How a Cobot Chooses to Stop Before it Hurts You?

In our last article, we tried to provide the reader with a general overview of safety norms and directives for collaborative robots.
Today, we will focus instead on how safety is actually implemented in collaborative robots, starting from their formal definition and requirements.
As already mentioned, according to the current normative ISO TS 15066 [1], collaborative robots are defined as industrial robots that can be used in collaborative operations. The approaches this collaboration can actually be achieved with are built upon 4 main features, namely:

1. Safety Monitored Stop
2. Hand Guiding
3. Speed and Separation Monitoring
4. Power and Force Limiting

Generally speaking, all these features allow standard industrial robots to be employed in new collaborative scenarios in which human operators and robots can share, to some extent, their workspace [2]. This is achieved by implementing safety mechanisms relying on external sensors for triggering a stop (1) or reduce the system’s speed (3) if a person accesses the workspace. In the case of (2), the human operator takes control of the robot by using a certain hand-operated device.
However, only the last feature (4) is able to permit human-robot safe cooperation, which is actually what cobots are born for. Therefore, in this article, we are going to focus only on power and force limitation.

Typically, cobots implementing power and force limitation (PFL) are capable of sensing all the external forces applied to their body thanks to on-mounted internal sensors and react accordingly, for instance by stopping the robot if these forces exceed a certain limit level. This helps to avoid injuries and brings to a new level of safety, not achievable in the first 3 classes of collaborative robots mentioned above.
Besides, since no external safety devices such as fences or lasers are needed (depending on the risk assessment output), the cost of the cell is considerably reduced.

The question is: how cobots manufacturers achieve reliable PFL?
Actually, there exist different sensing settings depending on the type of sensors involved and on which part of the robot they are mounted on.

Joint torque sensing: The joint’s torque is estimated by measuring the motor current or by mean of a torque sensor attached to the joint. A similar approach is developed in Universal Robots UR series in which the force is derived from motor current and position of the encoders [3].
6 dof force/torque sensing: Ring-shaped sensors are used to measure all moments and forces acting on them. They can be connected to each joint, to the robot’s base or to the robot’s end-effector.
Some examples of cobots implementing 6 dof f/t sensing at each joint are the IWAA by Kuka and the HC10 by Yaskawa [4].
Skin sensing: The sensors are mounted on the robot’s surface measuring the applied external force. Despite being mostly used for research purposes and not very popular at industrial level, we can find some examples of cobots provided with tactile sensors on the market like the Comau’s Aura or the APAS by Bosch.
Mechanical PFL: Power and force are limited mechanically by employing variable stiffness actuators or non-stiff elastic actuators. One famous example of Mechanical PFL is the Rethink Robotics’s Baxter which mounts Series Elastic Actuators (SEA) developed by MIT [3].

Even though cobots coming with PFL introduce a new level of safety in human-robot collaboration and pave the way to new applications by easy and safe integration, it is necessary to remark the importance of thoroughly carrying out the risk analysis for any new robotic application, regardless the kind of robots (or cobots) being used.

What will be the next game-changing technology applied to collaborative robots to become more and more reliable and easily integrable in production lines? How will risk analysis be evolving in the next years? Will it eventually get unneeded?