Qiang,
You should be able to accomplish what you want with torque control of the actuators.
The torque (effort) controller in the actuator acts on the sensed torque directly on the output, and the details of how the actuator controls the output torque can be tuned by adjusting the gains on the effort controller. Details on these gains can be found
here.
To start off with I would increase the P-gain of the effort controller (Effort Kp in the API) to something like 0.5 (significantly more than the default value). You can also try adding in an I-gain (Effort Ki in the API) of around 10 as a starting point. Play around with the values to see what works for your application. Taking the P-gain higher will help reduce the resistance at zero-torque and improved tracking, but at some point it will become unstable if the value is too large. The same goes (even more so) for the I-gain.
You can tune things interactively using the
Gains Tab in Scope. Once you change a parameter, click the "Send" button to have the new gains be applied to the actuator. If you want the gains to persist after the actuator is rebooted, turn on the "Persist Sends" slider before pressing "Send". Once you have the initial tuning done, the best practice is to save the gains to an
XML file and load them onto the actuator in whatever API you are using to control the actuator. This makes sure that the tuning of the actuators are consistently set and documented with the rest of your code.
Hope this helps,
-Dave