Despite their many successes and great computational power and speed, why are machines still so blatantly outperformed by humans in uncertain environments that require flexible behavior through real-time sensorimotor interactions like playing football or navigating a disaster zone? Answering this question requires understanding the mathematical principles of biological sensorimotor control and learning, which is not only essential for understanding the fundamental principles of intelligent behavior, but it is also of potentially great technological value. Over the recent years Bayes-optimal actor models have widely become the gold standard in the mathematical understanding of sensorimotor processing. However, these models become quickly intractable for real-world problems that go beyond simplified and well-controlled laboratory tasks. What is therefore needed is a bounded rational framework of sensorimotor processing that allows unifying action and perception within the same formalism and considers information-processing costs as the fundamental currency. It is the ambition of BRISC to establish such a framework in order to understand how bounded rational control strategies might cope better in real-world scenarios fraught with uncertainty and time pressure. Establishing such a framework requires drawing out theoretical predictions and gathering experimental evidence both in single- and multi-agent sensorimotor processing in humans. In particular, BRISC investigates the hypothesis that limited informational resources drive the development of hierarchical control architectures where different levels of the hierarchy process information with different degrees of abstraction and with different time scales. Understanding how abstract concepts are formed autonomously from the sensorimotor stream based on resource allocation principles will provide a decisive step towards a bounded rational framework for robust and flexible sensorimotor processing.