TITLE: Applying Metric Interval Temporal Logic to Robot Monitoring

PRESENTER: Prof. Froduald Kabanza

AFFILIATION: Computer Science Department, University of Sherbrooke, Quebec, Canada

DATE: Thursday 18 November, 2004

TIME:12 - 1pm

PLACE: CSE K17 1st Floor Seminar Room

ABSTRACT:

The behavior-based design principle indicates that decision making in a mobile robot should be split among several concurrent processes each of them being in charge of a simpler aspect of the global decision making process. Exhaustive debugging or off-line validation cannot guarantee, however, the absence of failures in robot behaviors such as obstacle avoidance, robot localization and object tracking, because these behaviors rely on real-time data that cannot be accurately predicted off-line. Therefore online robot monitoring becomes necessary to detect failures and launch recovery processes or even avoid failures by anticipating their occurrence and launching processes that prevent them from happening. However, writing reliable and logically consistent monitoring processes is often a very difficult task particularly because the behavior-based design can lead to unpredictable concurrent executions of robot behaviors.

In this talk I will discuss the merits of Metric Interval Temporal Logic (MITL) as a language for specifying robot monitoring conditions. One issue that I will discuss is whether it is easier to specify MITL monitoring properties than to explicitly write equivalent robot monitoring programs. Another issue is, given that MITL monitoring processes are constructively proven consistent with respect to the monitored behavior (as a result of the logical semantics of MITL), does this lead to higher reliability in the robot's behavior compared to using explicitly programmed monitors.