ENTANGLEMENT DETECTION OF A SWARM OF TETHERED

ROBOTS IN SEARCH AND RESCUE APPLICATIONS

T. R. Vishnu Arun Kumar and R. C. Richardson

School of Computer Science, University of Manchester, Manchester, United Kingdom

Keywords: USAR, swarm of robots, tether, entanglement detection.

Abstract: In urban search and rescue (USAR) applications, robots play a pivotal role. As USAR is time sensitive,

swarm of robots is preferred over single robot for victim search. Tethered robots are widely used in USAR

applications because tether provides robust data communication and power supply. The problem with using

tethers in a collapsed, unstructured environment is tether entanglement. Entanglement detection becomes

vital in this scenario. This paper presents a novel, low-cost approach to detect entanglement in the tether

connecting two mobile robots. The proposed approach requires neither localization nor an environment

map. Experimental results show that the proposed approach is effective in identifying tether entanglement.

1 INTRODUCTION

USAR is a time-critical process, which involves

complex and hazardous environment. Secondary

collapse, confined space, presence of fire and

poisonous gas in the environment pose serious

threats to the human and canine rescuers. Thus,

robots become inevitable in the field of search and

rescue. Mobile Robots can be used for a variety of

tasks such as localization, communication, victim

search, biomedical monitor delivery, environment

monitoring and reconnaissance. It is desirable to

deploy a swarm of robots as the survival rate of the

victims falls drastically after 72 hours.

Untethered autonomous robots depend on

wireless communication for information exchange.

When such robots are employed simultaneously in

the same area, issues such as interferences with other

systems, data security and international band

differences will arise as stated in (Fukushima et al.).

Tethered robots are used in a variety of applications

in ground, under-water and space environments

(Fukushima et al.). During recent times, tethered

robots are being employed in search and rescue

because tethers inherently provide robust data

communication and uninterrupted power delivery

(Fukushima et al.), (Hert et al., 1999). Tethers can

be used for navigating the robots through steep

slopes (Fukushima et al.) and also for pulling the

robot out when it gets stuck into debris (Perrin et

al.).

When a swarm of tethered robots is employed in

USAR environment, the tether might get entangled

with the obstacles or fellow robots. Entanglement

detection becomes an important problem that needs

to be addressed. Despite this, the merits of using

tethers make them a valid option even in search and

rescue scenario. Traditionally there are different

techniques to tackle entanglement detection. Most of

them need the robot to localize using an

environment map, which is then used to detect the

existence of tether entanglement. Navigation

planning could be done in such a way that there is no

tether entanglement (Hert et al., 1999). A method to

plan the shortest path for a tethered robot to a

destination point has been discussed in (Xavier,

1999).

In this paper, a novel, low-cost technique to

detect tether entanglement has been proposed. This

technique does not need the swarm of robots to be

localized. It does not require any environment map.

Section 2 discusses related work on localization and

map building in USAR. In Section 3, the tether

entanglement detection hardware is explained.

Section 4 throws light on experimental results and

Section 5 deals with the characterization of tether

entanglement using the experimental results. In

Section 6, a static model of tether is derived and the

experimental results are analyzed using the model.

Section 7 and 8 discuss current work and conclusion

respectively.

143

R. Vishnu Arun Kumar T. and C. Richardson R. (2007).

ENTANGLEMENT DETECTION OF A SWARM OF TETHERED ROBOTS IN SEARCH AND RESCUE APPLICATIONS.

In Proceedings of the Fourth International Conference on Informatics in Control, Automation and Robotics, pages 143-148

DOI: 10.5220/0001626001430148

 SciTePress

2 RELATED WORK

In USAR, communication and power supply play

crucial role. In case of autonomous robots using

wireless communication, the information exchange

is always noisy because of the thick concrete and

steel structures present in the search and rescue

environment (Perrin et al.). Also untethered robots

carry on board power supply for their operation.

This limits the life time of the entire system. So

tether based multi-robot systems are preferred over

untethered robots in search and rescue scenario. In a

tether based multi-robot system, a robot can detect

tether entanglement based on its pose in the

environment map and the pose of the fellow robots

in the same map. This technique is often referred to

as Simultaneous Localization and Map building

(SLAM) (Wijesoma et al., 2004).

One of the approaches for SLAM is to use

Inertial Measurement Unit (IMU) for localization

and Global Positioning System (GPS) for map

building. An Inertial Navigation system has errors

rising from factors like bias, scale factor

uncertainties, misalignment errors and noise

(Sukkarieh et al., Volume 19). Also in an uneven

terrain, gyrometer readings tend to be very noisy.

Fault detection and fault isolation form an integral

part of an IMU as stated in (Sukkarieh et al.,

Volume 19). In search and rescue scenario, GPS

becomes futile as discussed in (Gustafon et al.,

2005), (Cheng et al., 2004), and (Ramirez-Serrano et

al., SSRR).

Vision based approaches, which rely on

landmarks are used to localize the robot (Saeedi et

al., 2003), (Ramirez-Serrano et al., SSRR). As

search and rescue environment is complex and

unstructured, landmark based approaches are not

efficient (Cheng et al., 2004). The same reason can

be applied to (Saeedi et al., 2003) in which a vision

based approach for 3D localization and tracking has

been proposed. In this approach, distinctive scene

features extracted from the environment are used for

localization, but uncertainty in perception rising due

to different regions appearing similar is an issue to

be addressed.

The concept of Intelligent Dynamic Landmarks

is discussed in (Ramirez-Serrano et al., SSRR),

wherein some members of the robot group act as

portable landmarks for other robots to localize. In

(Gustafon et al., 2005), a swarm of robots have been

employed to achieve localization and target

identification. Line of sight approach is adopted to

localize heterogeneous teams of robots in

(Grabowski et al., 2004).

In order to detect tether entanglement, most of

the approaches need the robots to be localized and/or

an environment map. In this paper, a novel tether

entanglement detection technique has been proposed

that eliminates the need for localization and

environment map.

3 ENTANGLEMENT

DETECTION HARDWARE

The essence of this approach is that by recoiling

tethers and monitoring the force across the tether

during this process, entanglements, snags and

chafing effects on the tether can be detected. The

proposed system consists of two components (i) the

tether winding unit that pulls up tether slack (ii) a

sensor to detect horizontal forces across the tether.

The principle described here is to be applied to

swarms of interlinked tethered robots.

3.1 Tether Winding Unit (TWU)

Tether Winding Unit comprises of a pair of wheels

tightly coupled with a spring. One of the wheels is

driven by a 6 volt-5 Watt DC Motor. The tether

passes between a pair of wheels as shown in the

Figure-1. This unit is mounted on one of the robots

(Robot-A). There is an automatic wire coiling

system on the robot, which would hold one side of

the cable.

Figure 1: Tether Winding Unit (TWU).

3.2 Force Measurement Unit (FMU)

This unit comprises of a force sensor to measure the

force exerted on the tether. This unit is mounted on

the other robot (Robot-B). Tether entanglement

detection is carried out in the following steps:

Step-1: The tether connecting two robots is pulled

taut using the TWU.

Step-2: During that process, the horizontal force

exerted on the tether is measured using the FMU.

Drive wheel

Passive support wheel

Tension

spring

Multi-core tether

ICINCO 2007 - International Conference on Informatics in Control, Automation and Robotics

144

Step-3: Based on the pattern in which the force

exerted on the tether increases, it can be identified

whether the tether is snagged by an obstacle or not.

4 EXPERIMENTAL RESULTS

Tether Entanglement can be detected experimentally

through Snag Test, which makes use of TWU

attached to one robot and FMU attached to the other

robot. An open loop test is performed under four

different scenarios with or without obstacles. The

input voltage and the output force are recorded

during each test. Each of the four scenarios depicts

different levels of friction and different tether

dynamics involved when the tether is being pulled.

The scenarios are as follows:

Case-A is the scenario in which the tether is

freely hanging and there is no entanglement. Case-B

models the scenario in which the tether is stuck in

rubble and is subjected to friction at discrete points

along its length, as a result of which there is an

uneven movement when it is being pulled. In Case-

C, as the tether is wound around a pillar-like object,

the friction would be so high that the TWU would

not be able to hold the tether taut. Case-D depicts a

scenario in which the tether is bent by pillar-like

object and there is slow and steady movement of the

tether when it is being pulled. This is because the

friction is uniform through out the length of contact

with the obstacle. Figure 2 shows a snapshot of

rubble and pillar-like objects respectively.

Figure 2: Snapshot of rubble and pillar-like objects.

Figure 3: Schematic of Experimental Setup.

The snag test is conducted using unequal square

wave pulse, which excites the system so that the

dynamic effects of the tether being pulled are tested

on different lengths of the tether at different stages

of the experiment. Also the pulsing signal would

excite the stick-slip friction. A schematic of the

experimental setup is shown in Figure-3.

The force sensor readings are plotted against

time for all the four scenarios along with the unequal

square wave input signal as in Figure-4. Initially as

the input voltage is zero, the force exerted on the

tether remains a constant. After 6000ms, the voltage

raises to +2 V. This is reflected in the graph as steep

raise in the force value. It is also observed that there

is a lag between the time of application of the

voltage pulse and the time at which the force value

starts to raise. This is because the force exerted on

one end of the tether by TWU has to reach the other

end of the tether containing the FSU.

Figure 4: Input-Output Graph.

5 CHARACTERISATION OF

TETHER ENTANGLEMENT

In Figure-4, the force value for Case-A raises steeply

whenever the voltage pulse is applied because the

tether is hanging freely. For Case-B, there is uneven

raise in force value, as the friction from the rubble

acts at discrete point throughout the length of

contact of the tether. Case-C has no effect in force

value, as the tether is completely snagged and the

friction is so high that it is not possible for the TWU

to pull the tether. For Case-D, as there is uniform

friction throughout the length of the tether, there is a

smooth transition in the force value.

The force curves are analyzed using three

different methods namely Range of force analysis,

Area under the curve analysis and Static Model

Force Sensor

Robot - A

Robot - B

Object

(Obstacle)

Tether

ENTANGLEMENT DETECTION OF A SWARM OF TETHERED ROBOTS IN SEARCH AND RESCUE

APPLICATIONS

145

Analysis. From these analyses, an attempt has been

made to identify the type of snag from the force

sensor readings.

5.1 Range of Force Analysis

In this method, the force curve is preprocessed so

that any offset in force is eliminated. From the force

curves, it is very evident that when the tether is not

snagged by an obstacle (Case-A), the TWU holds

the tether taut and the maximum force is around 18

N. For all the scenarios where the tether is

entangled, the maximum force is around 6 N as

shown in Table-1. Thus Force Range can be used as

a parameter to model the type of snag.

5.2 Area Under Curve Analysis

If there are false spikes in the force curve due to

factors like slippage, drift, small object falling on the

tether, analysis using range of force would be

misleading. Area under curve analysis would reduce

such effects. After eliminating the offset from the

curves, the area is calculated as summation of the

product of the time interval (20ms) and

corresponding force value. The mean area of the

force curve for the three samples is calculated for all

the four different scenarios and listed in Table-1.

It could be seen that Case-A has maximum area

of around 400 square units. For Case-B and Case-D

the area is around 200 and 100 square units

respectively. Case-C has least area of less than

unity. Thus for a given input signal, based on the

area under the force curve, the type of snag can be

determined. For this analysis, time duration of the

test plays a significant role, as the area of the curve

is directly proportional to time.

Table 1: Average values of Force Range and Area under

Curve.

Case Force Range

(N)

Area under Curve

(square units)

18.195 400.0576

6.627 173.9507

0.124 0.1794

3.289 89.2996

6 TETHER MODELLING

From the experimental results it was observed that

the system is non-linear. This is evident from the

force curves in Figure-4. A non-linear model of the

system would give better insight into the behavior of

the system. A static model of the Tether

Entanglement Detection System (TEDS) is shown in

Figure-5. It comprises of two robots (Robot-A and

Robot-B) connected using a tether. Robot-A has the

tether linked with FMU. Robot-B has the tether

passing through TWU. The following are the

parameters, which influence the model.

pull

- Horizontal pulling force (N)

sag

- Sag angle of the tether (radian)

wheel

- Angular Velocity of drive wheel (radian /s)

- Half of the catenary length of the tether (m)

- Half of the horizontal length of the tether (m)

a - Distance between the vertex and the axis of

the catenary curve (m)

- Distance of the top of the catenary curve

from its axis (m)

6.1 Static Model of Freely Hanging

Tether - Derivation

A static model of the tether based robot system has

been derived. It is assumed that the dynamic effects

of the tether are negligible because the angular

velocity of the wheel α

wheel

is low. It is also assumed

that the mass is evenly distributed throughout the

length of the tether and the curve created by the

freely hanging tether is a catenary curve. A catenary

curve is the shape created by a chain-like object

fixed on both ends and hanging freely under the

force of gravity. The model can be used to determine

the horizontal pulling force (F

pull

) acting on the

tether and the sag angle (θ

sag

) of the tether (Flugge,

1962).

Figure 5: Tether Entanglement Detections System (TEDS)

Model.

pull

wheel

sag

Robot - B

Robot - A

ICINCO 2007 - International Conference on Informatics in Control, Automation and Robotics

146

6.1.1 Horizontal Pulling Force Derivation

The horizontal pulling force (F

pull

) is given by the

following formula:

aqF

pull

×=

(1)

where

q – weight per unit length of the tether (N/m)

In order to determine ‘a’, the following formulae are

used.

⎟

⎠

⎞

⎜

⎝

⎛

××

⎟

⎠

⎞

⎜

⎝

⎛

−= t

wheelinic

(2)

⎟

⎠

⎞

⎜

⎝

⎛

×=

sinh

(3)

where

ini

– Half the initial length of the tether (m)

R – Radius of the wheel (m)

Formula (2) is used to deduce the value of L

. This

value is used in formula (3) to find out the value of

‘a’ by creating the following non-linear equation:

⎟

⎠

⎞

⎜

⎝

⎛

⎟

⎠

⎞

⎜

⎝

⎛

×−=

aLaf

sinh)( = 0

(4)

Newton-Raphson iterative method is used to solve

the above equation. For that the derivative of f(a) is

needed.

⎟

⎠

⎞

⎜

⎝

⎛

⎟

⎠

⎞

⎜

⎝

⎛

⎟

⎠

⎞

⎜

⎝

⎛

⎟

⎠

⎞

⎜

⎝

⎛

−=

hhh

coshsinh)(

(5)

Assume the first guess

La =

then

⎟

⎠

⎞

⎜

⎝

⎛

−=

)(

⎟

⎠

⎞

⎜

⎝

⎛

−=

)(

…

This is repeated until

threshold

⎟

⎠

⎞

⎜

⎝

⎛

−

(6)

Threshold can be lower than 0.00001. Lower the

threshold higher is the accuracy of the value of ‘a’.

The value of a

k+1

is used as the value of ‘a’ in

formula (1).

6.1.2 Sag Angle Derivation

The formula for Sag angle is as follows:

⎟

⎠

⎞

⎜

⎝

⎛

sag

acos

(7)

where

⎟

⎠

⎞

⎜

⎝

⎛

×=

cosh

(8)

6.2 Analysis

The static model is verified experimentally by

measuring the force sensor readings for different

catenary length of the tether (L

) keeping the

distance between the robots (L

) as constant. Then a

graph is plotted with x-axis containing the ratio

between L

and L

and y-axis containing the

corresponding force readings. In the same graph the

force curve predicted using the static model for the

same value of L

is drawn as shown in the Figure-6.

It is observed that the actual force readings are very

close to the predicted values. This validates the static

model.

Figure 6: Predicted Force Curve Vs Experimental values.

Figure 7: Actual Vs Predicted Vs Corrected Force Curve.

Figure-7 shows the actual force readings and

those predicted using the static model for Case-A.

The predicted force curve closely follows the pattern

of the actual force curve except that it lags in time.

This is because the angular velocity of the drive

wheel will reduce when it is running under load

(tether passing through the wheels) compared to no

load condition. This is verified by simulating the

predicted force curve with 90% of the measured

angular velocity. The corrected curve matched very

closely to the experimental force curve as shown in

Figure-7. Another reason for the time lag could be

attributed to slippage of the tether. An optical

ENTANGLEMENT DETECTION OF A SWARM OF TETHERED ROBOTS IN SEARCH AND RESCUE

APPLICATIONS

147

encoder could be attached to the wheel to measure

the tether length, as it eliminates the time lag error.

From the above three analyses, it is evident that

the static model analysis is more promising than the

other two methods in terms of providing an accurate

model of freely hanging tether. Such a model can be

used to predict the force curve for freely hanging

scenario and the predicted curve can be compared

with the experimental curve. Based on the error

between the two curves, it could be identified

whether the tether is freely hanging or snagged with

obstacles. The static model can also be used to

identify different types of snags if dynamic effects

are introduced into it. One such approach could be

friction modeling.

7 CURRENT WORK

Currently friction modeling is being investigated to

understand the dynamic effects of the system. Also a

robust and low-cost 3D localization strategy for a

swarm of tethered robots is being developed. This

technique does not require an environment map for

localization. It includes a tether length measurement

unit (TLMU) and a tether orientation measurement

unit (TOMU) to localize the robot in 3D space.

TLMU comprises of an optical encoder attached to

the passive wheel of the TWU to measure the length

of the tether. TOMU consists of a joystick attached

to the end of the TWU to measure pitch and roll of

the tether.

8 CONCLUSION

In this paper a novel, low-cost and robust system,

which does not require localization or environment

map to detect tether entanglement has been

proposed. A static model has been derived for the

proposed system. Experiments have been conducted

to verify the validity of the approach. The results are

analyzed using three different methods. From the

analyses it is clear that the static model analysis is a

promising way of detecting entanglement because it

clearly identifies the scenario in which the tether is

freely hanging.

REFERENCES

Erik H. Gustafon, Christopher T. Lollini, Bradley, E.

Bishop, Carl E. Wick, “Swarm Technology for Search

and Rescue through Multi-Sensor Multi-Viewpoint

Target Identification”, Proceedings of the Thirty-

Seventh South-eastern Symposium on System Theory

(SSST 2005), Page(s):352 – 356

Edwardo F. Fukushima, Noriyuki Kitamara, Shigeo

Hirose, “A New Flexible Component for Field

Robotic System”, Proceedings of the IEEE

International Conference on Robotics and Automation,

San Francisco, CA, Page(s): 2583 – 2588

Doughlas P. Perrin, Albert Kwon, Robert D. Howe, “A

Novel Actuated Tether Design for Rescue Robots

using Hydraulic Transients”, Proceedings of the IEEE

International Conference on Robotics & Automation,

Page(s): 3482-3487

Patrick G.Xavier,“Shortest path planning for a tethered

robot or an anchored Cable”, Proceedings of the IEEE

International Conference on Robotics and Automation

(ICRA 1999), Volume 2, Page(s): 1011- 1017

Susan Hert, Vladimir Lumelsky, “Motion Planning in R

for Multiple Tethered Robots”, IEEE Transactions on

Robotics and Automation, Volume 15, Issue 4, August

1999, Page(s): 623 – 639

Chee Kong Cheng; Leng, G., “Cooperative search

algorithm for distributed autonomous robots”,

Proceedings of IEEE/RSJ International Conference on

Intelligent Robots and Systems, (IROS 2004), Volume

1, Page(s):394 - 399

P. Saeedi, D. G. Lowe, P.D.Lawrence, “3D Localization

and tracking in unknown environments”, Proceedings

of the IEEE International Conference on Robotics and

Automation (ICRA 2003), Volume 1, Page(s):1297 –1303

Salah Sukkarieh, Peter Gibbens, Ben Grocholsky, Keith

Willis, Hugh F. Durrant-Whyte, “A Low-Cost

Redundant Inertial Measurement Unit for Unmanned

Air Vehicles”, International Journal of Robotics

Research, Volume 19, No. 11, Page(s): 1089-1103

Alejandro Ramirez-Serrano, Giovanni C. Pettinaro,

“Navigation of Unmanned Vehicles Using a Swarm of

Intelligent Dynamic Landmarks”, Proceedings of

IEEE International Workshop on Safety, Security and

Rescue Robotics (SSRR), Page(s): 60 – 65

Robert Grabowski, Pradeep Khosla, Howie Choset,

“Development and Deployment of a Line of Sight

Virtual Sensor for Heterogeneous Teams”,

Proceedings of IEEE International Conference on

Robotics and Automation (ICRA 2004), Volume

1, Page(s):3024 – 3029

W.S. Wijesoma, L. D. L. Perera, M. D. Adams,

“Navigation in Complex Unstructured Environments”,

Proceedings of 8

International Conference on

Control, Automation, Robotics and Vision (ICRA

2004), Volume 1, Page(s): 167 – 172

W. Flugge, “Handbook of Engineering Mechanics”, First

Edition 1962, McGraw-Hill Book Company, Page(s):

4.7, 21.15

ICINCO 2007 - International Conference on Informatics in Control, Automation and Robotics

148