How maneuvers are computed while running the decision trees?

  1. How a basic maneuver is generated?

 

Generating a maneuver requires data from the input ‘event’. A basic maneuver is generated while running the corresponding decision tree. A basic Maneuver should be viewed as an input/output of the running algorithm of the decision tree. To each node, rules are attached. To fire a rule, the conditions should be satisfied. ‘event’ data are scanned. In the same time, a basic maneuver is also processed. Each time a rule is fired, the basic maneuver is updated. So from the root of the tree to the leaf, all the necessary data of ‘events’ have been taken into account to generate the most accurate basic maneuver. Here an example that illustrates this process:

 

 

It is a branch taken from the pursuit tree. It shows how the datum Maneuver is updated while the branch is being visited:

 

  1. How a maneuver is generated?

 

Now it has been explained how a one-decision trees generates a basic maneuver. Nevertheless there are three decision trees and a maneuver is based at least based on three basic maneuvers. How does it work?

 

The following policy has been chosen: a default basic maneuver is passed in the input of the first tree. An output basic maneuver is provided and re-injected in the running process of the second tree and so on. So that finally, as output of the third tree a basic maneuver is obtained that is supposed to take into account of the both previous ones. Actually, in the prototype, the basic maneuver updates made by a  tree replace those made by the previous one. This choice will be discussed further later. To sum up here is a diagram that illustrates this process: