FORCE LIMITING DEVICES

Even if their design is optimized to reduce their inertia and power, some manipulators remain too dangerous to interact physically with people. Indeed, during a collision between a human and a robot, the contact forces can be large, particularly if the person is constrained between the robot and a wall. Reducing their velocity allows the reduction of the danger associated with the manipulator. However, some robots can apply static forces exceeding a safe magnitude. In this case, mechanisms limiting the interaction forces can be used.

The desired behaviour of such mechanisms is to preserve the manipulator's stiffness until an external force applied on the robot exceeds an arbitrary threshold. When this happens, the impacted robot link disengages and its motion becomes mechanically disconnected from the rest of the manipulator. As a result, the effective inertia of the robot is reduced during the collision whereas its behaviour remains unchanged under normal conditions. It is possible to create this type of mechanism using torque limiters, as shown on Figure 1.

This type of mechanism is particularly well suited for suspended robots because only a fraction of the manipulator can collide with a person. Due to their parallel architectures comprising parallelograms, the force threshold which activates a torque limiter only depends on the orientation of the force and not on its application point. These force limiting devices are Cartesian because the relation between the applied force and the couples generated at the limiters is independant of the manipulator's configuration. In addition to the simple one degree of freedom version of Figure 1, it is possible to obtain architectures with two and three degrees of freedom which react to contact force in horizontal directions or in any possible direction.

Even though Cartesian force limiting devices are effective for suspended robots, many manipulators intended to physically interact with humans have serial architectures. In this case, it is possible to place torque limiters in series with each actuator of the robot. This positioning ensures that the efforts are limited in the directions in which the robot can move. However, the relationship between the forces applied on the manipulator and the articular torques is complex, because it depends on the contact point location and the robot's configuration. To alleviate this drawback, it is possible to use torque limiters with electronically adjustable thresholds and to optimize the architecture and motion of the robot. The next figure shows the architecture and a picture of a serial robot prototype for which adjustable torque limiters are placed in series with each actuator.