Very small simulation time step for successful walking!!!
Dear Gazebo community
I need to simulate and control a simple 3D biped robot using gazebo and ros. Currently I have developed the package (named mhr) which containts urdf model of the robot, low level controllers and stable offline saved trajectory for each actuated dof's of robot's walking on flat ground. Please consider that the trajectory is stable right now based on ZMP criteria.The problem is, I have to choose very small time step (dT = 0.00005 sec) to achieve successful walking in simulation which cause real time factor to be very smaller than 1. I am not sure about choosing proper solver between ODE, bullet, and Simbody as well as adjusting other related parameters of solver (e.g cfm, erp, etc). My parameters of ode solver are as follow:
<physics type='ode'>
<max_step_size>0.0001</max_step_size>
<real_time_factor>1</real_time_factor>
<real_time_update_rate>10000</real_time_update_rate>
<gravity>0 0 -9.8</gravity>
<ode>
<solver>
<type>quick</type>
<iters>200</iters>
<sor>1.0</sor>
</solver>
<constraints>
<erp>0.2</erp>
<cfm>0</cfm>
</constraints>
</ode>
</physics>
Complete package is accessible through Google Drive which should be extract in your "catkin_ws".
Mirror 2 Dropbox
Also please note that contact parameters of right and left foot are considered as follow:
<gazebo reference="RightFoot">
<material>Gazebo/Red</material>
<kp>8000000.0</kp>
<kd>1000.0</kd>
<mu1>1.5</mu1>
<mu2>1.5</mu2>
<maxVel>1.0</maxVel>
<fdir1>1.0 0.0 0.0</fdir1>-->
</gazebo>
I would be grateful if anybody could help me to increase simulation time step and improve my simulation.
Best Regards,
Masoud