1 September 2021 in Ecological Modelling
14 June 2021 in IEEE Transactions on Automation Science and Engineering
This article presents a novel information-based mission planner for a drone tasked to monitor a spatially distributed dynamical phenomenon. For the sake of simplicity, the area to be monitored is discretized. The insight behind the proposed approach is that, due to the spatiotemporal dependencies of the observed phenomenon, one does not need to collect data on the entire area, which is one of the main limiting factors in unmanned aerial vehicle (UAV) applications due to their limited autonomy. In fact, unmeasured states can be estimated using an estimator, such as a Kalman filter. In this context, the planning problem becomes the one of generating a flight plan that maximizes the quality of the state estimation while satisfying the flight constraints (e.g., flight time). The first result of this article is the formulation of this problem as a special orienteering problem where the cost function is a measure of the quality of the estimation. This results in a mixed-integer semidefinite formulation, which can be optimally solved for small instances of the problem. For larger instances, a heuristic is proposed, which provides suboptimal results. Simulations numerically demonstrate the capabilities and efficiency of the proposed path-planning strategy. We believe that this approach has the potential to increase dramatically the area that a drone can monitor, thus increasing the number of applications where monitoring with drones can become economically convenient. Note to Practitioners —This article was motivated by the problem of performing large-scale field monitoring activities using unmanned aerial vehicle (UAV), which at the moment is very time-consuming and limits the definitive adoption of UAVs for this kind of activities. This problem is caused by the limited autonomy of commercial UAVs and the lack of systematic ways to plan missions so as to maximize the amount of information collected. This work starts from the observation that, in many applications, the phenomena that one wants to observe have dynamics and statistical properties. Accordingly, data that are not directly measured can be estimated with a characterizable observation error. In this article, we develop the theoretical foundations for an information-based path planning and define the problem of designing the optimal mission as an optimization problem based on the knowledge of the monitored phenomenon. The presented results have the potential to dramatically improve the effectiveness of drones for monitoring applications.
17 January 2020 in IEEE Robotics and Automation Letters
In this letter, inspired by the needs of the European H2020 Project PANTHEON, we propose a full navigation stack purposely designed for the autonomous navigation of Ackermann steering vehicles in precision farming settings. The proposed stack is composed of a local planner and a pose regulation controller, both implemented in ROS. The local planner generates, in real-time, optimal trajectories described by a sequence of successive poses. The planning problem is formulated as a real-time cost-function minimization problem over a finite time horizon where the Ackermann kinematics and the presence of obstacles are encoded as constrains. The control law ensures the convergence toward each of these poses. To do so, in this paper we propose a novel non-smooth control law designed to ensure the solvability of the pose regulation problem for the Ackermann vehicle. A theoretical characterization of the convergence property of the proposed pose regulation controller is provided. Numerical simulations along with real-world experiments are provided to corroborate the effectiveness of the proposed navigation strategy.
1 January 2021 in Annual Reviews in Control
30 June 2026 in Ecological Modelling