Improved ant colony optimization for robotic navigation paper

Question

I am trying to understand this document and make a live implementation of the improved optimization of ant colonies for paper navigation for robots . While I was trying to implement, I had a few questions that hit my head:

• The author presented a negative deposition of pheromones (see second column of the above article). But I did not understand what it is or where it is used! Inside the paper, he does not talk about this, nor about this. What is it and where will we use it? We are already doing pheromone precipitation and evaporation.

• In the target search algorithm (on page 2), pheromones are deposited after all ants move to the next position, as well as after evaporation. So, at this time, the deposition of pheromones is carried out by iterating through all the ants and updating the concentration of the pheromone at their current location, is not it?

• In this goal search algorithm (on page 2), the author talks about Check if termination criteria met. So, does this mean to check if ant has reached its goal (i.e. Destination)? If so, execution should be discontinued. Is not it?

• In addition, I did not understand what he meant by these three lines in the target search algorithm on page 2:

  • Office ant distance from the wall

  • Backtracking prevention

  • 4-square loop prevention

I included a screenshot of the relevant part from the above article:

, .

, : https://softwareengineering.stackexchange.com/questions/238639/ant-colony-optimization-movement-of-ants

Answer 1

(. ). , ! googled . ? .

, , , . , , , , . , , , ...

, . , , .

( . 2) , . , , ?

( . 2) . , , ant (.. )? , . ?

, - - . . , 90% , .

(5).

, , . (3) (4) , (, , , ).

, , . 2:

    Control ant distance from wall
    Prevent backtracking
    Prevent 4 square looping

. , ( - , , ...). ant , - 4- (.. , ).

EDIT ( )

initialize all cells pheromone levels to some constant > 0

repeat
    set all ants to start location and erase their history
    repeat
        for every ant do
            if this ant is at the goal skip it
            make list of all neighbouring cells that are
                1. not too close to a wall
                2. not equal to the previous cell
                3. not equal to the cell that was visited 3 moves before
            calculate probability for all cells in this list
            choose next cell according to these probabilities
            update current position and history
        end for
    until 80% of all ants have reached the goal

    evaporate pheromones

    for every ant do
        if it reached the goal
            add pheromones to all cells in this ants history according to (3)
        else
            substract pheromones accoring to (4)
    end for
until length of shortest path has not changed for M iterations

, . 2. 3. , ant, ....