 |
Smart Its Project
Interconnected Embedded Technology for Smart Artefacts with Collective AwarenessIn this project the PCCV group will be responsible for the Development of perceptual computing methods for ad hoc connected sensors. In particular we will investigate the following topics:
Collective Perception and Awareness of Smart-ItsTraditionally, perceptual computing has been focused around audio and vision sensors employing computationally expensive methods for solving relatively constrained problems. Context awareness for ubiquitous and wearable computing not only demands for a large diversity of sensors but also for new methods how to process and combine information coming from a multitude of sensors. In contrast, the proposed Smart-Its approach will investigate perceptual computing based on the integration of a large and diverse set of simple sensors in a distributed system of low-end processing units. The innovative aspects are therefore distributed sensor integration from a large and diverse set of sensors, low-cost perceptual computing and collective perception in ad hoc collections of sensors.Microphones, cameras as well as many other sensors become so cheap that one can easily imagine to employ a large number of many different sensors in a distributed fashion. In this project we will investigate which sensors can be used to make objects more contextually aware. We propose to build different such sensor types into the Smart-Its. The ultimate goal is to combine the sensor and context information coming from the diverse sensors and coming from different Smart-Its in a distributed way. The investigation of simple feature extraction mechanisms will enable to implement them directly onto the Smart-Its. More complicated feature extraction mechanisms can be implemented on a set of Smart-Its. These different types of features are eventually combined in order to obtain higher level context information. This project investigates in particular how one can distribute this sensor integration process and how higher level context information may emerge from simple integration behaviours. The model for collective perception is to give sensor-based awareness devices access to sensor-based context of other devices in a distributed system. In this model perception methods will operate on features exported locally through a generic context API, and remotely through a communication API. The model assumes a dynamic system in which collections of awareness devices change over time. This requires investigation of perception methods that scale with number of devices involved and that still function when a device is temporarily isolated from others ('graceful degradation'). Publications:
Resources:
Contact: |
|
