Future Extensions

The presented version of our legged robot simulator (including this documentation) is just a $\beta$version that (I am sure) includes many errors, incoherences, and limitations. However, this prototype has resulted useful for us to test our controllers and we hope it can be also useful for other researchers. Our objective when starting this project was that of producing a minimal tool to test legged robot simulator and we think this objective has been achieved. In future versions (if they exists at all) many improvements can be done:

• We can allow more than one robot to be simulated simultaneously.
• Different leg structures could be used. Each leg structure requires of its own direct and inverse kinematic routines. We can offer a limited set of leg structures so that the user can choose between them or we can try to program a set of general kinematic routines (but this is a too complex problem in general).
• The robot to ground interaction should be performed in a more realistic way. A good idea would be that of adding a collision detection engine. In this way not only any leg to ground contact would be detected but also leg to leg or body to ground contacts would be also detected.
• We should provided more powerful (on-line) tools to define ground shapes.
• Better triangulization algorithms could be used to set up the triangular mesh that represents the ground. This can reduce the number of polygons to be managed and would increase the visualization speed.
• It would be a good idea to program the visualization routines directly in OpenGL and not to use Geomview as a geometry manager any more.
• The world object could be extended to include other objects (stones, trees,...) to make more realistic sceneries. Additionally, these objects could be animated (i.e., provided with its own dynamic).
• We can include our autonomous robot programming language [9] into the simulator package so that user controllers can be written using it and not directly in C. This would be an advantage since our language provides structures suited to program robot controllers (parallel behaviors, wires to communicate behaviors, clock triggers,...).
• Noise could be added to the robot's odometry so that the values it returns are more realistic.
• The whole code should be re-written in C++.
• The exception handler should be improved so that when an error situation is found the user is better informed about the problem.
• The syntax used in the configuration files should be more flexible.

Uff!! such amount of work to do and this in only a small fraction of it!! But for now, that's all folks.