next up previous contents index
Next: Objective Up: The Legged Robot 3D Previous: Abstract

Introduction

In general (but not always [14]), legged robots are special purpose robots constructed in research laboratories for experimental reasons and that result difficult to built, very expensive, and not very robust.


 
Figure 1.1: The Argos  six-legged robot and its corresponding simulation.
\begin{figure}
\centerline{
\includegraphics [width=0.9\linewidth]{images/robots.eps}
}\end{figure}

Performing experiments with a legged robot is a tedious and dangerous process since experiments tend to last long periods and because any programming error can result in a damage of the robot. For these reason it seem reasonable to test tentative controllers in simulation previously to test them in the real robot. However, simulator can not be as accurate as reality and can not replace it at all [3]. In general the more the realistic the simulator the more useful it becomes. Trivial simulators for legged robots can only deal with flat terrain and the robot's legs are reduced to a two degrees of freedom (dof) mechanism that move legs up and down or forward and backward. These simulators are only useful to study the generation of patterns of steps. For instance, we use one of those simple simulators in [8] and in [10]. To test more complex controllers, more realistic simulators are needed. The drawback is that the more the realistic the simulator the more the complex and slow it becomes. Consequently, we have to meet a trade off between fidelity to the reality and usefulness. In the simulator presented in this report, we include the simulation of abrupt terrain (it is the only terrain in which it make sense the use of legged robots) but in the most simple way: Leg-ground interaction are detected with a simple touch sensors (no force sensors are provided), ground friction is considered to be infinite (so all leg to ground contacts are valid footholds), and the robot body can only be propelled if the supporting legs are explicitly moved in a coordinated way. Additionally we only consider the kinematic aspects of the robot (leg positioning) and we do not simulate dynamic aspects (inertias in leg movements, gravity,...). In general this is not a drawback since in the main part of the legged robots only use statically stable gaits. For those aiming to produce dynamic gaits more complex simulators must be use [12].



 
next up previous contents index
Next: Objective Up: The Legged Robot 3D Previous: Abstract
Josep M. Porta Pleite
8/2/2000