI) signal can categorize panoramic views, acquired during learning walks and encoded in the MB, into "goal on the left" and "goal on the right" memory banks, enabling robust homing in natural outdoor settings. We validate this approach through four incremental experiments: (1) simulation showing attractor-like nest dynamics; (2) real-world homing after decoupled learning walks, producing nest search behavior; (3) homing after random walks using noisy PI emulated with GPS-RTK; and (4) precise stopping-at-the-goal behavior enabled by a fifth MB Output Neuron (MBON) encoding goal-views to control velocity. This mimics the accurate homing behavior of ants and functionally resembles waypoint-based position control in robotics, despite relying solely on visual input. Operating at 8 Hz on a Raspberry Pi 4 with 32x32 pixel views and a memory footprint under 9 kB, our system offers a biologically grounded, resource-efficient solution for autonomous visual homing.

Solitary foraging ants excel at route following using minimal neural resources, Robots don't. Recent biological studies proposed lateralized, nest-centric memories to explain ants' direct visual homing but did not address how ants follow curved visual routes away from their nest. We present a biologically inspired neuromorphic model for one-shot panoramic route learning and continuous route following, implemented on a compact car-like robot, Antcar. We demonstrate that route-centric lateralized memories, inspired by the insect mushroom body, enable Antcar to achieve bi-directional route-following, with motivation-driven recognition of route extremities and familiarity-based velocity control. With rigorous Lyapunov-based stability analysis and an empirical memory scalability evaluation, the model was tested over 1.6 km across 113 challenging real-world trials. The system achieves less than 25 cm median lateral error using minimal resources (800-pixel input, 300 MB RAM, 500 mW power, and 18.75 kB memory per 50 m route), offering insights into insect cognition and advancing autonomous robotics under strict resource constraints.

This work focuses on the design and control of a novel type of Flapping-Wing Micro Aerial Vehicle (FWMAV). The drone, known as the X-Fly, is a new under-actuated robotic platform that also has an inner control loop to stabilize its roll angle thanks to an onboard IMU. Such assistance makes the X-Fly easier to pilot. The under-actuation and the flapping oscillations make the modelling and the control of the X-Fly a challenging task. A dynamical model is introduced that is able to take advantage of the stabilized roll dynamics to separate the platform into two almost independent sub-systems, one for the altitude and another for the position on the x-y plane. A trajectory tracking controller for the altitude and a circular trajectory are then proposed and tested in order to corroborate the validity of the presented model.

This article introduces a novel self-triggering strategy designed to ensure the control of discrete-time linear systems with guaranteed stability, even in the presence of disturbances and uncertainties. This strategy aims to consistently maintain satisfaction of state constraints while accounting for the uncertainties in the system through a set-membership description. The self-triggering framework primarily relies on reachable and invariant sets. Reachable sets quantify the maximum deviation of the disturbed system from the predicted behavior, while an invariant set establishes triggering bounds for these reachable sets. This control method is intended to minimize the number of measurements required, thereby avoiding network bandwidth saturation. To validate the effectiveness of the proposed strategy, the experiments are conducted on an air extractor system, demonstrating a reduction in the number of measurement samples while ensuring stability and satisfying system state constraints.

Air ducts are integral to modern buildings but are challenging to access for inspection. Small quadrotor drones offer a potential solution, as they can navigate both horizontal and vertical sections and smoothly fly over debris. However, hovering inside air ducts is problematic due to the airflow generated by the rotors, which recirculates inside the duct and destabilizes the drone. In this article, we map the aerodynamic forces that affect a hovering drone in a duct using a robotic setup and a force/torque sensor. Based on the collected aerodynamic data, we identify a recommended position for stable flight, which is not the center of a circular duct. We then develop a neural network-based positioning system that leverages low-cost time-of-flight sensors. By combining these aerodynamic insights and the data-driven positioning system, we show how to improve the stability of a small quadrotor drone (here, 180 mm) inside small air ducts (down to 350 mm diameter) and fly autonomously over 2 m.

As the demand for autonomous robots grows, from self-driving cars to factory automation, so does the need for resource-efficient navigation algorithms. This challenge has inspired the development of sparse neural networks for visual navigation, drawing particularly from bio-inspired, ant-based neuromorphic artificial intelligence algorithms. Recent approaches leverage the neural architecture of the mushroom body in ants to learn and follow routes by recognizing familiar visual patterns. However, such models have not yet been implemented in closed-loop, resource-constrained robotic systems with satisfying performance. In this work, we present a mushroom body-inspired model embedded in the Antcar, a compact, car-like robot that processes panoramic images at low resolution. Our experiments reveal that reducing the scanning range and resolution improves visual compass performance for robot’s route-following. These findings highlight the trade-off between increased refresh rates and the benefits of selective perception, paving the way for more efficient real-time decision-making in resource-constrained systems.

Web-like obstacles, such as safety nets, represent a unique hazard for drones, and especially UAVs (Unmanned Aerial Vehicles). Fencing and netting are particularly difficult to distinguish from the background using either computer vision, lidar and sonar. In contrast, animals such as flying insects may detect these web-like obstacles using Optic Flow (OF), and more precisely motion parallax. A netting-avoidance solution was proposed using a OF-based detection method. The netting detection method was based on a signature defined by the shape of the OF magnitude across the visual field. We established that the OF shape depends on the orientation of the netting in relation to the hexarotor’s movement. This paper demonstrates netting detection in real-world experiments, according to any direction flight made by the UAV along the net. The proposed NOWA method (which stands for Netting Optical floW-based distinction Algorithm) separates the OF signatures belonging to these different surfaces -netting or background- whatever their orientations. By extracting the OF signatures of these different surfaces and separating them, the proposed visual method can estimate their relative locations and orientations. In a robotic simulations, the multirotor explores and navigates automatically using this netting detection method, using saccades to avoid obstacles. In the simulations, these saccades are also used to simplify netting detection by orienting itself systematically parallel to these planes, a behavior reminiscent of flying insects.

Solitary foraging ants excel in following long visual routes in complex environments with limited sensory and neural resources—an ability that remains challenging for robots with minimal computational power. Here, we introduce a self-supervised, insect-inspired neural network that enables robust route-following on the compact, low-cost Antcar robot. The robot leverages key aspects of ant brain and behavior: (i) continuous, one-shot visual route learning using panoramic encoding in a mushroom body-inspired network, (ii) categorization of low-resolution egocentric panoramas via oscillatory movements, (iii) opponent-process control of angular and forward velocities based on visual familiarity, (iv) recognition of places of interest along routes, and (v) motivation-based memory modulation. Antcar autonomously followed routes between indoor or outdoor destinations, forward or backward, while remaining stable in both theoretical analysis and real-world testing despite occlusions and visual changes. Across 1.3 km of autonomous travel, Antcar achieved challenging route-following with sub-20 cm lateral error at speeds up to 150 cm/s, requiring only 148 kilobits of memory and processing panoramas every 62 ms. This efficient, brain-inspired architecture stands out from more sensor-intensive and computationally demanding methods, presenting a neuromorphic approach with valuable insights into insect navigation and practical robotic applications.

Animal collective behavior is often modeled with self-propelled particles, assuming each individual has “omniscient” knowledge of its neighbors. Yet, neighbors may be hidden from view and we do not know the effect of this information loss. To address this question, we propose a visual model of collective behavior where each particle moves according to bioplausible visual cues, in particular the optic flow. This visual model successfully reproduces three classical collective behaviors: swarming, schooling, and milling. This model offers a potential solution for controlling artificial swarms visually.

Estimating distance traveled is a frequently arising problem in robotic applications designed for use in environments where GPS is only intermittently or not at all available. In UAVs, the presence of weight and computational power constraints makes it necessary to develop odometric strategies based on minimilastic equipment. In this study, a hexarotor was made to perform up-and-down oscillatory movements while flying forward in order to test a self-scaled optic flow based odometer. The resulting self-oscillatory trajectory generated series of contractions and expansions in the optic flow vector field, from which the flight height of the hexarotor could be estimated using an Extended Kalman Filter. For the odometry, the downward translational optic flow was scaled by this current visually estimated flight height before being mathematically integrated to obtain the distance traveled. Here we present three strategies based on sensor fusion requiring no, precise or rough prior knowledge of the optic flow variations generated by the sinusoidal trajectory. The “rough prior knowledge” strategy is based on the shape and timing of the variations in the optic flow. Tests were performed first in a flight arena, where the hexarotor followed a circular trajectory while oscillating up and down over a distance of about [Formula: see text] m under illuminances of [Formula: see text] lux and [Formula: see text] lux. Preliminary field tests were then performed, in which the hexarotor followed a longitudinal bouncing [Formula: see text]-long trajectory over an irregular pattern of grass.

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.

In this letter, we describe a new light-flashing shield to be used at flash-based imaging or event-based vision tasks performed in environments with poor light conditions. The shield incorporates a multiplexer that permits setting of the operation mode, the nominal current, the pulse width, and the trigger sensitivity of up to 2 high-brightness LEDs by means of a single inter-integrated circuit (I2C) bus. A trigger conditioning circuit permits synchronization of the LEDs light and an Intel RealSense d455 camera to obtain a set of clear images while the LEDs are illuminated, this feature being the main novelty and contribution of this work, in order to provide light on-demand in visual-based tasks performed in environments with poor light conditions. Images obtained during real tests prove the correct synchronization between the light-flashing shield and the camera.