Source: dm_control commit 773d5b5d94d8f58943cf56a66b9b9755818fe5b4; cartpole model with 2 chained poles
body_mass = [0.0; 1.0; 0.1; 0.1;;]
body_inertia = [0.0 0.0 0.0; 0.010833333333333334 0.01666666666666667 0.020833333333333332; 0.009424592767295598 0.009424592767295598 0.00010010377358490566; 0.009424592767295598 0.009424592767295598 0.00010010377358490566]
body_ipos = [0.0 0.0 0.0; 0.0 0.0 0.0; 0.0 0.0 0.5; 0.0 0.0 0.5]
dof_damping = [0.0005; 2.0e-6; 2.0e-6;;]
actuator_gear = [10.0 0.0 0.0 0.0 0.0 0.0]
actuator_ctrlrange = [-1.0 1.0]
Control interval 0.02 s, two upstream 0.01 s RK4 steps; force u maps to action u/10
Task additions: noisy partial sensing, bias/dropout/mass faults, constrained MPC; not original suite reward protocol
