Robots often require a model of their environment to make informed decisions. In unknown environments, the ability to infer the value of a data field from a limited number of samples is essential to many robotics applications. In this work, we propose a neural network architecture to model these spatially correlated data fields based on a limited number of spatially continuous samples. Additionally, we provide a method based on biased loss functions to suggest future areas of exploration to minimize reconstruction error. We run simulated robotic information gathering trials on both the MNIST hand written digits dataset and a Regional Ocean Modeling System (ROMS) ocean dataset for ocean monitoring. Our method outperforms Gaussian process regression in both environments for modeling the data field and action selection.

Robots often operate in built environments containing underlying structure that can be exploited to help predict future observations. In this work, we present a deep learning based approach to predict exit locations of buildings. This technique exploits the inherent structure of buildings to create a model. A convolutional neural network is trained using a database of building blueprints and used to guide a search within a building. This technique is compared to standard frontier exploration and a traditional image processing approach of extracting features through histogram of gradients (HOG) and training a support vector machine (SVM). After validation through simulation, we show that the proposed deep learning technique reduces the amount of building exploration required to find the goal by 36%.