# 5. Dynamic weights

This 5th step illustrates how to obtain the shortest path from a point to another one and to update the weights of an existing graph.

## Formulationâ€‹

- At initialization, the value of the
`destruction_coeff`

of the`road`

agents will be equal to 1. - Add a new parameter: the
parameter that represents the value of destruction when a people agent takes a road. By default, it is equal to 0.02.`destroy`

- When a people agent arrive at its destination (home or work), it updates the
`destruction_coeff`

of the`road`

agents it took to reach its destination: "destruction_coeff = destruction_coeff - destroy". Then, the graph is updated.

## Model Definitionâ€‹

### global sectionâ€‹

We add the ** destroy** parameter.

In the global section, we define the `destroy`

variable:

`float destroy <- 0.02;`

In the experiment section, we add a parameter:

`parameter "Value of destruction when a people agent takes a road" var: destroy category: "Road" ;`

We define a new reflex that updates the graph at each simulation step. For that, we use the `with_weights`

operator. This operator allows to update the weights of an existing graph.

`global {`

...

reflex update_graph{

map<road,float> weights_map <- road as_map (each:: (each.destruction_coeff * each.shape.perimeter));

the_graph <- the_graph with_weights weights_map;

}

}

### people agentsâ€‹

At each time-step, after a `people`

agent has moved over one or multiple road segments, it updates the value of the destruction coefficient of `road`

agents crossed (i.e. roads belonging to the path followed). We have for that to set the argument ** return_path** to

`true`

in the `goto`

action to obtain the path followed, then to compute the list of agents concerned by this path with the operator **.**

`agent_from_geometry`

`species people skills: [moving]{`

...

reflex move when: the_target != nil {

path path_followed <- goto(target: the_target, on:the_graph, return_path: true);

list<geometry> segments <- path_followed.segments;

loop line over: segments {

float dist <- line.perimeter;

ask road(path_followed agent_from_geometry line) {

destruction_coeff <- destruction_coeff + (destroy * dist / shape.perimeter);

}

}

if the_target = location {

the_target <- nil ;

}

}

...

}

## Complete Modelâ€‹

`loading...`