Mobile manipulators have a huge potential in fabrication processes as assistive co-workers, for the commission and pre-assembling of product components, or for the automatized refill of automatic machines. We focus at the Institute of Robotics of the Johannes Kepler University Linz (Austria) with our experimental platform (called omniRob) on the key challenges of this use cases:
Realization of a robot that offer a high maneuverability and dexterity
Cognition of the surrounding
Safe, collision free navigation of the robot base and arm
Traditional (car like) platforms are mostly built on a four-wheeled base with two steerable wheels. They are highly reliable but are not able to directly move sideways or turn on the spot. However, those operations are crucial for the operation in cost efficient and thereby narrow industrial shop floors. Our mobile platform overcomes these limits by the usage of special kinds of wheels (Mecanum wheels). They allow the robot to move instantaneously in any direction which is often referred to as an omnidirectional motion.
Robots that operate in industrial shop floors have to cope with dynamic environments. Dynamic means, on the one hand, that we have to consider that some objects (like fork lifts, humans, ...) move in the vicinity of our robot. Thus, the navigation scheme has to deal with changing situations or even predict the behavior of the surrounding. For sensing near objects we equipped our platform with ultrasonic distance sensors that sense the environment in a 360° view.
On the other hand, an observation of the surrounding is used for localizing the position of the robot. Hence, if the environment changes (pallets are for example moved) the robot has to recognize this and rebuild the map which he uses for localizing. The ultrasonic sensors are neither safe devices nor precise enough for the localization. We added therefore an additional LIDAR (Light detection and ranging) sensor for the localization.
The platform is additionally equipped with a light weight robot arm with seven rotational joints and a fin ray gripper attached at the end point. The arm is already used for picking up material and will in future also be used within preassembling tasks. Since the arm is not inside a closed and structured environment, it is highly important to sense the environment for potential collisions objects before starting the motion and plan collision free paths.
The arm is powered by Pluto servo drives which are operated in the interpolation position mode. The communication between the higher level components, which computes the desired path, and the Pluto servo drive, that controls the arm along this path, is done using the CANopen protocol. Future work will investigate more sophisticated control schemes (force control) that are capable to compensate small positioning errors at the endpoint during a preassembling task.