-1

I'm currently doing a project using Webots to create a 3d simulation world - a container terminal yard whereby multiple robots(AGVs) reach their designated destination to load/unload containers.

Here's a glimpse of what I've been doing for the past few weeks.

http://www.youtube.com/watch?v=Rt6NlGP9wpA

The round bubbles you see act as a wireless range to send directions to the AGVs.

Seeing some similar threads on shortest path algorithm like Dijkstra or A* Algorithms, I'm pretty sure it can be done but i was hoping if anyone could perhaps provide some insights whether it's possible? and which algorithm is preferred to use?

Thanks and Regards

user2749692
  • 21
  • 5
  • 1
    The question is not clear - is each robot free to find the topographically shortest way to its designated target, does it have to pass a given chain of way points on the way, or is it free to travel anywhere as long as it's within range of some waypoint at any given time? Another question is - can you plan the perfect course in advance given perfect knowledge of the waypoint locations, or do you have to pick the course in real-time? – Leeor Sep 09 '13 at 20:08
  • Each robot will be given a chain of "directions" for every waypoints it passes. So far I thought of Using dijkstra algo(Directed weighted graph) 's nodes to represent the waypoints. So yes, the location of waypoints (xy coordinates) are pretty fixed and I think I could plan the perfect course in advance before dispatching the robot. Still, its pretty hard to guess if I could pull this off.. – user2749692 Sep 10 '13 at 19:12

2 Answers2

0

Well if there was a certain path you had that we could use as example, explaining how the algorithm would work would be simpler.

But i'll try to explain anyways:

Considering you have a Starting point A and a destination Z, which i will represent in this Pyramid:

. . . . 3 . . . .  
. . . 7 4 . . .  
. . 2 4 6 . .  
. 8 5 9 3 .

In this pyramid we will try and find the minimum path sum, starting from the first line and reaching the last line, by picking any of the two numbers benieth the one we picked previously.

On a small tree, it would be simpler to measure all possible solutions and pick the smallest. This is not effective on bigger trees.

The way to solve it would be to start from the bottom and go up, take 8 and 5, and add the smallest to the number above them (2 here) then 5 and 9, then 9 and 3. You repeat the process on the line above, until you reach the top line, and the Minimum path sum.

I hope this clears up a bit on the nodes and minimum path. Though this might be hard to code for AGV paths with several paths and nodes intercrossing.

A simple way i would use, if you have the coordinates of each AGV node (stopping point that connects different paths) would be to draw a line between the destination and the start, and add node by node the closest ones to that line that are connected to the previous node.

Good luck :)

Pshemo
  • 122,468
  • 25
  • 185
  • 269
ThaBomb
  • 702
  • 6
  • 11
  • Hi thanks for the advice! :) I made the AGV paths to have several paths and nodes intercrossing. i think the last point you mentioned is possible since each AGV have an GPS embedded. But with the dijkstra Algo(Directed weighted graph), maybe i can perhaps give a list of waypoints to the AGV and assign a chain of directions along way? – user2749692 Sep 10 '13 at 19:23
  • You could do that But do keep in mind that having several AGV's means they will need to each calculate routes on each nodes that have several outgoing routes Try computing routes from the start for each AGV, and adding it to a route list of nodes, that way AGV's need only to follow after compution. This may save you alot of memory and calculations when running. – ThaBomb Oct 10 '13 at 19:21
0

Given your clarifications, you need to choose your game here. The most straightforward solution is indeed running a Dijkstra algorithm and building the shortest paths from any waypoint to any other waypoint (this means a single algorithm run for each possible source point). Note however that this is only done once (at least until some waypoint is moved or removed). By the way A* isn't related here, it's intended to run on decision/game trees and find min/max optimal path (max benefit for you, min for the opponent), that's a different story. Another alternative is to run Floyd–Warshall which find all paths in a single run.

Anyway, once you ran that, you can keep for each waypoint a table, saying what is the next node in the shortest path for each destination waypoint. You don't need to entire path, just the next vertice. Once the robot gets there, it's told where to go next. This is basically the same thing done in most network routing algorithms.

Now, if you want to showcase how the robots are working, you could make them run this algorithm online and calculate the paths for themselves, but then the communication with the waypoints is rather pointless.

Either way, looks like you're having some fun coming your way, enjoy :-)

Leeor
  • 19,260
  • 5
  • 56
  • 87