The A.I. revolt has begun! Robot kills Man!

Page 4 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

Engineer

Elite Member
Oct 9, 1999
39,230
701
126
I've almost been hit myself once or twice myself in the last 30 years, once they are set to do something they will just do it and do not care if a person is in the way or not in some circumstances,

On many robots, you can set collision detect (on the fly) for the moves to help eliminate the possibility of damage, even to someone getting hit. However, when you have a robot capable of lifting 350 pounds and it's rapidly swinging a 150 pound brake drum around, the collision detect has to be set so high that it wouldn't matter if it hit you or not. The results wouldn't be nice.

Not sure how Baxter does it's job so well and avoids hurting people when it hits them. Maybe sensors on all contact points. But Baxter is so damn slow, not sure that it would matter.
 

pcgeek11

Lifer
Jun 12, 2005
22,347
4,973
136
I'll say that you may do that in a production world but you will not do that in a development / machine building world, and I still say that your situation is unique. If it's possible to do it that way, sure....the more safety the better. In the real world, most of the time somebody is going to have to grab a pendent and go inside lots of cells.

So you are into the machine and automation cell building business and have never had anyone go inside a cell with a teach pendent and teach anything at all? Was all taught outside of the cell area, especially high precision pickup and place points? Have you done anything like that for automotive parts? From experience, it's damn near impossible considering the shapes of the machines and the parts produced. I've done it on a few machines but they are few and far between (using reference points for the other points).

https://www.youtube.com/watch?v=f9vnwhBvbIc

I do not work in the " development / machine building world ". Electronics tech in automated assembly, manufacturing automotive parts. Gasoline Fuel Injectors, Fuel Pumps etc... Very precise placement and taught from outside of the cell, using several fixed teach points and the remainder of all placement are computed. Granted the fixtures for the parts trays and the Assemblies have to be very precise also in order for this to work.

I guess in the initial development and initial system design you would in fact have to go into the enclosure with a pendant. I wasn't saying it never happens. I was only being short sighted looking at my end in a production environment repair scenario. My apologies.
 
Last edited:

Engineer

Elite Member
Oct 9, 1999
39,230
701
126
I do not work in the " development / machine building world ". Electronics tech in automated assembly, manufacturing automotive parts. Gasoline Fuel Injectors, Fuel Pumps etc... Very precise placement and taught from outside of the cell, using several fixed teach points and the remainder of all placement are computed. Granted the fixtures for the trays and the parts have to be very precise also in order for this to work.

I guess in the initial development and initial system design you would in fact have to go into the enclosure with a pendant. I wasn't saying it never happens. I was only being short sighted looking at my end in a production environment repair scenario. My apologies.

Nah, don't worry about it. Not a big deal (even if I did my usual knee jerk reaction). No need to apologize.

I was just trying to wrap my head around how to build a machine / cell and have it safe from the first time the robot was fired up. Notice in my latest video, there was no cell. Many times, the machinery is built in the shop, transferred to the customer and guarding put in place then. The gate circuity, e-stop switches, etc. are usually already wired but many times, just jumpered out so that it can be run without guarding while developing. Oh, and with robot simulation software and using SolidWorks models, quite a bit of initial testing and development can be done inside of a virtual model and then transferred to the robot for testing / debugging.

Oh, and sorry for the knee jerk reaction.

I can see how what you're talking about would be done. Seems many of the machines would be more on a linear plane instead of twisting and turning inside itself.

Good stuff....
 
Last edited: