private void SetUpTree()
{
alice = new Thing("Alice");
bob = new Thing("Bob");
charlie = new Thing("Charlie");
dave = new Thing("Dave");
edna = new Thing("Edna");
frank = new Thing("Frank");
george = new Thing("George");
harry = new Thing("Harry");
alice.AddChild(bob);
alice.AddChild(charlie);
alice.AddChild(dave);
dave.AddChild(edna);
dave.AddChild(frank);
edna.AddChild(george);
}
//fill this method in with code which will execute
//once for each cycle of the engine
public override void Fire()
{
Init(); //be sure to leave this here
ModuleView naSource = theNeuronArray.FindModuleByLabel("UKS");
if (naSource == null)
{
return;
}
uks = (ModuleUKS)naSource.TheModule;
Thing mentalModel = uks.GetOrAddThing("MentalModel", "Visual");
Thing currentlyVisibleParent = uks.GetOrAddThing("CurrentlyVisible", "Visual");
Thing motionParent = uks.GetOrAddThing("Motion", "Visual");
//Things which disappeared in the last cycle need special treatment to remove from mental model & relationships but not the known object
foreach (Thing t in motionParent.Children)
{
if (t.Label.Contains("Disappeared"))
{
Thing thingToRemove = t.Children[0];
for (int i = thingToRemove.Children.Count - 1; i >= 0; i--)
{
Thing child = thingToRemove.Children[i];
thingToRemove.RemoveChild(child);
}
uks.DeleteThing(thingToRemove);
}
}
uks.DeleteAllChilden(motionParent);
//build little lists of things which moved or appeared so we can also be left with things which things which disappeared
List <Thing> matchedThings = new List <Thing>();
List <Thing> movedThings = new List <Thing>();
List <Thing> movedThingsNew = new List <Thing>();
List <Thing> newThings = new List <Thing>();
foreach (Thing visibleArea in currentlyVisibleParent.Children)
{
Thing matchedThing = null;
foreach (Thing storedArea in mentalModel.Children)
{
Debug.Assert(visibleArea.Children.Count == 1);
if (storedArea.Children.Count != 1)
{
continue;
}
if (storedArea.Children[0] == visibleArea.Children[0]) //are these instances of the same object type?
{
var values = uks.GetValues(visibleArea);
var prevValues = uks.GetValues(storedArea);
Point c1 = new Point(uks.GetValue(visibleArea, "CtrX+"), uks.GetValue(visibleArea, "CtrY+"));
Point c2 = new Point(uks.GetValue(storedArea, "CtrX+"), uks.GetValue(storedArea, "CtrY+"));
float dist = (float)(c1 - c2).Length;
if (dist < (uks.GetValue(visibleArea, "Siz+") + uks.GetValue(storedArea, "Siz+")) / 2)
{
//these are probably the same object...is motion detected?
var changes = uks.GetChangedValues(visibleArea, storedArea);
if (changes.ContainsKey("Siz+") ||
changes.ContainsKey("CtrX+") ||
changes.ContainsKey("CtrY+") ||
changes.ContainsKey("Ang+"))
{
Thing t1 = uks.AddThing("Moved", motionParent);
t1.AddChild(storedArea);
foreach (var change in changes)
{
uks.SetValue(t1, change.Value, change.Key, 0);
}
movedThings.Add(storedArea);
movedThingsNew.Add(visibleArea);
}
matchedThing = storedArea;
goto objectMatch;
}
}
}
//there is no matching object, this just appeared!
newThings.Add(visibleArea);
Thing m = uks.AddThing("Appeared*", motionParent);
m.AddChild(visibleArea);
continue;
objectMatch :;
if (matchedThing != null)
{
matchedThings.Add(matchedThing);
}
}
for (int i = mentalModel.Children.Count - 1; i >= 0; i--)
{
Thing t = mentalModel.Children[i];
if (!matchedThings.Contains(t))
{
Thing m = uks.AddThing("Disappeared*", motionParent);
m.AddChild(t);
mentalModel.RemoveChild(t);
Thing attn = uks.Labeled("ATTN");
if (attn != null)
{
attn.RemoveReference(t);
}
}
}
if (motionParent.Children.Count > 1)
{
//analyze motion
//if only a few objects moved, they moved
//if all objects moved in same direction, it's becaue POV turned
//if all objects moved in directions radiating to or from a point, POV is moving toward or away from that point.
//TODO if all objects moved in same direction but varying magnitudes, POV has moved sideways...further objects move less
//calculate the motion vector for all associated objects
List <Segment> motions = new List <Segment>();
foreach (Thing t in motionParent.Children)
{
if (t.Label.StartsWith("Moved")) //ignore appeared/disappeared
{
var values = uks.GetValues(t);
PointPlus p = new PointPlus(0, (float)0);
if (values.ContainsKey("CtrX++"))
{
p.X = values["CtrX++"];
}
if (values.ContainsKey("CtrY++"))
{
p.Y = values["CtrY++"];
}
Segment s = new Segment();
s.P1 = p; //This is a relative motion value
values = uks.GetValues(t.Children[0]);
p = new PointPlus(0, (float)0);
if (values.ContainsKey("CtrX+"))
{
p.X = values["CtrX+"];
}
if (values.ContainsKey("CtrY+"))
{
p.Y = values["CtrY+"];
}
s.P2 = p; //This is an absolute position
motions.Add(s);
}
}
if (motions.Count > 1)
{
//do most move the same? ANGULAR POV MOTION
//TODO: use clustering to get most likely motions
PointPlus pMotion = new PointPlus(motions[0].P1);
int count = 0;
foreach (Segment s in motions)
{
if (s.P1.Near(pMotion, .1f))
{
count++;
}
}
if (count > motions.Count - 2) //only two objects can be exceptions
{
for (int i = 0; i < motions.Count; i++)
{
if (motions[i].P1.Near(pMotion, .1f))
{
motions.RemoveAt(i);
Thing t = motionParent.Children[i];
uks.DeleteAllChilden(t);
uks.DeleteThing(t);
i--;
}
else
{
//an object has its own motion too...subtract off the POV motion
Thing t = motionParent.Children[i];
float x = uks.GetValue(t, "CtrX+");
float y = uks.GetValue(t, "CtrY+");
x -= motions[i].P1.X;
y -= motions[i].P1.Y;
uks.SetValue(t, x, "CtrX", 3);
uks.SetValue(t, y, "CtrY", 3);
}
}
Thing POVMotion = uks.AddThing("POVMotion*", "Motion");
uks.SetValue(POVMotion, pMotion.X, "CtrX", 3);
uks.SetValue(POVMotion, pMotion.Y, "CtrY", 3);
}
else
{ //this is not a rotation...is it a motion toward/away?
//TODO there are no exceptions allowed
foreach (var m in motions)
{
m.P1 += m.P2;
}
List <PointPlus> intersections = new List <PointPlus>();
for (int i = 0; i < motions.Count; i++)
{
for (int j = i + 1; j < motions.Count; j++)
{
Utils.FindIntersection(motions[i].P1, motions[i].P2, motions[j].P1, motions[j].P2, out Point intersection);
intersections.Add(intersection);
}
}
//testing clustering...
//TODO try all this with points of interest rather than centers
double[][] rawData = new double[intersections.Count][];
for (int i = 0; i < intersections.Count; i++)
{
rawData[i] = new double[2] {
intersections[i].X, intersections[i].Y
}
}
;
int[] clusters = KMeansClustering.Cluster(rawData, 2);
//now ask if there is a single cluster with most of the data
PointPlus aveIntersection = new PointPlus();
for (int i = 0; i < intersections.Count; i++)
{
aveIntersection.X += intersections[i].X;
aveIntersection.Y += intersections[i].Y;
}
aveIntersection.X /= intersections.Count;
aveIntersection.Y /= intersections.Count;
float dist = 0;
float dir = 0;
for (int i = 0; i < intersections.Count; i++)
{
dist += (intersections[i] - aveIntersection).R;
//moving toward or away?
dir += (motions[i].P1 - intersections[i]).R - (motions[i].P2 - intersections[i]).R;
}
dist /= intersections.Count;
dir /= intersections.Count;
if (dist < 10) // ???
{
uks.DeleteAllChilden(motionParent);
Thing POVMotion = uks.AddThing("POVMotion*", "Motion");
uks.SetValue(POVMotion, aveIntersection.X, "CtrX", 3);
uks.SetValue(POVMotion, aveIntersection.Y, "CtrY", 3);
uks.SetValue(POVMotion, dir, "CtrZ", 3);
}
}
}
}
//transfer currently visible objects to previously visible
Debug.Assert(movedThings.Count == movedThingsNew.Count);
for (int i = 0; i < movedThings.Count; i++)
{
uks.SetValue(movedThings[i], uks.GetValue(movedThingsNew[i], "CtrX+"), "CtrX", 0);
uks.SetValue(movedThings[i], uks.GetValue(movedThingsNew[i], "CtrY+"), "CtrY", 0);
uks.SetValue(movedThings[i], uks.GetValue(movedThingsNew[i], "Siz+"), "Siz", 0);
uks.SetValue(movedThings[i], uks.GetValue(movedThingsNew[i], "Ang+"), "Ang", 0);
}
foreach (var t in newThings)
{
t.AddParent(mentalModel); //TODO, this may leave multiple things in the mental model with the same label
}
//This hack puts predictable labels on things in the mental model so we can select them more easily in the description
//Area0-n from top->bottom, left->right
List <sortable> vals = new List <sortable>();
foreach (Thing storedArea in mentalModel.Children)
{
var values = uks.GetValues(storedArea);
vals.Add(new sortable {
t = storedArea, x = values["CtrX+"], y = values["CtrY+"],
});
}
vals = vals.OrderBy(w => w.y * 1000 + w.x).ToList();
for (int i = 0; i < vals.Count; i++)
{
vals[i].t.Label = "Area" + i;
}
//for debugging:
//if there is a new image, cancel the existing relationships
ModuleImageFile mif = (ModuleImageFile)FindModule("ImageFile");
if (mif == null)
{
return;
}
string curPath = mif.GetFilePath();
if (curPath != prevPath)
{
foreach (Thing t in mentalModel.Children)
{
var refs = t.GetRelationshipsByType(null, null);
foreach (Link ref1 in refs)
{
t.RemoveReference(ref1.T);
}
}
prevPath = curPath;
}
//if you want the dlg to update, use the following code whenever any parameter changes
// UpdateDialog();
}
