Noah's Reply to AlsOpeningSimEmail

Al and Uluc are excellent software engineers; Al seems to hit on many important points, and in terms of the RHex project, nobody is better suited to be thinking about these things than Uluc and Al. Uluc wrote the original simulation environment for Rhex (SimSect), as well as its successor, (SimSect)++. I think SimSect++ is the one that is controllable from RHexLib the software environment for RHex, but I'm not sure.

It sounds like they are in the frame of mind to "grow their own" robot simulator. They do outline some important issues if they use commercial products, but I would bet they want to build their own system; these days I don't know if that is necessary, but then again it has several clear advantages, for example, I really like their idea of a "creature description language" and I'm sure that is novel in terms of a robot simulation package. Could also be very useful for simulating organisms.

-- MarkCutkosky - 10 Jun 2003
Noah et al., One reason we went to Adams for vSprawl was that it made the details of modeling the leg compliance, damping, piston dynamics, etc. easier that trying to do it from scratch. As you point out below, modeling foot/surface interactions will be crucial. I can see how we might build appropriate models in Adams for this and would be interested to know how well SimSect and other packages would handle it. The weak point of Adams has been its ability to interact with a controller -- it can be done in theory (e.g. with Simulink) but we've found it difficult and fragile.

The thing that seems to be missing from all commercial packages, as far as I know, is a really good model for how surface interactions 'work'. Obviously, nothing could be more critical for scansorial robots. If you get that wrong, then you can forget about making any useful climbing predictions whatsoever. However, I don't know whether or not this group can get that right, either. It is just really hard. First of all, you also need a surface model, in addition to a robot model, and Al hasn't even mentioned that. Trees, rocks, teflon and glass are all different (stiffness, roughness, polarizability, etc). Is there a good surface description language? It would have to have a statistical component, too. This seems extremely challenging to me.

-- NoahCowan? - 10 Jun 2003

This site is powered by the TWiki collaboration platformCopyright &© by the contributing authors. All material on this collaboration platform is the property of the contributing authors.
Ideas, requests, problems regarding TWiki? Send feedback