Partners: HKUST RI & KTH RPL
Project description: For the purpose of making robots autonomous and be able of interacting with the environment, the ability to sense and to understand the environment is integral. As humans, robots can use various types of sensing, even some not present in humans, such as distance measuring sensors. One of the challenges has been to develop control and decision-making algorithms that can in an optimal manner use information from several sensory modalities and also do this in real time. As such, given various sensory data, we will in RoMRO explore geometrical and topological tools to facilitate the design of effective representations, which will enable the robots to better understand the task and the problem, and in turn to boost the optimization of their planning and executions. In particular, we will focus on developing representations that allow us to exploit the geometrical variances among the encountered particular problems to make the system adaptive, and also to capture the topological invariances to be used for understanding higher-level semantics that better generalize over problems.