Many robotic systems use active perception methods to sense the environment and execute specific tasks with pre-defined signal processing chains. The paradigm of active perception tries to enhance perceptual performance by dynamically interacting with the signal processing chain and their containing algorithms. In this article we propose a concept that uses such active perception techniques to perform mission relevant tasks onboard UAVs. The concept utilizes a resource base containing information on available active resources such as computer vision algorithms, hardware resources, available sensors and actuators as well as background knowledge such as previously trained models or geographic information. This resource base also includes attributes on applicability, capability and limitation of each item. Our idea is to generate a signal processing chain dynamically dependent on current mission needs. This is achieved by understanding which resources are likely to fulfill a specific mission task. Thereby specific focus shall be given to effectively balancing and distributing the resulting processes to be performed on available hardware resources, while taking into account limited processing power and energy supply.    «

Many robotic systems use active perception methods to sense the environment and execute specific tasks with pre-defined signal processing chains. The paradigm of active perception tries to enhance perceptual performance by dynamically interacting with the signal processing chain and their containing algorithms. In this article we propose a concept that uses such active perception techniques to perform mission relevant tasks onboard UAVs. The concept utilizes a resource base containing informatio...    »