The aim of this project is to integrate runtime diagnosis of hardware
and software of autonomous robot systems in a unified way. For this, a
common model needs to be developed on which to apply model-based
diagnosis techniques. Autonomous mobile robots represent a
particularly well suited application for diagnosis tasks. Their
autonomous nature implies that they need to reason about their state
as there might not always be a supervising human operator. A robot
should at all times aim to maintain functionality or try to retain as
much system functionality as possible. Repair actions might lead to a
fully functioning state after a fault has been diagnosed.
Reconfiguration might retain functionality to a certain degree. In the
worst case, the robot might reconfigure itself to switch to a
fail-safe mode to avoid further damage or threats to nearby people.
For example, diagnosis that the vision sensor is malfunctioning could
lead to a reconfiguration where other sensors, e.g., a laser range
scanner, are used. This might also require to use a different mode of
operation, i.e., software reconfiguration. If, however, the video
device driver causes problems while the actual video device is still
fully functioning, a restart of the video device driver and all
dependand services might be a sufficient repair action to restore
normal behaviour. As can be seen from this example hardware and
software are tightly coupled and need to be modeled together.
In this project the feasibility of a unified diagnosis approach is to
be evaluated. The modular robotic platform created at Graz Technical
University will serve as a test-platform for case-studies on modelling
hardware, software, interfaces between hardware and software and
interfaces between the abstract control and diagnosis layers. These
robots are used for office delivery tasks and robotic soccer, and will
serve as target for a case-study. Diagnosis on these models is to be
integrated with repair and reconfiguration.