Understanding the environment is an essential ability for robots to be autonomous. In this sense, Convolutional Neural Networks (CNNs) can provide holistic descriptors of a scene. These descriptors have proved to be robust in dynamic environments. The aim of this paper is to perform hierarchical localization of a mobile robot in an indoor environment by means of a CNN. Omnidirectional images are used as the input of the CNN. Experiments include a classification study in which the CNN is trained so that the robot is able to find out the room where it is located. Additionally, a transfer learning technique transforms the original CNN into a regression CNN which is able to estimate the coordinates of the position of the robot in a specific room. Regarding classification, the room retrieval task is performed with considerable success. As for the regression stage, when it is performed along with an approach based on splitting rooms, it also provides relatively accurate results.
29 September 2021 in Artificial Intelligence Review
This paper reports and evaluates the adaption and re-training of a Convolutional Neural Network (CNN) with the aim of tackling the visual localization of a mobile robot by means of a hierarchical approach. The proposed method addresses the localization problem from the information captured by a catadioptric vision sensor mounted on the mobile robot. A CNN is adapted and evaluated with a twofold purpose. First, to perform a rough localization step (room retrieval) by means of the output layer. Second, to refine this localization in the retrieved room (fine localization step) by means of holistic descriptors obtained from intermediate layers of the same CNN. The robot estimates its position within the selected room/s through a nearest neighbour search by comparing the obtained holistic descriptor with the visual model of the retrieved room/s. Additionally, this method takes advantage of the likelihood information provided by the output layer of the CNN. This likelihood is helpful to determine which rooms should be considered in the fine localization process. This novel hierarchical localization method constitutes an efficient and robust solution, as shown in the experimental section even in presence of severe changes of the lighting conditions.
This paper presents an extended study about the compression of topological models of indoor environments. The performance of two clustering methods is tested in order to know their utility both to build a model of the environment and to solve the localization task. Omnidirectional images are used to create the compact model, as well as to estimate the robot position within the environment. These images are characterized through global appearance descriptors, since they constitute a straightforward mechanism to build a compact model and estimate the robot position. To evaluate the goodness of the proposed clustering algorithms, several datasets are considered. They are composed of either panoramic or omnidirectional images captured in several environments, under real operating conditions. The results confirm that compression of visual information contributes to a more efficient localization process through saving computation time and keeping a relatively good accuracy.
In this paper, a hierarchical localization framework within indoor environments is proposed and evaluated, considering severe variations of the illumination conditions. The only source of information both to build a model of the environment and to solve the localization problem is a catadioptric vision system, which is mounted on the mobile robot. The images captured by this system are processed globally to obtain holistic descriptors. The position of the robot is estimated by comparing these descriptors with the information contained in a topological visual model, which is previously created using a clustering approach and is composed of a hierarchy of layers. Compacting the information via clustering proves to be an efficient alternative to estimate the position of the robot hierarchically and with robustness. The proposed localization strategy is tested with some sets of panoramic images, captured in large indoor environments under real operating conditions, including illumination changes that change substantially the appearance of the scenes. The results show a reasonable tradeoff computation time-accuracy when the localization is addressed in a hierarchical way.