This paper proposes an environmental-learning-based coverage control algorithm using a group of multi-rotor UAVs. The proposed method is a decentralized control system where data management and computation are performed by each agent. The environment is learned using Gaussian Process Regression based on data from the agent itself and its Voronoi neighbors. However, as the acquired training data increases, the computational load also increases. Therefore, in this paper, each agent restricts the shared data to those within its own Voronoi region. Additionally, the aim is to shorten operation time by concurrently conducting agent movement and environmental learning. The validity of the proposed method is demonstrated through numerical simulations and experiments.

This paper proposes a novel passivity-based full-order observer for mechanical port-Hamiltonian (pH) systems. We demonstrate its applicability in designing second-order sliding mode observers. Observer design can be broadly categorized into two types: minimal-order and full-order observers. While minimal-order observers prioritize reducing complexity, full-order observers provide greater design flexibility, enabling the realization of various observer structures to meet specific requirements. For mechanical pH systems, this work proposes a full-order observer, where the dynamics of the estimation error between the observer states and the plant states are represented as a Hamiltonian system. This design offers significant flexibility, facilitating the implementation of diverse observer types. Notably, this flexibility allows us to design an observer with estimation error dynamics achieving second-order sliding mode behavior and finite-time convergence. The effectiveness of the proposed method is demonstrated through a numerical example.

Control systems for mass-produced products are allowed to have tolerances for each of their components.herefore, the model parameters of the controlled object also have a certain tolerance, and are designed based on nominal values while taking into account perturbations. As a result, the target value response characteristics also have a certain range of variation.A design method has been proposed that uses data from a single experiment to restore the nominal target value response characteristics.In this paper, the effectiveness of this method is verified through experiments using an “arm-type LEGO crane”. Furthermore, two features that hold for a closed-loop system using a controller obtained based on this method are presented. A method for utilizing these properties during implementation is then presented.