Social Aspect, Relationship, Mission, Activity, Project

[mokun]

What do you guys see how important "relationship" should be in mars-sim ?

Currrently, the use of the relationship score is somewhat limited. See Relationship wiki

Do we want to leave it this way ?

What would be the one social aspect you would like to add ?

btw, we should also consider removing the use for the phoenixt's maven package (plexus-core) in both RelationshipManager and CreditManager.

plexus-core is an old and outdated graphing API.

import com.phoenixst.plexus.DefaultGraph;
import com.phoenixst.plexus.EdgePredicate;
import com.phoenixst.plexus.EdgePredicateFactory;
import com.phoenixst.plexus.Graph;
import com.phoenixst.plexus.GraphUtils;
import com.phoenixst.plexus.NoSuchNodeException;
import com.phoenixst.plexus.Traverser;

If we are going to transition to save in JSON, we may as well get rid of the dependency on plexus-core and come up with our own in-house data structure to store those relationship scores,

or we'll need to create a translator class for converting those graph objects into JSON and vice versa.

[Urwumpe]

I think it is very important, but I don't think it is well represented by the current data structure. It should better be a local property of every Settler. Maybe we can migrate to something Dwarf Fortress like and instead of storing a weighted graph (which is better for simulating a large number of people on a single CPU), we store how a Settler thinks about the people he heard from or interacted with.

This way, we could also safely migrate a Settler to be processed on an other node in a Cluster, instead of having to transmit many updates on the global relationship graph. And it could put some more action into Commander mode, if the social aspect is also important for the success.

If we really need to see such a graph, we should "hire" an psychologist settler to produce this as part of his work.

[mokun]

I think it is very important, but I don't think it is well represented by the current data structure. It should better be a local property of every Settler.

Yes. It's definitely a cleaner approach.

Just finished moving the relationship data into the Mind of a person. See commit 1 and commit 2

[mokun]

instead of having to transmit many updates on the global relationship graph.

There's no more global relationship graph from now on.

[Urwumpe]

What would be needed to allow long-distance communications with settlers in other Settlements? We should get to something like a hierarchy of communication options to be considered.

1. Talk directly, locally, in the same spot.
2. Local network radio or phone call.
3. Remote radio call
4. Visit remote location physically as a business trip.

Also, setting up such a business trip should maybe be a kind of mission itself. The settler might join others on the transport mission, but there would be a purpose, why this settler goes there and the settler will organise a transport mission again for returning home. And of course, plan to include as many activities in a short time in such a business trip, even at the expense of recreation (some people might even cut short sleep then...innocent whistling)

Can we support a "Mission of missions?" Or would that be too complex? The alternative, of a separate independent mission, would be wasteful for the resources. Even a small rover could take passengers with it and you should never depart with less than 2 people onboard (also for safety)

And again, I wish for some ways of exception handling. Can we maybe "teach" settlers certain activities to take place under certain triggers and conditions? And adapt those sets of exception handlers for new conditions, like "When travelling in a rover and somebody on EVA is incapacitated, then perform the 'Recover astronaut into current rover" activity."

[mokun]

We should get to something like a hierarchy of communication options to be considered.

Currently, we pretty much assume that if a scientist needs collaborators or assistants on a scientific study, he will find them globally.

Other than that, there's no communication per se between settlers from different settlements.

Also, trading or delivery mission could be considered an official contact between two settlements when negotiating the value and price of goods.

It was done by selecting a trader from each settlement and letting the two decides what to trade.

There isn't any inter-settlement communications so far.

[mokun]

1. Talk directly, locally, in the same spot.
2. Local network radio or phone call.
3. Remote radio call
4. Visit remote location physically as a business trip.

Yes intuitively, we will need to perhaps add new parts and equipment or even the necessary infrastructure in each settlement to make the communication method 2 and 3 above possible.

Then we'll need to create some Tasks to make use of those equipment.

[Urwumpe]

Yeah, also we could sure add a number of "team building problems" to Mars-sim, if we could find a real psychologist to assist us there. I know that such people exist for spaceflight missions, but I doubt I could learn enough from literature to abstract this into a simulation aspect. Having somebody as collaborator, who really studied this and is really competent there, might be the best option there.

I only really know motivation theories since it is one of my favorite topics about team and community management.

[bevans2000]

Well, one possible idea would be, to have a set of standard variables on a Mission or Project, which Java activity building blocks can access.

Certainly some standard variables will be needed.

Another option key-value pairs, which are available in the context (Mission/Project) but then we would need an exception handler activity to deal with missing data for each required value.

Too generic and not type safe.

More complex, but more flexible solution: We give things in the simulation two traits by defining an Java interface: Some things can be settler-accessible.

There will still be a need to classify the different Missions/Projects. E.g in the dialogue we need to know which involve a Vehicle

For the logic how to proceed to the next phase, what about having milestones, that trigger when a number of activities leading to it have finished successfully, like it is done in commercial workflow engines? So when we define a mission step or phase, We assign each activity to two milestones: A starting milestone and an ending milestone. This way, we could also define some coarse time estimate for each milestone, by estimating for each activity how much time it roughly needs - and react to milestones slipping more than a certain threshold, for example 5 sols.

Yes we could do. I was thinking more a simple linear sequence of steps as that is all we need currently. Once a step is completed it is removed from the sequence. To abort a mission, we abort the current step and replace the sequence with a travel step going home.

[bevans2000]

So I've had some thoughts on this during my morning run today. I think it's a lot of work but I can see significant benefits is centralising around the concept of a Project. Main one is a Person can be involved in multiple projects. But we need an "evolution not revolution" approach.

I have put together this target model of where we should get finish. This includes the proposed "Activity" style projects. (Using the cool PlantUML)

[Urwumpe]

I like that structure. Instead of a direct relation between Mind and Project, could there be a ProjectRole as relation object, which sets:

• Which role the the Settler has in the Project
• Which settler is assigned to this role.
• Maybe a short descriptive internal name for this role assignment for debugging (e.g. "Driver 1" and "Driver 2")?

[bevans2000]

Yes good point. I will add an Association class to the relationship. It will help identify the Leader and other Projects might have other roles

[bevans2000]

I have updated the target model and added a ProjectRole association now.

[mokun]

Now, I have another entirely different question as follows :

Q : How would you handle a person having two meetings assigned to him ?

In real life, I have a few meetings assigned to me from time to time on a given day to attend.

So it's not difficult to imagine that while a person was just assigned a mission, he may also have to attend a mandatory safety briefing, etc, while his supervisor may want to see him and have a meeting with him.

So what need to be changed in the code to be able to hand multiple missions/trainings ?

[bevans2000]

Thinking of what we discussed above, each Person would have the ability to be a member of multiple Projects. There would be a priority to influence which Project a Person worked on when free.

[Urwumpe]

Also a meeting priority might be helpful. Maybe just like in Outlook, being able to tell, if a person invited to a meeting is mandatory or optional.

[mokun]

Also a meeting priority might be helpful. Maybe just like in Outlook, being able to tell, if a person invited to a meeting is mandatory or optional.

Yes. I imagine we'll need a mission/meeting/activity priority score in TaskManager to determine which task to take.

Also, some tasks can be done in parallel.

e.g. If a person is very hungry/thirsty and he needs to go attend a safety meeting, he can grab his snack and water bottle and go to the conference and have the meeting while eating.

Having said that, I'm starting to think that at some points in future, if we ever switch to simulate things at a macroscopic scale, eating and drinking should be considered background tasks, such as breathing.

We can get rid of EatDrinkMeta and don't need to have PersonTaskManager calculate if a person should eat and drink.