I’m going to take a break from my regular topics and blog about something that’s been a long term interest of mine: Spatial and Visual Searching.
“Huh,” you say? What’s “Spatial and Visual Searching?” Well, there’s no real definition, it’s just a phrase I made up back in school when I was working on autonomous Robots. Basically, what I’m referring to is the problem of an autonomous Robot (or similar) spatially searching a space using a vision system as a primary sensor (though this isn’t exactly necessary, but yet given the state of sensing systems now and then, it’s a reasonable assumption that this will be a Robot’s primary searching sensor).
My particular interest was in a Robot searching indoor spaces, such as in a house or office building. I’m most keen on indoor Robotics because after all that’s where many of us spend much of our time and it seems natural that a Robot would too. Indoor environments also are more of a constrained environment than let’s say searching a mountainside or possibly even a lake bed. In fact, I imagine a whole additional set of challenges and discussions could center around searching unrestricted spaces, such as outdoor venues, multiple buildings, forests, you name it. Anyway, I was interested in how an autonomous Robot might try to find something–let’s say a bucket that it was asked to retrieve or the whereabouts of a person.
Before I get too deep, I’ll lay out a couple of boundaries to my thinking:
For now, this discussion will be geared around a physical Robot searching for something as it slides across a single-level floor, with a known floorplan. There’s no doubt that there are many other possibilities. For instance, back in the day, I used a gantry Robot when I initially tried tackling this problem. It was a conventient way of navigating across a space with minimal problems, but it sure isn’t a configuration that would be likely in a home. I’m also keen on virtual Robots, but that’s something for another post.
Vision will also play a lead role in how I envision a Robot searching for something indoors. That doesn’t mean that other sensor aren’t useful–it’s just that a vision system is a good, general purpose sensor that’s fairly practical to use.
With these boundaries in place, let’s start at the beginning: What do I mean by “spactial searching” for Robots? Well, it’s really no different at a conceptual level than what you have to go through to find something in your home. If someone asks you to find their glasses, how do you go about doing it?
You might start by thinking in terms of how easy it might be to see what you’re looking for. You might also consider where the object might be–with the primary objective here I presume to minimize the time and energy it takes to locate the object. From this list of possibilities, you might prioritize it and create a series of plans on how to search this space. These search plans might be hierarchical or exhaustive or who knows. It might depend on what you are searching for and how easy it is for you to use your sensing abilities to locate it. In other words, if it’s a small, translucent object you might have to be very close to it to see it or you might need to be careful when there’s not enough light or enough contrast or too much light or, or, or. Much depends on the sensors used and their capabilities, if you will. Don’t think object oriented, think sensor oriented.
Ooops….I’m being beckoned. I didn’t get very far….but I’ve got to run. This is chemo week so I’ll do what I can to get back ASAP….So much for the best laid plans…lol.