They kept repeating: it is not about research, this is about product creation!
One of the things that shocked me when I started to work for PAL Robotics some years ago was the required change in mentality. A change was necessary from a research based mind to a company based one. I almost remember myself thinking: you are in the army now!.
.thumbnail.JPG "Yes! It worked once! Let's write a paper!")
Yes! It worked once! Let's write a paper!Let's put it clear: a robot product must work correctly all the time. It is not enough to make it work for a promotional advertising, for a short demo, or to make a video for a conference paper on a very specific situation and condition. It must work under all the circumstances it is supposed to be when in its operating environment. Under all light conditions, under all noise conditions, under all crowds size. And let's face it, this is a tough job. Especially because we are talking about real hardware, no computer simulations.
Because of that, the mentality inherited from the time spent at University doing research is not valid any more. At least in its purest form...
Some time ago, I heard a well reputed SLAM researcher say that SLAM was a solved problem. So, he was proposing, why don't we move on to studying another thing?.
Well, I wouldn't go that far. One can say at most that in theory the problem is solved. It is like those problems on physics about using spherical cows.
I would say from my experience that robust SLAM is far from being solved. At least when using real robots with no adapted environment. Spherical cows do not exist in real life.
Experimenting with real robot REEM-BAnd that is what we are talking about here: real robots with real hardware, with real interferences between sensors, real battery problems, real CPU constraints, and more important than anything, real unpredictable environments. And under those conditions, current SLAM techniques do not work. What for a research mind seems a solved problem, for a product-generation mind seems almost useless.
So here we are. We need to solve a practical problem, which has a theoretical solution which doesn't work at practice. And here is where, paradoxically, our research mind comes to the rescue. A research mind is required to find a new solution to the problem, but having as a target the generation of a real life solution, rather than a paper publication one.
This means that one can apply the research methodology and add the necessary twist to adapt an existing research solution, modify it, or completely create a new one for that problem. Using this approach you can, for instance, construct a complete SLAM system which works on a humanoid robot.
So after all, the complete change of mind was not so necessary but instead, a change of focus.
The Zen master said, you must learn the path, follow the path and then forget the path. And that is exactly what happened to our mind when we moved from research to company work. However, in the process nothing remained the same. The path remains in our essence, hence, it can come to the rescue when needed, applying it to the place that is most required.
Just remember to enjoy the path.
