Dynamic simulation


old new

The model of the virtual robot is shown above. The size of this model represents that of a possible real robot, not of insect. This is because we want to identify possible dynamic problems before constructing the real robot. However, it is possible to scale the entire model to that of the insect if it is necessary to compare the controller with the original insect scale. Insects move leg segments with antagonist muscles pulling in both directions. This creates a resultant torque in that joint and the segment is accelerated. For this model, there is an interface that receives a resultant torque for each joint in standard units. Therefore, any type of controller can be implemented as long as it uses this interface to control and receive information from the simulation. At present, two types of controllers have been implemented to control the virtual model. The first is a simple PID controller with variable set-points for each of the joints of the robot. The other controller is another simulation itself, it is a neural simulation based on the "integrate and fire" neural model. This is used for testing simple neural models and to investigate possible networks and neural parameters. This is expected to be replaced by the second version of NRS, which is currently under development.

Ode libraries can be found here.


This image shows the leg geometry use in the simulation.