This paper deals with the problem of multiple robots working together to explore and gather at the global maximum of the unknown field. Given noisy sensor measurements obtained at the location of robots with no prior knowledge about the environmental map, Gaussian process regression can be an efficient solution to construct a map that represents spatial information with confidence intervals. However, because the conventional Gaussian process algorithm operates in a centralized manner, it is difficult to process information coming from multiple distributed sensors in real-time. In this work, we propose a multi-robot exploration algorithm that deals with the following challenges: 1) distributed environmental map construction using networked sensing platforms; 2) online learning using successive measurements suitable for a multi-robot team; 3) multi-agent coordination to discover the highest peak of an unknown environmental field with collision avoidance. We demonstrate the effectiveness of our algorithm via simulation and a topographic survey experiment with multiple UAVs.

The event-triggered model predictive control (MPC) reduces energy consumption for updating control sequences while maintaining the originality of the MPC, which copes with hard constraints on dynamical systems. In the presence of large uncertainties, however, the standard event-triggered MPC may generate too frequent event occurrences. To compensate for unknown uncertainties, this paper applies a statistical learning to event-triggered MPC. The stability and the feasibility of the proposed control system are analyzed in regard to the statistical learning influences, such as the number of training samples, model complexity, and learning parameters. Accordingly, the event-triggering policy is established to guarantee the stability. We evaluate the proposed algorithm for the tracking problem of a nonholonomic model perturbed by uncertainties. In comparison with the standard event-triggered control scheme, the simulation results of the proposed method show better tracking performance with less frequent event triggers.

This letter presents a combination of reinforcement learning (RL) and deterministic controllers to learn a quadrotor control. Learning the quadrotor flight in a standard RL approach requires many iterations of trial and error, which may bring about risky exploration and battery consumption. In this letter, we integrate a classical controller such as PD (proportional and derivative) or LQR (linear quadratic regulator) with a RL policy using their linear combination. The proposed method is not only simple to use, but also fast in learning convergence. When the algorithm is evaluated for a quadrotor trajectory tracking by means of a velocity control for both simulation and experiment, it demonstrates the faster convergence rate and better control performance in comparison with an existing rapid model-based RL method.

Accommodation of a situation when a prior map is not available in an indoor localization system is valuable to cost-effective operations by removing a need for map drawing and map updating. This paper suggests a trajectory learning method using crowdsourced measurements in order to support the absence of map. A localization framework based on a particle filter is formalized by machine-learning-based feature extraction and Gaussian process (GP) regression. The feature extraction algorithm reduces dimensionality of sparse measurement vector, and it is applied to detect floor level and designated landmarks. Also, the combination of the feature extraction and the GP regression is used for modeling nonlinear relationship between location and measurement. By this combination, locations of Wi-Fi access points are not required to be known. From the field experimental results, we confirm that the detections of floor level and landmarks are accurate, the learned trajectories are close to the true map, and positioning accuracy is improved thanks to the learning-aided localization.

A set of Wi-Fi RSSI (Received Signal Strength Indicator) measurements is one of basic sensory observation available for indoor localization. One major drawback of the RSSI based localization is maintenance of the RSSI fingerprint database, which should be periodically updated against measurement pattern changes caused by relocation, removal and malfunction of Wi-Fi APs (access points). To address this problem, a new change detection method is proposed in this paper. First, by machine learning techniques, the RSSI database is reconstructed to a probabilistic feature database by the implementations of PCA (Principal Component Analysis) and GP (Gaussian Process). Then, KL (Kullback-Leibler) divergence is used as a metric to measure the similarity of the existing database and a newly arrived test sets. The proposed method is evaluated by a real experiment at a multi-storey building. For experimental study, different cases that provoke changes of RSSI patterns are considered, and the positioning accuracy is examined by the k-NN (Nearest Neighbor) method. From the experimental results, it is found that the bigger the RSSI pattern changes, the large the KL divergences become. Also, when a modified change detection algorithm as the benchmark, which does not implement the PCA feature extraction, is compared, the proposed algorithm yields accurate and fast computing performances. In addition, the required number of survey points is empirically found associated with the threshold value to trigger the detection alarm.

This paper studies the indoor localization based on Wi-Fi received signal strength indicator (RSSI). In addition to position estimation, this study examines the expansion of applications using Wi-Fi RSSI data sets in three areas: (i) feature extraction, (ii) mobile fingerprinting, and (iii) mapless localization. First, the features of Wi-Fi RSSI observations are extracted with respect to different floor levels and designated landmarks. Second, the mobile fingerprinting method is proposed to allow a trainer to collect training data efficiently, which is faster and more efficient than the conventional static fingerprinting method. Third, in the case of the unknown-map situation, the trajectory learning method is suggested to learn map information using crowdsourced data. All of these parts are interconnected from the feature extraction and mobile fingerprinting to the map learning and the estimation. Based on the experimental results, we observed (i) clearly classified data points by the feature extraction method as regards the floors and landmarks, (ii) efficient mobile fingerprinting compared to conventional static fingerprinting, and (iii) improvement of the positioning accuracy owing to the trajectory learning.