Implementation of asymmetric multiprocessing framework in humanoid robot

In this paper, the characteristics and implementation of a new asymmetric multiprocessing (AMP) framework in humanoid robot are discussed. This proposed framework is used to replace computer network as the high level processing unit for the humanoid robot that is designed to perform object localization based on visual and auditory information. In this AMP framework, a multi-core computer is divided into several smaller virtual machines that own a part of the physical resources including processing core, memory and input/output (I/O) devices. Each virtual machine executes a guest operating system (OS) and dedicated applications, including colored object localization, sound source localization and multisensory information fusion and motion control. Xen paravirtualization technology is used to conveniently manage these guest OSes.