A method for ego-motion estimation in micro-hovering platforms flying in very cluttered environments

We aim at developing autonomous miniature hovering flying robots capable of navigating in unstructured GPS-denied environments. A major challenge is the miniaturization of the embedded sensors and processors that allow such platforms to fly by themselves. In this paper, we propose a novel ego-motion estimation algorithm for hovering robots equipped with inertial and optic-flow sensors that runs in real- time on a microcontroller and enables autonomous flight. Unlike many vision-based methods, this algorithm does not rely on feature tracking, structure estimation, additional dis- tance sensors or assumptions about the environment. In this method, we introduce the translational optic-flow direction constraint, which uses the optic-flow direction but not its scale to correct for inertial sensor drift during changes of direction. This solution requires comparatively much sim- pler electronics and sensors and works in environments of any geometry. Here we describe the implementation and per- formance of the method on a hovering robot equipped with eight 0.65 g optic-flow sensors, and show that it can be used for closed-loop control of various motions.

Optic-Flow Based Control of a 46g Quadrotor

Adrien Briod, Dario Floreano, Jean-Christophe Zufferey

We aim at developing autonomous miniature hov- ering flying robots capable of navigating in unstructured GPS- denied environments. A major challenge is the miniaturization of the embedded sensors and processors allowing such platforms to fly autonomously. In this paper, we propose a novel ego-motion estimation algorithm for hovering robots equipped with inertial and optic-flow sensors that runs in real- time on a microcontroller. Unlike many vision-based methods, this algorithm does not rely on feature tracking, structure estimation, additional distance sensors or assumptions about the environment. Key to this method is the introduction of the translational optic-flow direction constraint (TOFDC), which does not use the optic-flow scale, but only its direction to correct for inertial sensor drift during changes of direction. This solution requires comparatively much simpler electronics and sensors and works in environments of any geometries. We demonstrate the implementation of this algorithm on a miniature 46g quadrotor for closed-loop position control.

2013

A Sensor Fusion Approach to the Estimation of Instantaneous Velocity Using Single Wearable Sensor During Sprint

Kamiar Aminian, Salil Apte, Farzin Dadashi, Vincent Gremeaux, Frédéric Meyer

Power-Force-Velocity profile obtained during a sprint test is crucial for designing personalized training and evaluating injury risks. Estimation of instantaneous velocity is requisite for developing these profiles and the predominant method for this estimation assumes it to have a first order exponential behavior. While this method remains appropriate for maximal sprints, the sprint velocity profile may not always show a first-order exponential behavior. Alternately, velocity profile has been estimated using inertial sensors, with a speed radar, or a smartphone application. Existing methods either relied on the exponential behavior or timing gates for drift removal, or estimated only the mean velocity. Thus, there is a need for a more flexible and appropriate approach, allowing for instantaneous velocity estimation during sprint tests. The proposed method aims to solve this problem using a sensor fusion approach, by combining the signals from wearable Global Navigation Satellite System (GNSS) and inertial measurement unit (IMU) sensors. We collected data from nine elite sprinters, equipped with a wearable GNSS-IMU sensor, who ran two trials each of 60 and 30/40 m sprints. We developed an algorithm using a gradient descent-based orientation filter, which simplified our model to a linear one-dimensional model, thus allowing us to use a simple Kalman filter (KF) for velocity estimation. We used two cascaded KFs, to segment the sprint data precisely, and to estimate the velocity and the sprint duration, respectively. We validated the estimated velocity and duration with speed radar and photocell data as reference. The median RMS error for the estimated velocity ranged from 6 to 8%, while that for the estimated sprint duration lied between 0.1 and -6.0%. The Bland-Altman plot showed close agreement between the estimated and the reference values of maximum velocity. Examination of fitting errors indicated a second order exponential behavior for the sprint velocity profile, unlike the first order behavior previously suggested in literature. The proposed sensor-fusion algorithm is valid to compute an accurate velocity profile with respect to the radar; it can compensate for and improve upon the accuracy of the individual IMU and GNSS velocities. This method thus enables the use of wearable sensors in the analysis of sprint test.

2020

Human-Centered Scene Understanding via Crowd Counting

Weizhe Liu

Human-centered scene understanding is the process of perceiving and analysing a dynamic scene observed through a network of sensors with emphasis on human-related activities. It includes the visual perception of human-related activities from either single image or video sequence. Scene understanding with focus of human-related activities is becoming increasingly popular which results in the demand of algorithms that can efficiently model crowd activity in different real-world scenarios. In this thesis, we exploit human-centered scene understanding through crowd counting. Counting people is a challenging task due to perspective distortion and occlusion. We tackle these problems by developing algorithms to leverage a variety of data modalities including single image, video sequence and scene perspective map. First, we introduce an end-to-end trainable deep architecture for crowd counting that combines features obtained using multiple receptive field sizes and learns the importance of each such feature at each image location. In other words, our approach adaptively encodes the scale of the contextual information required to accurately predict crowd density. This yields an algorithm that outperforms previous crowd counting methods, especially when perspective effects are strong.Second, we explicitly model the scale changes and reason in terms of people per square-meter. We show that feeding the perspective model to the network allows us to enforce global scale consistency and that this model can be obtained on the fly from the drone sensors. In addition, it also enables us to enforce physically-inspired temporal consistency constraints that do not have to be learned. This yields an algorithm that outperforms previous methods in inferring crowd density from a moving drone camera especially when perspective effects are strong.Third, for video sequence, we advocate estimating people flows across image locations between consecutive images and inferring the people densities from these flows instead of directly regressing them. This enables us to impose much stronger constraints encoding the conservation of the number of people. As a result, it significantly boosts performance without requiring a more complex architecture. Furthermore, it allows us to exploit the correlation between people flow and optical flow to further improve the results. We also show that leveraging people conservation constraints in both a spatial and temporal manner makes it possible to train a deep crowd counting model in an active learning setting with much fewer annotations. This significantly reduces the annotation cost while still leading to similar performance to the full supervision case.