|
The Wallenberg laboratory
for research on
Information Technology
and Autonomous Systems
|
System Architecture
The first version of the
WITAS UAV command and control architecture (CCSA)
was developed for the
1997 and 1998 demo systems, which are described in a separate menu item
here. Based on the experience of that early work we defined a revised
architecture that is the basis for the continued work and that is
described in
a detailed report
that was published by Linköping University Electronic Press.
The following is a brief summary of the WITAS CCSA. It can be characterized
as a three-layered hybrid deliberative/reactive architecture
with functionality and structure similar in spirit to, but not the same as,
the three-layered architectures proposed by Firby and Gat.
The CCSA contains the following layers which function as
three separate asynchronous
computational processes:
- Deliberative Layer -- The deliberative layer is responsible
for computation intensive
tasks such as plan generation, plan monitoring, high-level episode recognition,
diagnosis, and responding to queries from other parts of the system. Time and space
complexity for tasks executed in the deliberative layer are theoretically
unrestricted, although the nature of the application obviously places bounds on these
parameters.
- Reactive Layer -- The reactive layer is responsible for the contingent
sequencing, parameterization, and execution of tasks associated with the achievement
of temporally extended goals explicitly generated during run-time by the deliberative layer and
implicitly defined at compile-time by the system designers in the reactive layer. Time and
space complexity for tasks executed in the reactive layer are restricted to have much tighter
bounds then the deliberative layer, with very little use of world state and limited time bounds on
computation.
- Process Layer -- The process layer is responsible for the concurrent
computation of collections of feedback control loops tightly coupling sensing with
actuation. Time and space complexity are tightly bounded with essentially no use of world
state and very short sampling periods.
In addition, the CCSA contains a Knowledge and Data Repository (KDR) storing two
types of knowledge and data:
- Static KD -- This part of the KDR contains longer term symbolic knowledge
structures such as those found in Geographic Information Systems, assertional
knowledge bases, or reactive task description repositories.
- Ephemeral KD -- This part of the KDR contains shorter term knowledge
structures and caches used by the different CCSA layers to implement both
low and high-level functionality. For example, the dynamic data structures used
by the image processing system and scene information manager in the vision module
create short term ephemeral KD structures during image interpretation.
In general, much of the knowledge and data used by the system is stored in a local
manner, especially in the lower layers of the architecture. Global
knowledge and data can be retrieved and updated by all three layers of the
CCSA via access to a central data bus, although the lower layers primarily use
and update the ephemeral KD structures and the deliberative layer uses and updates the
static KD structures. In addition, it is the responsibility of the deliberative layer to
fuse ephemeral KD structures with static KD structures when this is deemed useful.
Maintenance information:
Latest update 10.5.2000 by WITAS secretariat.
Edit mode aml, position code D.witas.info.arch.
To proceed, click other item in menue |