# Why are bipedal robots difficult?

Not sure if this has been asked, but there are lots of simulations of bipedal locomotion algorithms online, some of the evolutionary algorithms converge to very good solutions. So it seems to me that the algorithm part of bipedal locomotion is well-understood.

If you can do well on simulations, you should be able to do it well in the real world. You can model delay and noise, you can model servo's response curve.

What I don't understand is then why is it still difficult to make a walking robot? Even a robot like the Big Dog is rare.

• Although we can simulate walking algorithms/gaits (and I don't agree that these are easy at all), actually building the mechanical bits is much more complex and expensive than just building a wheeled vehicle. Walkers are not as common as wheeled robots because it's difficult to do the control, provide a compact power source etc in practice.
– Andy
Sep 28 '15 at 15:45

If you can do well on simulations, you should be able to do it well in the real world.

There's this lovely saying:

In theory, there is no difference between theory and practice.

And what that implies for practice goes without saying.

You can model delay and noise, you can model servo's response curve.

A simulation is only as good as the model it is simulating. We like simple models. We don't like nonlinear things. We don't like to deal with too many variables...

That something works well in a simulation is often a necessary condition to be able to do it in the real world, but it's not necessarily (pun intended) a sufficient one.

Given all the necessary information doesn't mean it's possible to do it or reasonable to do so. Maybe we know how to do the calculations and have all the input, but cannot do the calculations quick enough with a reasonable sized processor.

It appears to me that your reasoning is that because we are able to do it, we should be doing it. There are plenty of other reasons to do (or not do) something aside from the mere possibility of doing it. Maybe it is in fact possible to build a robot with bipedal locomotion. It's quite possible that the difficulty in actually doing it is in finding a reason to do so.

Would that bipedal robot be faster than a robot with wheels? More stable? Would it require less energy? How well would it work on different terrain? Will the money required for further research and testing pay off later in any way?

A lack of benefits or at least lack of obvious benefits is a major difficulty to develop something. Running a computer simulation is one thing, but building a physical robot (and designing one before that) requires funding which requires making rich people interested in finding a solution to the problem.

some of the evolutionary algorithms converge to very good solutions

Speaking of evolution, bipedal motion is not exactly the most popular one. And it isn't the best one either, because no matter how successful you consider the bipedal homo sapiens, one thing he did was to invent other forms of locomotion superior to his bipedal one.

It looks like this is one of those research topics that we'd like to solve so we can claim that we are able to do it, but other technologies (drones fro example) seem to offer more and more obvious advantages.

• I think it's pretty obvious that bipedal walking is superior to wheels. It's adaptable to terrains, use less energy compared to many-leg designs, and is lightweight. Remember that somebody is trying to make a robot ostrich? That'd be very useful on battlefields (unfortunately, robot applications are still mostly limited to wars). Sep 28 '15 at 8:32