watch what happens. By changing the location of the cubes and the settings on the slider, they can refine their
simulation — dynamically. They can fine-tune a specific area of the virtual heritage site, just addressing a few
NPCs, while the rest of the NPCs are happily moving around and taking the existing environment and current layout
of cubes into account. Indeed, historians can see NPCs approaching a specific cube, clearly with the intention of
sitting on top of it, and just delete the cube — or place another one near the NPC — just to see what happens. Or
they can delete a building, block a street, demolish a wall, and see how NPCs immediately take these changes into
account. The simulation adapts dynamically to everything that happens in the environment.
A further development is to have ‘combined boxes’, which can make the simulation even more interesting.
Consider the following scenario: a church provides happiness, but it provides more happiness if a priest is present.
To model this, historians drop a money cube for the priest (meaning that priests earn their ‘daily wages’ by holding
mass, and that acts as their motivation for being inside the church), as well as a few red cubes for happiness. When a
priest NPC sits on its money cube, the red cubes get ‘activated’ — for instance, their rate of bestowing happiness
increases dramatically. The expected behaviour of the rest of the NPCs is that they will try to prefer entering a
church with a priest holding mass, because the rate at which happiness is increased will weigh more on the fitness
function of the GA. But there might be a limit of just a few cubes available in the church; what will the GA select? It
could wait until the cubes are free — but by then, the priest might have reached its ‘daily wage’ and leave its place,
and thus ‘deactivating’ the happiness cubes. Similar approaches can be used to simulate a tavern or market stall:
when the innkeeper is at work, the energy cubes are active, but when he’s away, they are not.
‘Combining’ cubes that way requires a bit more work for the historians. Once they are dropped, they could be
assigned to a certain class (for instance, only for priests), which is accomplished via the SL/OpenSimulator
mechanism of sending a dialogue box with options to the 3D viewer. A further option would be to ‘combine’ nearby
cubes, which can either be accomplished automatically (all cubes in a certain radius) or by inviting the historians to
click on each cube that they wish to ‘link’ with the one currently be confirmed. This can even be done visually, by
using particle beams linking the combined cubes, thus giving historians some feedback on what has been configured
so far.
8. Conclusions and avenues for further work
Turning the visual programming of complex crowd simulation by historians into something very much like a
game has some interesting possibilities. While the prototype, at the time of writing, is still far from functional, and
has not underwent user acceptance testing, it should allow easy configuration of the whole simulation, and thus fulfil
the initial requirements of the project. This, however, has not yet been validated.
Unlike strict rule-based systems, this model is expected to be much less ‘predictable’, while still maintaining
‘realism’: NPCs navigate the virtual world according to goals and motivations, and not randomly, but their
behaviour, due to the ever-changing, dynamic nature of past interactions, will not be deterministic. The way that the
crowd simulation is configured, by dropping cubes and setting sliders, is quite similar to many strategy/simulation
games (like the popular The Sims). This can be used to develop educational games.
Further research could also focus on the ‘cubes’ themselves. Right now, these are supposed to be relatively static,
and remain at the place where the historian has dropped them. But as was explained before, an interesting case is
when the cube is picked up by an NPC and carried around. Thus, future development might have moving cubes (not
‘carried’ but autonomously moving); or cubes might interact by being attracted/repelled to each other.
One possible complication with this approach is the requirement of having many coloured cubes spread all over
the scene of the simulation, being potentially difficult for historians to track the overall model. In fact, it can be
argued that such a method is not practical for large-scale simulations with hundreds or thousands of intelligent
agents. However, in the case of SL/OpenSimulator, system constraints limit the number of agents in a scene to a few
dozens; thus, the proposed framework is expected to be adequate for small-scale crowd simulation in
SL/OpenSimulator without being too confusing to follow for the historians setting it up, but might require a different
approach for other platforms able to model large-scale crowds.
Also, this model is assuming somewhat ‘modern’ perspectives, due to its urban 18th century application scenario:
concepts such as ‘work’ and ‘money’ may not be adequate motivators in earlier settings or, indeed, in rural settingsof the same century. Defining sets of motivational/control cubes is a probable later requirement of historians and
watch what happens. By changing the location of the cubes and the settings on the slider, they can refine their
simulation — dynamically. They can fine-tune a specific area of the virtual heritage site, just addressing a few
NPCs, while the rest of the NPCs are happily moving around and taking the existing environment and current layout
of cubes into account. Indeed, historians can see NPCs approaching a specific cube, clearly with the intention of
sitting on top of it, and just delete the cube — or place another one near the NPC — just to see what happens. Or
they can delete a building, block a street, demolish a wall, and see how NPCs immediately take these changes into
account. The simulation adapts dynamically to everything that happens in the environment.
A further development is to have ‘combined boxes’, which can make the simulation even more interesting.
Consider the following scenario: a church provides happiness, but it provides more happiness if a priest is present.
To model this, historians drop a money cube for the priest (meaning that priests earn their ‘daily wages’ by holding
mass, and that acts as their motivation for being inside the church), as well as a few red cubes for happiness. When a
priest NPC sits on its money cube, the red cubes get ‘activated’ — for instance, their rate of bestowing happiness
increases dramatically. The expected behaviour of the rest of the NPCs is that they will try to prefer entering a
church with a priest holding mass, because the rate at which happiness is increased will weigh more on the fitness
function of the GA. But there might be a limit of just a few cubes available in the church; what will the GA select? It
could wait until the cubes are free — but by then, the priest might have reached its ‘daily wage’ and leave its place,
and thus ‘deactivating’ the happiness cubes. Similar approaches can be used to simulate a tavern or market stall:
when the innkeeper is at work, the energy cubes are active, but when he’s away, they are not.
‘Combining’ cubes that way requires a bit more work for the historians. Once they are dropped, they could be
assigned to a certain class (for instance, only for priests), which is accomplished via the SL/OpenSimulator
mechanism of sending a dialogue box with options to the 3D viewer. A further option would be to ‘combine’ nearby
cubes, which can either be accomplished automatically (all cubes in a certain radius) or by inviting the historians to
click on each cube that they wish to ‘link’ with the one currently be confirmed. This can even be done visually, by
using particle beams linking the combined cubes, thus giving historians some feedback on what has been configured
so far.
8. Conclusions and avenues for further work
Turning the visual programming of complex crowd simulation by historians into something very much like a
game has some interesting possibilities. While the prototype, at the time of writing, is still far from functional, and
has not underwent user acceptance testing, it should allow easy configuration of the whole simulation, and thus fulfil
the initial requirements of the project. This, however, has not yet been validated.
Unlike strict rule-based systems, this model is expected to be much less ‘predictable’, while still maintaining
‘realism’: NPCs navigate the virtual world according to goals and motivations, and not randomly, but their
behaviour, due to the ever-changing, dynamic nature of past interactions, will not be deterministic. The way that the
crowd simulation is configured, by dropping cubes and setting sliders, is quite similar to many strategy/simulation
games (like the popular The Sims). This can be used to develop educational games.
Further research could also focus on the ‘cubes’ themselves. Right now, these are supposed to be relatively static,
and remain at the place where the historian has dropped them. But as was explained before, an interesting case is
when the cube is picked up by an NPC and carried around. Thus, future development might have moving cubes (not
‘carried’ but autonomously moving); or cubes might interact by being attracted/repelled to each other.
One possible complication with this approach is the requirement of having many coloured cubes spread all over
the scene of the simulation, being potentially difficult for historians to track the overall model. In fact, it can be
argued that such a method is not practical for large-scale simulations with hundreds or thousands of intelligent
agents. However, in the case of SL/OpenSimulator, system constraints limit the number of agents in a scene to a few
dozens; thus, the proposed framework is expected to be adequate for small-scale crowd simulation in
SL/OpenSimulator without being too confusing to follow for the historians setting it up, but might require a different
approach for other platforms able to model large-scale crowds.
Also, this model is assuming somewhat ‘modern’ perspectives, due to its urban 18th century application scenario:
concepts such as ‘work’ and ‘money’ may not be adequate motivators in earlier settings or, indeed, in rural settingsof the same century. Defining sets of motivational/control cubes is a probable later requirement of historians and
การแปล กรุณารอสักครู่..
