23.11.2012 Views

Master Thesis - Hochschule Bonn-Rhein-Sieg

Master Thesis - Hochschule Bonn-Rhein-Sieg

Master Thesis - Hochschule Bonn-Rhein-Sieg

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

2. Overview on human-robot <strong>Master</strong> <strong>Thesis</strong> Björn Ostermann page 11 of 126<br />

2.2 State of science – Related Work<br />

The first approaches on safe interactions of humans and industrial robots were made in the 80s. After<br />

the first death of a worker related to a robot accident, reported in 1982 [1], safety studies in 1983<br />

concluded that a robot could only be safe if the robot could detect the approach of a human and react<br />

accordingly [2].<br />

In 1985 [3] Graham and Meagher of the Rensselaer Polytechnic Institute in New York investigated<br />

microwave, ultrasound, infra-red and capacitive sensors on their applicability in collaborative<br />

workplaces. They continued their work, together with Derby, in 1986 [4] resulting in the conclusion<br />

that “much research still needs to be done on robotics safety”. The Rensselaer Polytechnic Institute is<br />

still working in the field of safe robot-human collaboration. In one of their latest projects from 2007<br />

[5] Meisner, Isler and Trinkle used a galvanic skin response sensor, measuring the workers stress<br />

level, to generate robot trajectories.<br />

In 1989 [6] Takakura et al. outfitted a robot with torque sensors and programmed an algorithm that<br />

worked on contact sensing and memorization of objects. Although this approach was not feasible in<br />

praxis, due to the reduced speed of the robot, in order to avoid any damage by collisions, and problems<br />

with dynamic objects, whose changed positions can not be tracked, the idea of using torque sensors<br />

persists until today (see chapter 2.2.2).<br />

In 1992 and 1994 Novak and Feddema used capacity sensors for object avoidance [7]. The advantage<br />

of this approach is that the robot does not need to touch the obstacle, but can sense it in its vicinity.<br />

While increasing the sensor range from touch to close vicinity was an improvement, the problem in<br />

this approach is the different reactions of capacity sensors to different materials.<br />

Ebert and Henrich presented in 2002 a solution that used a multiple camera system to detect collisions<br />

between the robot and obstacles [10]. The problems they reported for this approach existed in cells<br />

being reported as occupied while free and vice versa and in the robot blocking the view of the cameras<br />

in certain positions. In later projects by Henrich this problem has been reduced but not completely<br />

solved (see chapter 2.2.3).<br />

Heiligensetzer and Wörn demonstrated in 2002 [20] [21] a Kuka Robot, equipped with capacity<br />

proximity sensors and haptic devices. Heiligensetzer continued the approach, using capacitive<br />

proximity and tactile sensors, in 2004 [19] and 2007 [57]. A problem in approaches using force<br />

sensors is always the reduced speed of the robot, which is necessary to avoid harm to humans. The<br />

robot presented in 2007 was therefore also equipped with collision absorbing material, making it<br />

possible to detect a collision before harm is done to the worker. Wörn continued his work in this field<br />

in 2004 [22], together with Yigit and Burgart, by developing a reflex based control for industrial<br />

robots, in which reflexes take over the control if a dangerous situation is detected.

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!