Dealing with an housing estate

[kmamor]

I'm currently trying to teach myself GAMA. As such I have built a small project where a people agent moves between home and work. This is based on a road/building network taken from OSM via QGIS. Sometimes when I run it instead of moving between the two buildings the people agent jumps from one to the other. The main cause seems to be where a road leads onto an estate of buildings rather than a building itself . This is illustrated in the attached image. As can be seen a yellow road leads into a cluster of buildings. The agent then leaves the road (and presumably the graph) goes to the target building. When told to move to the next building it simply jumps directly to the next building rather then use the network.

Is there a simple fix for this problem?

I have included the code in case it helps. By the way - do I need the clean_network procedure (in global / init) ? I came across it in a model I was looking at but haven't seen it used in any tutorials.

model MK

global {
	file shape_file_roads <- shape_file("../includes/MK_roads.shp");
	file shape_file_buildings <- shape_file("../includes/MK_buildings.shp");
	geometry shape <- envelope(shape_file_roads);
	graph the_graph;
	int nb_people <- 1;
	int nb_people_in_model  ;
	int nb_roads_in_model;
	int nb_buildings_in_model;
	float min_speed <- 40.0 #km / #h;
    float max_speed <- 60.0 #km / #h; 
	int totalcounter <- 0;
	
	init {
		
		create roads from: shape_file_roads with:[type::string(get("highway"))]  ;
		
		//the_graph <- as_edge_graph(roads);
			
		list<geometry> var0 <- clean_network(shape_file_roads.contents, 100, true, true);
		create roads from: var0;
		
		
		the_graph <- as_edge_graph(roads);
		
		create building from: shape_file_buildings with: (type:get("building")){
		color <- #green;
		}
	
		create people number:nb_people {
		home <- one_of(building);
		location <- home;
		
		ask home {
			color <- #red;
		}
		work <- one_of(building);
		ask work {
			color <- #red;
		}
		speed <- rnd(min_speed, max_speed);
		objective <- "go_work";		
		the_target <- work;
		}

list<people> nb_people_list <- people; 		
nb_people_in_model <- length (nb_people_list);
list<roads>nb_roads_list <- roads;
nb_roads_in_model <- length (nb_roads_list);
list<building>nb_building_list <- building;
nb_buildings_in_model <- length (nb_building_list);


ask roads{
	if (type = "motorway"){color <- #black;}
	if (type = "trunk"){color <- #blue;}
	if (type = "primary" or type = "secondary"){color <- #green;}
	if (type = "tertiary"){color <- #yellow;}
}
}
	}

species roads  {
	string type;
	/* 
	rgb color <- type = "motorway" ? #black : #red;
	*/
	rgb color <- #black;
	aspect base {
		draw shape color: color ;
	}
}

species building {
rgb color <- #blue;
string type;
aspect base {
draw shape color: color;
}
}



species people skills: [moving] {
	building home <- nil;
	building work <- nil;
	string objective <- nil;
	point the_target <- nil;
	rgb color <- #purple ;
	int counter;
	int work_time <- 5 + rnd(5);
	int home_time <- 10 + rnd(5);
	
	aspect base {
		draw circle(100) color: color border: #violet;
	}
	
	
	reflex moving_home_work  {
	do goto target:the_target on:the_graph; 
		
	if location = the_target  {
	
	if objective = "go_work" {
		objective <- "at_work" ;
	the_target <- work;
	
		}
		
	if objective = "go_home" {
		 objective <- "at_home" ;
	the_target <- home;
			}
		
	if objective = "at_work"{
	counter <- counter + 1;
    	if counter >= work_time {
    		the_target <- home;
    		objective <- "go_home";
    		counter <- 0;	
		}	
		}
	
		
	if objective = "at_home"{
	counter <- counter + 1;
    	if counter >= home_time {
    		the_target <- work;
    		objective <- "go_work";
    		counter <- 0;
    	    	}	    	
    	     }	
		}
		}	
		}
	
experiment MK type: gui {
	
	output {
		monitor "Number of people" value: nb_people_in_model;
		monitor "Number of roads" value: nb_roads_in_model;
		monitor "Number of buildings" value: nb_buildings_in_model;
		monitor "Counter" value:counter of people(0);
		monitor "Target" value: the_target of people(0);
		monitor "Objective" value: objective of people(0);
		
		display city_display {
			
			species roads aspect: base ;
			species building aspect: base;
			species people aspect: base ;
			}
}

}

Thanks,

K

[kmamor]

I've added a line to each move instructions:

location <- roads closest_to(self);

In the moving_home_work reflex:

reflex moving_home_work  {
	do goto target:the_target on:the_graph; 
		
	if location = the_target  {
	
	if objective = "go_work" {
		objective <- "at_work" ;
	the_target <- work;
	location <- roads closest_to(self);
	
		}
		
	if objective = "go_home" {
		 objective <- "at_home" ;
	the_target <- home;
	location <- roads closest_to(self);
			}
		
	if objective = "at_work"{
	counter <- counter + 1;
    	if counter >= work_time {
    		the_target <- home;
    		objective <- "go_home";
    		counter <- 0;	
    		location <- roads closest_to(self);
		}	
		}
	
		
	if objective = "at_home"{
	counter <- counter + 1;
    	if counter >= home_time {
    		the_target <- work;
    		objective <- "go_work";
    		counter <- 0;
    		location <- roads closest_to(self);
    	    	}	    	
    	     }	
		}
		}

As far as I can tell this has done the trick. A few queries though.

a) I am still curious about the clean_network operator - do I need that?
b) Despite using the clean_network operator I did not produce a complete graph in the sense that some buildings do not appear to be adjacent to a road - hence the need to force the agent to move to the nearest road from the building. How can I guarantee that each building is identified with a road? Can GAMA do that or is there something I should have done using QGIS when I created the shapefile?

Thanks for any help,

K

[lesquoyb]

@ptaillandier