ModPlan - Modern Action Planning
Validation
For plan validation we have included VAL, as
provided by the Strathclyde planning group. The main capabilty of VAL
is the simulation, i.e. the execution of almost any plan in PDDL
syntax. It has recently
been extended to capture continous effects, exogenous events
and processes.
Temporal Plan Visualization
For plan visualization we included the
animation system Vega allowing a
magnification to an arbitrary part of the plan. Vega is
implemented as a Client-Server architecture that runs an annotated
algorithm on the server side to be visualized on the client side in a
Java frontend. Annotation are visualization requests that (minorly)
extend the existing source code by (simple) commands like send
point(x,y). In the system, visualization objects can be
associated with hierarchical structured identifiers. The client is
used both as the user front-end and the visualization engine. Thus, it
allows server and algorithm selection, input of data, running and
stopping algorithms, and customization of the visualization.
Vega allows both on-line and off-line presentations.
For the latter only the client is needed.
Gantt charts are plots for temporal plans, in which a horizontal open
oblong is drawn against each activity indicating estimated
duration. To access precise e.g. temporal information on the operators
representatives can selected with the mouse. Our visualization module
depicts the Gantt chart of plans in competition format. As the
essence of this visualization task is translating temporal operators
into rectangles and associated text labels, we were able to visualize
all plans generated at the 2004 planning competition in a scalable
Gantt chart with Vega.
Domain Visualization
Domain-dependent visualization is more challenging. As the semantics
of predicates are not known in advance, domain experts assist the
animation design. From a knowledge engineering aspect, the efforts
have to be as small as possible.
Propositional
atoms can often be illustrate by showing or hiding an image at a
certain location, while numerical quantities such as fuel can best be
expressed by using scalable graphical items.
The according images for displaying domain objects are
collected with an image web search engine.
Previously, the domain visualizer worked in cooperation with a planner
extension that wrote (sequential) plans together with associated
state information to disk. Currently, we exploit VAL's capabilities to allow
visualization of parallel and temporal plan execution. For this
purpose, we have extended the validator to flush a sequence of states at
each happening together with its time stamp for further
post-processing.
|