Recent advances in aerial robotics have enabled the use of multirotor vehicles for autonomous payload transportation. Resorting only to classical methods to reliably model a quadrotor carrying a cable-slung load poses significant challenges. On the other hand, purely data-driven learning methods do not comply by design with the problem’s physical constraints, especially in states that are not densely represented in training data. In this work, we explore the use of physics-informed neural networks to learn an end-to-end model of the multirotor-slung-load system and, at a given time, estimate a sequence of the future system states. An LSTM encoder-decoder with an attention mechanism is used to capture the dynamics of the system. To guarantee the cohesiveness between the multiple predicted states of the system, we propose the use of a physics-based term in the loss function, which includes a discretized physical model derived from first principles together with slack variables that allow for a small mismatch between expected and predicted values. To train the model, a dataset using a real-world quadrotor carrying a slung load was curated and is made available. Prediction results are presented and corroborate the feasibility of the approach. The proposed method outperforms both the first principles physical model and a comparable neural network model trained without the physics regularization proposed.

We address the problem of trajectory tracking control of a quadrotor-slung-load system. A novel control framework is proposed that relaxes the differentiability class order of the reference. In lieu of traditional approaches that primarily regard the position of either the quadrotor or the load in the design of tracking controllers, in this work, we consider avirtualpoint located along the cable between the vehicle and the payload. The resulting control framework, which draws parallels with a typical single quadrotor dynamic system, is characterized by an utmost simplicity since it relies only on three error variables, and requires only up to the third time derivative of the reference trajectory. The overall control scheme ensures almost global asymptotic stability of the closed-loop system in the presence of constant external disturbances. Extensive experimental results featuring time-varying exogenous perturbations are presented to validate and illustrate the effectiveness of the proposed method.

As climate change prospects emphasize the need for local adaptation, tackling heat exposure becomes a priority in Southern Europe. In the Greater Lisbon Area, heatwaves (HW) are already frequent, and their severity is expected to aggravate in the coming decades. Accordingly, there is a pressing need to pinpoint the neighborhoods where people are relatively more exposed to the excess heat. In this study, air temperature measurements from citizen owned stations are retrieved from open data sources, quality controlled and co-located with Earth Observation (EO) data to downscale the short-term numerical weather predictions (NWP) from 2.5km up to a metric scale (200m), using a Machine learning (ML) approach. While the NWP model resolves the regional physical processes, the ML model is able to disclose the air temperature response to more detailed local features, such as the urban heat island effect, supporting the local public health stakeholders in their decision-making.

In this article, we address the task of cooperative path following control of multiple autonomous quadcopters in the presence of unknown external disturbances. Under the assumption that the communications among the vehicles are bidirectional and continuous, a synchronized path following strategy is proposed that regulates the speed of each vehicle along its path to reach consensus in relative position. Moreover, collision-free transient paths from the vehicle initial positions to a group of suitable selected positions along the desired paths are designed. Building on the backstepping technique, the proposed path following controller for each individual vehicle drives the quadcopter toward, and to stay within an arbitrarily small neighborhood of its corresponding desired path, achieving global uniformly ultimately boundedness. In addition, the devised controller guarantees that the actuations are bounded with respect to the position error. The controller is also made robust to external constant or slowly time-varying disturbances by the introduction of dynamic estimators for the disturbances. A projection operator is used to ensure that the estimates remain within the prescribed bounds and are sufficiently smooth to be backstepped. In order to validate the effectiveness and performance of the proposed methodology, we present and analyze both simulation and experimental results.

This paper addresses the landing problem of a vertical take-off and landing vehicle, exemplified by a quadrotor, on a moving platform using image-based visual servo control. Observable features on a flat and textured target plane are exploited to derive a suitable control law. The target plane may be moving with bounded linear acceleration in any direction. For control purposes, the image of the centroid for a collection of landmarks is used as position measurement, whereas the translational optical flow is used as velocity measurement. The proposed control law guarantees convergence to the desired landing spot on the target plane, without estimating any parameter related to the unknown height, which is also guaranteed to remain strictly positive. Moreover, convergence is guaranteed even in the presence of bounded and possibly time-varying disturbances, resulting, for example, from the motion of the target plane, measurement errors, or wind-induced force disturbances. To improve performance, an estimator for unknown constant force disturbances is also included in the control law. Simulation and experimental results are provided to illustrate and assess the performance of the proposed controller.

This work addresses the design and experimental evaluation of a robust controller for a quadrotor landing maneuver comprising the approach to a landing slope and sliding on that slope, before coming to a complete halt. During the critical landing flight phase the dynamics of the vehicle change with the type of contact with the ground and a hybrid automaton, whose states reflect the several dynamic behaviors of the quadrotor, is employed to model the vehicle throughout the complete maneuver. The quadrotor landing problem is broken down into separate maneuver generation and robust trajectory tracking problems, which are combined to achieve a successful maneuver that is robust to possible uncertainties. Experimental results are provided to attest the feasibility of the proposed landing procedure.

In this paper, we tackle the problem of trajectory tracking for a particular class of underactuated vehicles with full torque actuation and a single force direction (thrust) that is fixed relative to a body attached frame. Additionally, we consider that thrust reversal is not available. We present the design of a hybrid controller that, under some given assumptions, is able to globally asymptotically stabilize the vehicle to a reference position trajectory while minimizing the angle to the desired attitude trajectory. This objective is achieved robustly and globally, in the sense that small perturbations do not lead to instability and it is achieved regardless of the initial state of the vehicle. The algorithm is tested in a experimental setup, using a small scale quadrotor vehicle and the VICON motion capture system.

This paper presents an on-line three dimensional trajectory planner for an unmanned vehicle following a Leader vehicle, where the Follower remains at a specified relative position with respect to the Leader. The planner defines the trajectory for a virtual vehicle that will be used as reference for the real Follower vehicle. The virtual Follower behaves like a three dimensional trailer attached to the Leader and its reference frame is used to specify the relative position vector. At the kinematic level, the proposed planner requires only the knowledge of the Leader's linear speed, but not its angular speed, and the Follower's trajectory can be generated on-line, in the sense that no a priori knowledge of the Leader's trajectory is required. Experimental results obtained with quadrotor vehicles are presented, which demonstrate the richness of the planned trajectories.

This paper addresses the problem of robust take-off of a quadrotor unmanned aerial vehicle (UAV) in critical scenarios, such as in the presence of sloped terrains and surrounding obstacles. Throughout the maneuver, the vehicle is modeled as a hybrid automaton whose states reflect the different dynamic behaviors exhibited by the UAV. The original take-off problem is then addressed as the problem of tracking suitable reference signals in order to achieve the desired transitions between different hybrid states of the automaton. Reference trajectories and feedback control laws are derived to explicitly account for uncertainties in both the environment and the vehicle dynamics. Simulation results demonstrate the effectiveness of the proposed solution and highlight the advantages with respect to more standard open-loop strategies, especially for cases in which the slope of the terrain renders the take-off maneuver more critical to achieve.

This paper presents an experimentally evaluated solution to the problem of estimating the attitude of a rigid body using rate gyros and a pan-tilt camera. A nonlinear attitude observer combines angular velocity measurements obtained from rate gyros with images of a planar scene provided by the camera. By directly exploiting the sensor information, a stabilizing feedback law is introduced, and exponential convergence to the origin of the estimation errors is shown. Additionally, an active-vision system is proposed that relies on an image-based exponentially input-to-state-stable control law for the pan and tilt angular rates of the camera to keep the features in the image plane. Using recent results in geometric numerical integration, a multirate implementation of the observer is proposed, which exploits the complementary bandwidth of the sensors. Practical considerations, such as the lens-distortion compensation and the computation of suitable observer feedback gains, are considered. Experimental results obtained with a high-accuracy motion rate table demonstrate the high level of performance attained by the proposed solution.

This paper addresses the problem of robust takeoff and landing control of a quadrotor UAV (Unmanned Aerial Vehicle). During the critical flight phases of takeoff and landing the vehicle dynamics change according to the possible existence of contact with the ground. To model the vehicle during the overall maneuver a hybrid automaton is used where each state corresponds to a different dynamic behavior exhibited by the UAV. The original takeoff and landing problems are then addressed as a problem of tracking suitable reference signals in order to achieve the desired transitions between different hybrid states of the automaton. Both reference trajectories and feedback control laws are derived to explicitly account for measurement noise and uncertainties, in both the environment and in the vehicle dynamics. Simulation results demonstrate the effectiveness of the proposed solution.