ACES research

Pervasive computing

Members

Michel Banātre (INRIA Research Director), Paul Couderc (INRIA Researcher), Mathieu Becus (INRIA Engineer), Julien Pauty (Ph.D. Student)

Research :

An important trend in computing is the spreading of computing systems in the nature and their ubiquity in everyday activities. Several factors are driving this trend:

• the success of mobile phones, which effectively make computing system and information services available virtually anywhere
• the universality of the internet as an interface to information and computing services. More and more devices are "WWW pluggable" in the sense that they are network aware and offers Web interfaces.
• wireless communication technologies are becoming very common, providing seamless global network connectivity to both users and devices, as well as allowing spontaneous cooperation of mobile devices.
• sensing technologies, such as positionning, image monitoring, tagging allow computing systems to automatically capture data and build representation of the real world. This feature provides environment awareness to services and application, freeing the user attention from the interactions with the devices (invisible computing).

The aspect in which we are especially interested about pervasive computing is the coupling of computing processes and the real world. We focus on how the real world should be modeled at the system and application levels, and how real world events and beheavior of physical entities should be mapped in the computing domain. The motivation for this research is that processing in applications for mobile computing is often related to interactions of physical objects. A typical trend in system design is to hide physical aspects such as instability and dynamicity raised by mobility and providing to applications the illusion of a standard (static) computing environment. Our approach is at the opposite: our goal is to exploit physical interactions of objects to convert them into actual information processing. Consider for example a real shopping cab. It can be seen as a data structure allowing the calculation of a price from a set of objects. The data structure is defined spatially by the volume including the articles stored in the cab. Adding an article in the cab or removing one can be considered as implicit data processing (ie. updating the current total price). The two key benefits to this approach (that we call spatial computing) are:

1. simpler application programming and system design, as physical objects are considered as data symbol and we re-use into the data processing some processing that implicitly exist at the physical level.
2. the approach results in efficient and extensible solutions, as it favor processing at the local level by attaching semantic to physical interactions.

• Modeling the notion of context, and providing system support around this abstraction
• The design of sensing system (entities and/or data discovery, spatial awareness, event reporting...)
• Analyzing and optimizing the performance or cost of our solutions in terms of communication load and energy use, as our approach is typically dedicated to autonomous devices depending on battery.
  
  

Existing system : SPREAD

SPREAD (Spatial PRogramming Environment Ambient computing Design) is a software system enabling simple and elegant programming of ubiquitous computing applications. SPREAD is based on the concept of physical ``tuples'', which are structured piece of data associated to a shape surrounding a physical object. SPREAD allow processes to use the physical space as a simple database or associative memory, where data are represented by physical objects and data flow are implicitely related to object movements.

The system offers a simple yet powerful API to applications. This API allows to read and write tuples in the physical space, and to synchronize operations. SPREAD is based on a wireless peer to peer architecture, where all the participating objects host an instance of the system and the relevant application components. The targeted execution platform for SPREAD nodes are very low cost, low power devices integrating a cpu, some memory, and a short range wireless communication capability.

Currently, there are two implementation of the SPREAD system. The first one is a native Windows CE 3.0 version. PocketPC PDAs can be use for application prototyping, using either WLAN (IEEE 802.11) wireless interface or Bluetooth radio.

The second version is written in Java, and run on J2ME compliant platform.

SPREAD has been used to develop and experiment several ubquitous computing application including Ubi-Bus, an operational urban transportation assistant for disabled people, and Web Walker, a contextual Web system.

SPREAD has been deposited as an INRIA Software

Publications

• See ACES main publication page for academic work