Closed-Loop Control

      Closed-loop motion control is absolutely essential to soft robotics. The term closed-loop refers to this conceptual circle: The input of sensors represents positions or locations. The output of these sensors is sent to a computing device. The computing device determines desired target positions or locations as the robot moves. The computing device sends control signals to the effectors (sources of motion like motors or pistons). The effectors work to achieve the calculated position or location. Finally, the new positions change what the position sensors are detecting. This completes the conceptual circle called the closed-loop. This is the circle of information transfer that regulates position. There can be many closed loops in a robot. This means that the computer (according to its program) may make the actual position the same as the desired position despite environmental influences. The computer continually checks the sensors and adjusts position by means of effectors.

      The absence of closed-loop control has come to be known as open-loop control. The precision of open-loop control depends upon the precision with which the parts of the robot are made, while the precision of closed-loop control does not.

      Open-loop control uses an effector, but without the benefit of the sensing and correction. This can be useful only when the environment is known to an adequate level of precision and the parts of the robot are made with enough precision to render such actions accurate.

      People exhibit a great deal of closed-loop control. It is well worth observing this property in our ordinary actions.



Compliance and Force Feedback

      A robotic arm or manipulator is compliant if it can bend, stretch, or compress when it is subjected to mechanical stress. This cushions the blow in the case of collisions. When a compliant part collides with an object, the initial force may not be very great. Although the force may continue to climb, the force does not instantaneously climb to a destructive level. This can give the computing device enough time to change the direction of motion.

      Force sensors are used to monitor the rise in force associated with the deformation of compliant parts. This force sensing is also known as force feedback.

      Compliance is useful also in another way: When used together with force sensing, compliance can facilitate searching for objects in the environment by intentionally causing collisions. At appropriately low forces, these intentional collisions provide an opportunity to sense the presence and location of an object being sought. It is much less expensive than vision systems and very often, it is all that is need.

      Like closed-loop control, compliance and force feedback are essential to soft robotics.



Precision Gain: Highly Precise Motion from Low-Precision Parts

      Closed-loop control, compliance and force feedback make possible low-cost construction methods. This is because robotic devices designed in this way have a very important property: The precision with which such devices move does not depend upon the precision with which their mechanical parts are made!

      This can be understood from the way people move and navigate. Suppose a person wakes up one morning with an arm that is one eighth of an inch shorter than it was the previous day. Reaching for a morning beverage, the hand starts its journey. It comes closer and closer to the handle of a cup. It is unlikely that anything unusual would be noticed. Feedback from the senses tell one when to stop. A person does not strain to remember how long he or she should press ahead or how fast to move. Anybody would simply move the hand until it gets there.

      Contrast this with the world of machined, ground, and polished steel parts used in most of today's industrial equipment. A part being an eighth of an inch short would be absolutely unthinkable!

      This means that instead of using hardened steel parts made with great precision, parts may be rough-cut or extruded, or even made from materials having poor shape stability such as wood, rubber, or spring wire.

      Here is another example: A piece of lumber is on a table saw. A robotic arm pushes the board back toward a backstop. This backstop prevents the board from falling off of the far side of the table. One end of the board happens to reach the backstop first, but soon the board is flush against the backstop.

      If the robot arm continues to move past this point, the board could be dented or crushed. If the accuracy of movement is low and there is no compliance, the robot arm may crush the wood or stop short of pressing the board flush against the backstop.

      On the other hand, if force feedback and compliance are used, we have two fortunate circumstances: 1.) the force sensing provides an indication to the computing device that further movement should cease; and 2.) the bending or compression of the robot arm provides the time that the computing device needs to stop or redirect the effectors.

      Without compliance, damage could be done in the time between sensing the opposition of the backstop and the computer's action to stop the effectors involved.

      In the same way, the board can then be moved (let's say to the left) toward an end stop.

      In this case, the dimensions of the table, the backstop, and the end stop provide the accuracy of positioning that is needed to determine the placement of the cut. Robotic devices using closed loop motion control, force feedback, and compliance can utilize the same sort of tools, jigs, and fixtures that people do.

      Here is an example that does not derive measurement from an external tool: A wrench is use to torque down a bolt. Turning must stop when the force at the end effector reaches a certain value. The accuracy of the force sensor in the robotic device determines the accuracy of the result. This accuracy was not determined by the precision with which the mechanical parts were made. If the wrench handle can be held at various distances from the bolt, a distance sensor of some kind would be involved; but again, the sensors and not the structural parts provide the precision. Since the price of sensors and computers are going down faster than the price of precisely made mechanical parts, this should be the source of the precision. Soft Robotics saves money while doing a the job better.



Indirect Actuation

      Indirect actuation is a design policy that centralizes the source of mechanical power. It is one of the frontiers of soft robotics. Instead of using an expensive motor on every limb and finger (one effector on each joint), work is applied through linkages and intervening mechanical devices controlled perhaps through solenoids. If it can be achieved, this feature can reduce expense substantially.



Durability and Repairability

      Technical innovations are so frequent that it is better to emphasize repair than endurance. If a robot can survive a task for five years and is easy to repair, it seems a waste of money to extend its expected repair-free life too much longer.