|
|
|
| e-Pub |
Previous | 
Home
Section: New Results
Synchronous Approach to Activity Recognition
Participants : Daniel Gaffé, Sabine Moisan, Annie Ressouche, Jean-Paul Rigault, Ines Sarray.
Activity Recognition aims at recognizing and understanding sequences of actions and movements of mobile objects (human beings, animals or artefacts), that follow the predefined model of an activity. We propose to describe activities as a series of actions, triggered and driven by environmental events.
Due to the large range of application domains (surveillance, safety, health care ...), we propose a generic approach to design activity recognition systems that interact continuously with their environment and react to its stimuli at run-time. Such recognition systems must satisfy stringent requirements: dependability, real time, cost effectiveness, security and safety, correctness, completeness ... To enforce most of these properties, our approach is to base the configuration of the system as well as its execution on formal techniques. We chose the Synchronous Approach which provides formal bases to perform static analysis, verification and validation, but also direct implementation.
Based on the synchronous approach, we designed a new user-oriented activity description language (named ADeL) to express activities and to automatically generate recognition automata. This language relies on two formal semantics, a behavioral and an equational one [48]. We also developed a component, called Synchronizer, to transform asynchronous sensor events into synchronous “instants”, necessary for the synchronous approach. This year, we mainly worked on the ADeL compiler to generate synchronous automata, on the graphical tool of this language and on the Synchronizer component.
ADeL Compilation:
To compile an ADeL program, we first transform it into an equation system which represents its synchronous automaton. Then we directly implement this equation system, transforming it into a Boolean equation system. This equation system provides an effective implementation of the initial ADeL program for our runtime recognition engine. The internal representation as Boolean equation systems also makes it possible to verify and validate ADeL programs, by generating a format suitable for a dedicated model checker such as the off-the-shelf NuSMV model-checker.
Synchronizer:
The role of the Synchronizer is to filter physical asynchronous events, to decide which ones may be considered as “simultaneous” and to aggregate the latter into logical instants. The sequence of these instants constitutes the logical time of our recognition systems. The runtime recognition engine interacts with the synchronizer and uses these instants to run the automata corresponding to the activities currently recognized. In general, no exact decision algorithm exists but several empirical strategies and heuristics may be used e.g., for determining instant boundaries. This year we completed the specification and implementation of a first version of the Synchronizer. It is parametrized by heuristics to manage events and data coming from various sensors, to define instant boundaries, and to cope with possible high level interruptions (preemptions).
Moreover, to facilitate the job of the synchronizer (to build the instants) and of the runtime engine (to wake up only the relevant automata), each automaton provides information about the awaited events at each state, i.e the events which may trigger transitions to a next state. The ADeL compiler has in charge to generate this information. In a first attempt, we computed statically all the awaited events in all states of an automaton. However, this approach implied to build the entire explicit automaton from an equation system, which was not realistic. Thus, this year we added specific equations to the equation systems of the operational semantics to compute the awaited events of each operator of the language. The information about next awaited events is now computed at runtine, when a state of the automaton is reached.