Can Roborealm do this?
Bill from United States  [2 posts]
5 years
There are five robotic arms in the cooler at a local convenience store. Each arm is holding a different brand of soda and is responsible for replenishing two rows of soda as they are removed by customers.

All 5 robotic arms are watching as the cooler door opens. The customer took a soda from the second row from Robot #3's shelf. #3 Robot saw the absent soda bottle from its section so when the cooler door closed (telling the robot to act) it replaced the soda in the second row because it saw that the first row was still full. The other robots did nothing.

Will this software allow for this scenario?
Steven from United States  [1425 posts] 5 years

This is certainly possible but I'm not sure why one would use vision for such an application. Using a couple of strategically placed IR or contact sensors would probably allow you to do the same thing. In my mind, this is very similar to vending machine and thus subject to automation techniques rather than using robotic arms (i.e. one would build an entire machine around this problem rather than use an expensive robotic arm).

Is there something about the product that could not be detected using anything but a camera? Or am I missing something?

Bill from United States  [2 posts] 5 years
I used the soda replacement scenario to mask the real machine to keep secret my idea for creating a new machine.

The missing item has to be detected by something so that something can tell the robotic arm to replace it moments after it goes missing. I know of no other way to detect the missing object other than a camera noticing it gone. It cannot  be touched nor can any sensors, i.e. lasers, switches, the items own weight, etc. be used.

An ideal situation for my application would be for one single camera to view the entire area, determine when and how many sodas went missing at a given instance, then relay that information to one or more of the robotic arms telling them which slot(s) to refill.

For example: There are five robots covering five cooler doors. Each robot is responsible for two rows of sodas in each door. Three customers open three of the five doors simultaneously and each customer removes one soda from each door. When the last cooler door closes the camera takes a picture of the area in which the five robots are covering. It "sees" that three sodas are missing from the rows and directs three of the five robots to replace those sodas from the missing rows. The appropriate robots respond to the command and replace the soda in the appropriate missing row. Can one camera and Roborealm software handle that scenario?

And let's say that to handle the entire cooler system in the building, 10 cameras and 50 robotic arms are needed for this task. Can Roborealm handle all of that at the same time using only one computer?

Steven from United States  [1425 posts] 5 years

Understood of the example scenario. Just making sure that's the case as we often get asked questions where other sensors are more ideal.

In your case, yes this is possible .. until we get a chance to see what type of image you are referring to. For example, if the object can be segmented from its background i.e. its a different color/intensity/pattern from its background then that's possible. We'd really have to see an example image to confirm this though.

In terms of handling that many cameras and arms, that's tough to know at this point. If the image size is only 320x240 then I don't see an issue. Assuming the robotics arms are controlled via a simple lightweight ethernet protocol (like Modbus) then that is also not a problem. Up the image size to 1024x768 and the communications bus might get taxed if you are running at 30fps ... but that might also not be needed since you probably only need to check 1 times a second. So, unfortunately, I can only say that it might be possible given the information you've provided so far.

What I would recommend is setting up a single camera with a single robot arm and work out the issues with that prototype. Then we'll know what the issues are in expanding to 10 cameras and 50 arms. There are just too many variables otherwise to predict usage patterns.

Also note, RR is software and will run as fast as the hardware allows it. There isn't anything within RR that would NOT support this combination (i.e. we only support X number of cameras) so its really the hardware that limits this solution.

Please feel free to use the site contact form to send us your email take this conversation private. We can sign an NDA to get actual images to discuss further possibilities.


This forum thread has been closed due to inactivity (more than 4 months) or number of replies (more than 50 messages). Please start a New Post and enter a new forum thread with the appropriate title.

 New Post   Forum Index