Teach a Cobot New Tricks
Originally published on May 22nd, 2018.
I blogged recently about Cobots and Tony Stark’s robotic arm and before publication that blog veered off into a different topic, but it’s one I’d like to dive into now. Throughout the Iron Man series, Tony repeatedly tells the articulated cobot that he’s doing it wrong, he even goes so far to place a dunce cap on its head and put it in the corner. And this brings me to an important aspect of a cobot: teachability.
There is a marrying of the control and logic of a robot with its mechanics. If you think about it, this is a human trait too. My son is learning to tie his shoes and he knows how to do it; he can walk me through the steps, but he can’t physically do it well yet. He is still processing his kinesthetic learning. His muscle memory for feel, touch, grip, and adjustment hasn’t developed yet. For industrial robots, we are still at this young child stage for robotics. That’s not to say that there aren’t intelligent people putting these together, but it’s not as simple as Hollywood makes it look (Anthony Daniels and Kenny Baker were still humans playing robots in Star Wars). These newer robots will have similar problems to our children. Companies will be programming the brains and the how to, place the decision logic into the controls, but the mechanics still need to be able to learn kinesthetic feedback like humans do.
For example, when you go to lift a flat heavy bag of sugar from the table, there are a few techniques people tend to use. It usually involves sliding the bag to get your hand around the smaller width of the bag and finish lifting. This is easy for an adult, but for a child with smaller hands, it becomes a challenge. Assuming the cobot is capable of picking up the bag of sugar, how hard does it squeeze? Will it use another arm or tool to slide its end effector under the bag? Will it slide the bag to the end of the table and balance it on the end and then lift? Will its gripper be larger than a human hand and will it decide to pick the sugar up width wise? Finally, how much pressure will it squeeze with? There needs to be enough pressure to lift the sugar but not so much that it busts through the paper and spills sugar everywhere.
I imagine that like human children (and adults learning a kinesthetic process) teaching the robot will be based on trial and error. The cobot will be lifting that bag of sugar many times. The bag will need to be placed on the table in different layouts- flat, wide, diagonal, upright, and so on. Once the cobot can pick up the bag of sugar, will it be presented a new item like an apricot to pick up? Will it be able to do so without damaging an apricot? Teaching the robot will need to be easy for the end user and the learning process for the robot will need to take place quickly as well. All of the work depends on the robotic controls design in conjunction with the robotic joints and mechanics to perform the physical work.
As a side note - I’m a trainer - do my students look different in a few years?
About the Author