IEEE Robotics & Automation Magazine - December 2011 - 27

HF1 = 1
S1

W
CF = 1

Ekin ≤

CF = 1
S2

S3
FT1 = 1

Symbol

Meaning

States

W
R
S1
S2
S3

Wait State
Running State
Collision Strategy 1 State
Collision Strategy 2 State
Collision Strategy 3 State

Events

Go Event for Starting Process

Signals

HF1
FT1
Ekin

Collision of Severity-Level HF1
Collision of Severity-Level FT1
Kinetic Energy of the Manipulator
Kinetic Energy Threshold

∋

Investigated Tools
The variety of tools one could analyze are basically countless and, therefore, a representative selection of tools with
different sharpness was carried out (Figure 6). Up to now,
there is no benchmarking test of tools, since the underlying
biomechanics is not fully understood yet. Therefore, we
chose typical household objects such as knives of different
sharpness, a scalpel, and a screwdriver. These tools were
selected as a reasonable choice of potentially dangerous
ones one could think of in robotic applications. They were
removed from their original fixtures and glued into new
mountings. Therefore, a fixed connection between the tool
and a robot can be guaranteed, and no compliance reduces
the transferred forces. This selection enables us to test a
wide range of cuts and stabs with varying blade characteristics and sharpness. Currently, we are preparing a systematic biomechanical test to provide a set of benchmarking
objects for investigating soft-tissue injury. The tools were
tested in the same condition they were bought, except for
the fact that they were glued into a rigid mounting to
remove eventually beneficial compliances.

∋

possibly occurring if a robot with a sharp tool penetrates a
soft material is analyzed. The dynamics of such an impact
is especially worth being investigated since, during rigid
(unconstrained) collisions [13], the dynamics is so fast that
a realistic robot is not able to reduce the impact characteristics by the collision detection and reaction. However,
during our previous investigations, a subjective safe feeling
could be definitely experienced by the users. Despite this
limitation in reactivity to blunt impacts, it was shown that
the necessity of countermeasures is not absolutely crucial
since rigid free impacts pose only a very limited risk at the
typical robot velocities up to 2 m/s. This is definitely not
the case for soft-tissue injuries caused by a stab because the
injury severity due to penetration can reach lethal dimensions. The particular worst case depends on the exact location by means of potentially injured underlying organs.
Because of the much slower dynamics compared with rigid
impacts, the requirements on a reactive robot concerning
detection and reaction speed are somewhat relaxed and not
unachievable for such situations as exemplified in the "A
Simulation Use Case with the LWR-III" section. It seems
surprising at first glance that it is not possible to counterbalance rigid blunt robot-human impacts by means of controls
that are definitely not life threatening, but at the same time,
dangerous or even lethal contacts with tools seem manageable to a certain extent. (Please note that we refer to impact
speeds of up to 2 m/s.) One purpose of the present experiment is to prove this statement.
In the framework of this article, the situation in which
the robot moves in position control with/without collision
detection by utilizing joint torque sensing is considered.
The contact force is measured with a JR3 force-torque sensor in the wrist. Please note that this sensor is only used for
measurement and not for collision detection.

Figure 5. Combining the different collision reaction techniques
to complex reflex behaviors.

Silicone Block
As a first experimental contact material, a silicone block
was used to get a feeling for the sensitivity and effectiveness of the collision detection and reaction for soft contact. (The used silicone was Silastic T2 with a Shore
hardness of A40.) These first tests were conducted at a
Cartesian velocity of 0:25 m/s, which is the recommended velocity according to International Standards
Organization (ISO) 10218 for collaborative robots [2].
The mounted tool is the kitchen knife. The desired goal
configuration was located at a depth of 8 cm in the silicone block. Without any reaction strategy, the achieved
penetration was 35 mm at a contact force of 220 N with

Blade
Fixture
Mounting

(a)

(b)

(c)

(d)

(e)

Figure 6. Investigated tools: (a) Scalpel, (b) kitchen knife, (c)
scissors, (d) steak knife, and (e) screwdriver.

DECEMBER 2011

IEEE ROBOTICS & AUTOMATION MAGAZINE

Table of Contents for the Digital Edition of IEEE Robotics & Automation Magazine - December 2011