INTELLIGENT ASSIST DEVICE PROTOTYPE
Several technologies developed in the human-robot collaboration research axis were integrated in a full-scale prototype in order to respond to the specific needs of the laboratory's industrial partner GM Canada. This prototype consists in an intelligent assist device which helps workers to displace and install large components during the vehicle's assembly. More specifically, the target application is the assembly of instrument panels, pre-assembled in a different plant than where the final assembly is located. This constitutes a difficult task due to the load which can reach 113kg and which needs to be displaced and installed during a typical 75 second cycle.
The required degrees of freedom for this task are the translations in all directions and rotation along a vertical axis. For safety reasons, velocities are limited to 1 m/s for both horizontal directions, 0.25 m/s for vertical motion and 1 rad/s for rotation. Accelerations are limited to 1 m/s2 for all translation directions and 1 rad/s2 for rotations. Figures 1 and 2 present an overview of the intelligent assist device prototype.
Different approaches were implemented to reduce the inertia and power of the prototype. The actuators were placed on the manipulator's base using belt routings (and a cable for vertical displacements). All the routings are decoupled, i.e. the X actuator cannot create motions in Y, Z or Theta and vice versa. Placing the actuators on the base reduce the moving inertia which in turn reduce the actuators' efforts to obtain the required performances. Figure 3 presents the belt routings for the actuation of the X, Y and Theta axes.
For vertical displacements, a gravity compensation system based on counterweights is implemented in addition to fixing the actuator on the base. Because the cable routing is decoupled from the other routings, the motion in the X, Y and Theta axes are not affected by the additional inertia of the counterweight. Also, because the counterweight is mobile on a lever arm, it is possible to adjust it to compensate various loads. Therefore, the actuator must provide a force to compensate the routing friction and the inertia of the load (and the counterweight), but not its weight. For displacements with small accelerations, this results in a significant power reduction for the Z actuator. The cable routing and the adjustable balancing system are represented in Figures 4 and 5, respectively.
From the design view point, the prototype mass was reduced using aluminum instead of steel for various components. Also, the end-effector is conceived to place its centre of mass on the rotation axis. In addition to reducing the rotating inertia (and the required actuator power), this simple measure allows the reduction of the bending moments on the end-effector and the trolley. These components can thus be designed to be lighter without reducing their strength. This is realized by designing the end-effector in a C shape instead of a conventional L shape, such that the instrument panel can be inserted into the car being assembled without interference with its roof. The two types of end-effectors are illustrated in Figure 6.
In order to measure the user intentions, a handle composed of photo-interruptors and compliant mechanisms is used. This allows the measurement of the efforts associated with the manipulator's degrees of freedom. The electromagnetic noise of the signal is very low and it does not drift over time, which constitutes important advantages compared to commercial alternatives. The forces are measured with precision up to 40N, although it is conceived to support larger forces for robustness. Figure 7 presents a CAD model of the developed handle.
Safety is a major issue in the design of an intelligent assist device which is capable of lifting heavy loads. Generally, industrial standards do not allow a simple computer to directly control a robot. Programmable logic controllers (PLC) are prefered due to their reliability. However, they do not possess the necessary flexibility to implement complex control algorithms.
In order to obtain a flexible yet safe controller, a programmable logic controller dedicated to safety is used. It constantly supervises the computer which controls the robot and can intervene, for example, by applying brakes if the computer does not respond. The programmable logic computer is also responsible for managing the three operation modes. The autonomous mode allows the manipulator to move without human intervention to, for example, pick up the next component to be assembled. The interactive mode is the principal mode of operation in which the user controls the assist device via the sensitive handle. Finally, the unpowered mode allows the deactivation of the actuators and brakes in order to move the device manually, which is useful to free the assembly line in the case of a major failure. Figure 8 presents the safety control scheme of the prototype.
The intelligent assist device prototype is an important development tool for the robotics laboratory. It allows the study of different control schemes for physical human-robot interaction in the scope of a realistic full-scale application. The robustness of the developed algorithm can thus be tested, which is not possible with the small manipulators usually seen in research environments.