Your First Visual Tracking Application Part Two

Filed under: FANUC Line Tracking Vision

In part one of this series, we covered gripper design and testing, vision considerations, robot selection, cell layout and choosing the correct number of robots for visual tracking applications. This post will cover some other considerations: conveyor flow, hardware requirements, setup and configuration, and finally, programming.

Here’s the full list with links to each section:

  1. Gripper
  2. Vision
  3. Robot Selection
  4. Layout
  5. Number of Robots
  6. Relative Conveyor Flow Direction
  7. Hardware
  8. Setup
  9. Programming

Relative Conveyor Flow Direction

If you consider a simple system with two conveyors: one for parts coming in, and one for parts going out, which direction(s) do they go? Do they both move in the same direction, or do they oppose eachother?

We call it “parallel-flow” when the conveyors both move in the same direction, and we call it “counterflow” when the conveyors move in opposite directions.

Some people may instinctually imagine a parallel-flow system while others picture a counterflow system.


  Parallel-flow:
  -------------------------------------
  Infeed            flow direction --->
  -------------------------------------
  -------------------------------------
  Outfeed           flow direction --->
  -------------------------------------

  Counterflow:
  -------------------------------------
  Infeed            flow direction --->
  -------------------------------------
  -------------------------------------
  Outfeed           <--- flow direction
  -------------------------------------
 

Each options has its own advantages and disadvantages, and those become more evident and important as the number of robots increases.

Parallel-flow

What happens when a part is missed on the infeed? It simply goes off the end of the conveyor. If there is something collecting excess at the end of the system, it can simply place to any empty slot on the outfeed without affecting the robot operation since it will be done downstream.

Imagine a long line of 10 robots. How does the most upstream robot operate? This robot has a very easy job: it can pick 100% of the incoming parts and place to 100% of the outgoing slots. What about the 10th robot? This one has it a little bit of harder. Assuming equal load distribution, this robot has only 10% of the infeed choices the first robot had, and it also has only 10% of the outfeed slots available. Depending on the decisions the other robots have made, this robot’s picks and drops may or may not be in sync with eachother. This causes misses on both the infeed and outfeed sides of the system.

Counterflow

If a missed part on the infeed is placed directly onto the outfeed, this immediately causes an imbalance in the system. Assuming the system has equal part coming in and no excess going out, you’ve just filled a slot that was meant for a robot to use.

What about the balancing problem? Picture the same line of 10 robots: how does the upstream robot operate now? It has 100% pick opportunity, but now it’s starved on the outfeed side, with only 10% to work with. The 10th robot has the opposite problem: 10% of the available parts to pick, but it can always place immediately.

So which one do I choose?

It usually comes down to space and money. Parallel-flow typically makes more sense from an automated line perspective (everything always moving in the same direction from start to finish), but counterflow does have some balancing advantages.

Load-balancing becomes extremely important as soon as you start working with more than one robot no matter which direction your conveyors flow. Counterflow systems tend to self-balance better, but you’ll probably end up having to control it yourself anyway if you can’t tolerate many missed parts on either side.

Go with parallel-flow if you can stop one or both conveyors without affecting the rest of the system. Ideally you’ll be able to have a slight buffer upstream on both sides that will allow the system to start and stop both sides as necessary to prevent any lost parts.

Hardware

At minimum, a dual-track visual tracking system will require:

  1. (2) Encoders
  2. (2) Encoder cables
  3. (1) Camera (2 if you want to use a camera on the outfeed too)
  4. (1) Camera cable (2 if you have 2 cameras, obviously)
  5. Depending on your type of robot, you’ll need one or two line tracking boards depending on the # of connections the board allows

If you are using multiple robots, you’ll need to decide how to get the encoder signals to the other robots. You can either use a multiplexer + additional cables or use the Ethernet Encoder option to send the signals over ethernet. You’ll need a good switch (pretty sure FANUC recommends a fancy and expensive managed switch, but you might be able to get away with a cheap one) and a couple ethernet cables.

Here’s how the pieces fit together:


      .-------.
.-----|encoder|
|     .----------------------------------------------------.
|     |////////////////////////////////////////////////////|
|     | .--------.                                         |
|   .---| camera |      Infeed  --->                       |
|   | | `--------`                                         |
|   | |////////////////////////////////////////////////////|
|   | `----------------------------------------------------`
|   |
|   | .-------.
| .-|-|encoder|
| | | .----------------------------------------------------.
| | | |////////////////////////////////////////////////////|
| | | |                                                    |
| | | |                 Outfeed --->                       |
| | | |                                                    |
| | | |////////////////////////////////////////////////////|
| | | `----------------------------------------------------`
| | |
| `--------------------------------------------.
|   |                                          |
|   |           .--------------.            .--------------.
`---------------|              |            |              |
    `-----------| controller 1 |---.    .---| controller 2 |
                |              |   |    |   |              |
                `--------------` .--------. `--------------`
                                 | switch |
                                 `--------`
 

With the ethernet encoder option, you can connect one encoder to each controller and then the signals will be passed over ethernet to the other controller(s). The vision offsets will also be passed over ethernet.

Without the ethernet encoder, you would have each encoder connect to a multiplexer which then outputs each encoder signal to each controller.

Encoders are typically connected to the conveyor with either a coupling or friction wheel. Both have advantages/disadvantages. Whatever you choose, make sure you end up with ~20-30+ encoder counts/mm of conveyor travel as your final encoder scale.

Setup

Setup is pretty easy:

  1. Connect all hardware
  2. Enable and test each encoder on all robots to make sure the counts are changing correctly
  3. Go to iRVision setup and create a Camera configuration. Make sure you see a picture when you take a snapshot
  4. On each conveyor, teach your encoder scale and tracking frame for all robots
  5. If using multiple robots, run a tracking frame validation to make sure things look right
  6. Calibrate vision

I could probably write an article for each item 3-6, but I’ll leave it at this for now.

Programming

When it comes to programming a visual tracking system, you have two choices:

  1. Start from scratch
  2. Use PickTool

So the question is, should I use PickTool? The answer depends on a couple of things:

  1. How complex is your system? (# of robots, # of different product variants, buffering requirements, conveyor switching, etc.)
  2. How much time do you have?

The real question is, “will using PickTool save me time and/or money?” Visual tracking is difficult, especially at high throughputs. Regardless of whether or not you use PickTool, you will have to learn how visual tracking works: how you teach your tracking frames, how you calibrate vision, how reference positions work, etc. Adding PickTool requires you to learn an additional set of conventions, programs, setup requirements, etc., and it gets you further from the “metal” of the application. Instead of writing a few hundred lines of code yourself, you have to learn to work within the PickTool framework’s way of doing things.

If your application is simple (i.e. just one or two robots, nothing too fancy), I would contend that writing your own TP programs would be a very worthwhile exercise. You’ll develop a deeper understand for how visual tracking works and eventually understand PickTool better if you decide to use it.

I haven’t spent much time with the new iRPickTool software, but I understand that FANUC has made the integration between iRPickTool and the underlying visual tracking architecture much tighter. As a result there should be less confusion over redundant setup screens, setup conflicts, etc.

That’s Visual Tracking in a Nutshell

I hope visual tracking is not so intimidating now. I know I only skimmed over a couple of topics (setup, programming), but I’ll expand on those topics in future articles.


Want posts like this delivered right to your inbox?

If you liked this post, please sign up for my mailing list!