Active state estimation for high degrees of freedom robotic applications

Planning actions under uncertainty has always played a significant part in robotics. As robots become more prevalent in dynamic home, industrial, and harsh environmental settings, they will be increasingly required to make independent decisions about how to accomplish their tasks, very often under (severe) uncertainty. Sources of uncertainties in robotics are derived from sensor noise, uncertain actuators, and uncertainties in environmental scenes. When equipping robots with an intelligence, a key objective is to achieve specific tasks that handle with these uncertainties in a robust way. In the past, several techniques have been developed decision making. However, these techniques are computational very intractable since they ensure that the consumption of computational load and memory are very high, which results in large computing times, especially in applications with high degrees of freedom. The main objective of this research project is to develop a new framework that decreases the computation time drastically by using an active sensing approach. By actively controlling the sensors, active sensing tries to decrease the uncertainties described in the paragraph above. This framework could be attractive for a large variety of applications and industries where uncertainty in high degrees of freedom are common, e.g. mobile manipulators, eCommerce AGVs, or electrical wheelchairs.