机器人学：力控（引用）

From <Robotics Research 101: Getting Started with Force Control> by Alex Owen-Hill
机器人学研究：力控

Types of Force Control

Pure force control uses a similar principle to the pure position control described above. The only difference is that you command the robot with a force instead of a position. Thus the robot will try to maintain the force which has been designated on the environment.

Pure force control also has some limitations. For example, it is not so great when moving the robot through free space (i.e. when there is no applied force).

In practice, a combination of motion and force control are used to get the best of both worlds.

There are a variety of different force control techniques, some of which are described below. These are all “active” force controllers (i.e. where force data is incorporated into the control loops). You can also have “passive” force control, such as when a bendy material or springs are integrated into a manipulator to ensure that it cannot apply a high force.

1、Indirect force control

These methods are primarily motion controllers, which only apply a force constraint when the position of the robot deviates from the target position. These controllers do not explicitly “close the force feedback loop.” This means that, technically, they don’t need force torque sensors to be implemented. However, often it is a good idea to use force torque sensor data to improve the controller.

• Impedance control - This applies a mass-spring-damper between the target position and the actual position of the robot. Imagine that you attach a spring between your finger and some target position in the air. The further you move your finger away from that point, the more force the spring applies to pull you back to that point.This video shows what impedance control looks like on a robot.
  
• Admittance control - This is the inverse of impedance control (more or less). It can be better imagined as pushing your finger through a very viscous substance, like honey or wet sand. The more force you apply to the substance, the further your finger will move.This video shows what admittance control looks like on a robot.
  
  There are also special cases of impedance and admittance control, such as stiffness control which actively varies the stiffness of the robot joint.(demonstrated quite clearly in this video)

2、Direct force control

These methods incorporate both force and motion into the control of the robot. They require two different inputs - the target position/motion and the target force. As a result of this, some direct force controllers must have a very clear model of the task.

This can sometimes be restrictive when you are using the controller in unknown environments or for unknown tasks. However, when you do know explicit details about the task, these controllers allow you to independently control each axis for either force or motion depending on the needs of the task.

• Hybrid force/position control - This control method is quite popular. It separates all 6 axes of the task (3 force and 3 torque) and applies either a motion based control or a force based control onto each of the axes. Unconstrained (free) axes are controlled in position while constrained axes are controlled by applying a constant force. This video shows what hybrid force/motion control looks like on a robot.
• Parallel force/position control - This is an option if you want to use direct control, but do not know details of the task. It works by implementing both force control and motion control simultaneously. The controller is set up so that the force control has more effect on the output than the motion control does. This means that the robot will move under the motion control until an external force is applied.