Top-down attention control for device communication manager on mobile robot platform

In this paper, we present the results of our recent work about design and implementation of a top-down attention control model in Device Communication Manager (DCM) layer on an omnidirectional mobile robot platform. Attention control is one of the best ways to reduce the complexity of sensory data, whereas in nowadays, we face with various types of sensors and actuators that all of them need to be managed through specific algorithms. Also, most of these sensory data is not usable in raw form, and some may need to be preprocessed for extraction of reliable and practical information. Most researches in this field are about presenting a model for more efficient use of sensory data in specific situations. So, these approaches don't control the hardware directly. Whiles by controlling sensors and actuators directly through attention model, we have increased the accuracy of sensory data, improved the accessibility to actuators and reduced the power consumption of whole system. In this research, we focused on a low-level attention control algorithm to adapt DCM layer with higher-level attention controllers. In this research, we present a top-down attention model based on incoming requests for each device form higher level algorithms and adapt each module's update frequency according to amount of attention on it. This model is inspired by simple sale forecasting techniques in economics like Exponential Moving Average (EMA) and some intuitive constraints to predict user's request rate in future and adapt each module's update frequency. This novel DCM and our method are implemented on ReMoRo mobile robot platform and results confirmed the efficiency of our method, which can improve the accuracy of sensory data and rate of access to actuators and distributed modules in a robot.