Fritz Heckel

Since I want the ActionModel to take care of action resolution, and there to be only one per agent, it can’t be used directly by the agents. Really, though, this is a good thing. The ActionModel also needs to store some basic state information for behaviors—if following a path, it needs to keep the path around. It should at least track the effectors so it can figure out when they actuallly run.

So I’ve added an ActionModelInterface class. It has storage for some basic state, hold onto the SubsumptionPolicy of the layer, and has simplified move/speak/act request functions. This should actually simplify behaviors, and to some extent, the ActionModel, quite a bit.

This functionality could just be folded into the Behavior base class, but I think this approach is better. If nothing else, it makes for easier testing, which is important right now.

Scientific:

Accomplishments:

• Decided to use probabilistic roadmaps for within-region navigation. Probably what we’ll do is generate a roadmap when a region is first entered, and store that with SSPS. If the dynamic objects in a region change, we can either regenerate the roadmap or fix the original one. Don’t need to worry about that just yet, so I’ll make the basic version work.
• Have the basic Perception/Effector interface nailed down

Challenges:

• Need to think about the ActionModel more. It’s going to be substantially more complex in 1.0 than it was before, as it needs to track the current actions, transitions between actions, and handle resolving actions. Possibly the last should be handed off to the SubsumptionPolicy instead, but if so, SP will still need to use AM to decide how to combine actions.
• Still working on exactly how to deal with hierarchical subsumption priorities. This will be an issue with effector resolution, since we now have priorities within the lower subsumptions. My initial thought was to use a float representation, which would be nice, as we still have comparators that would work. I think now instead I’ll use a max size (10 element) boost tuple of, say, unsigned shorts. Turns out that boost::tuple has comparators that do exactly what i want. So this is really just an engineering problem, in the end.

Engineering:

Accomplishments:

• Most of the skeleton for 1.0 is in place, but everything needs some fleshing out.
• Started testing the perceptual model, but got hung up on SSPS-related stuff
• Created a boost::graph object which maps to our SSPS representation
• Got the boost A* algorithm working on that graph object. This was a big ol’ PITA, but now that I’ve done it once, using any of the other algorithms should be much easier. I have learned a lot about boost in the process. For now, I’ll just call the A* algorithm when I need distances. This will be hell of inefficient, so I’ll probably at least cache these distances, but this will obviously change.
• Created new schema/behavioral language for agents.

Problems:

• This list is long, so I’ll skip most of it, since it’s mostly in mantis.
• UDP comms for 1.0: Casey is working on this. He had some trouble with boost and SVN, but I think he’s doing alright with them now. I’ll check up on him again Wednesday.
• PerceptualModel testing is incomplete
• Action Model is high priority
• Floyd-Warshall is next targeted boost graph algorithm to get working
• Probabilistic road map implementation for SSPS

Papers

• Working on the AIIDE draft, starting from some FLAIRS content