Workshop on Informative Path Planning and Adaptive Sampling

Robots rely on models of themselves and the environment to understand and act in the world around them. In many cases, these models are trained on observed (sampled) data, and the goal is to collect the set of data that will generate the most useful model within the resource constraints of the operating robot system. This process is known as adaptive informative sampling, and is applicable to a wide range of robotic applications, from modeling environmental phenomena to approximating value functions in reinforcement learning. However, despite the prevalence of adaptive sampling methods in state-of-the-art robotic applications, there are still many challenging and open problems. For example, how should we estimate the utility of future samples, or design information sharing protocols for a multi-robot team such that they can effectively reason over their joint sampling action? The main goal of this workshop is to discuss and share ideas related to informative path planning and adaptive sampling. This is a topic that spans all robotic domains and we want to bring together researchers from all fields—marine, ground and aerial robotics, as well as the multi-agent and learning communities—who might otherwise not be aware of the valuable techniques being developed in each of these domains, and the correspondence between their research. This workshop will look at the various aspects of informative path planning, including, but not limited to, its theoretical foundations, active sampling, spatio-temporal variability, and multi-robot planning. One of the aims of this workshop is to generate candid discussion on how these theoretical components interact with real world robotic constraints such as imperfect sensing and resource limitations. To that end, we also invite talks from potential end-users; e.g. oceanographers, agriculturists, meteorologists. Topics of interest Theoretical foundations of informative path planning and adaptive sampling Exploration, mapping, surveillance, and inspection missions in unknown or dynamic environments Persistent environmental monitoring, particularly methods for handling spatial and/or temporal variability Information sharing and data fusion for multi-robot teams Coordination algorithms for multi-robot missions Handling real robot constraints, such as large volumes of data or communication constraints Budgeted sampling under resource limitations Employing these methods in practice for real-world applications