sensor networks or smart artifacts? - ubicomp 2007

Sensor Networks or Smart Artifacts?An Exploration of Organizational Issues of an Industrial Health and Safety Monitoring System

Gerd Kortuem David AlfordLinden BallJerry BusbyNigel Davies Christos Efstratiou Joe FinneyMarian Iszatt White Katharina Kinder

Computing Department Psychology Department

Management School

Lancaster University

Gerd Kortuem - Lancaster University - 2007 - [email protected]

What role can Ubiquitous Computing play?

mailto:[email protected]


Designing for People and Organizations The NEMO Project




Organization

Health & Safety

Regulations

Health and Safety System

Policies & Rules

Training

Equipment

Assessment Procedures

Surveilance

Data Records

Legal & Government Agencies

Management

SupervisorOperative

Health & Safety

Compliance

Business pressure

Social pressure

Shareholders / Clients

Society

Actors

Designing for People and Organizations The NEMO Project




Technology

Design process

Challenges & lessons




NEMO Drill NEMO Dosimeter



Duration of use (measured)

Vibration characteristics (known for each equipment, measured by manufacturer)

Soil condition (assumed)

operating heavy vibrating machinery such as hydraulic drillsand breakers. Our experimental system comprises a collec-tion of tools, augmented with wireless sensor nodes, personaluser devices and mobile computers that collaborate in anad-hoc manner in order to collect HAV exposure informa-tion. This information is available in real-time to workersin the field and subsequently to management via a back-enddatabase. Our design allows health and safety regulationsspecified at the enterprise level to be embedded within thewireless sensor nodes, allowing them to operate without anyinfrastructural support. This is crucial for workers who oftenhave to operate in areas with limited or no wireless commu-nication infrastructure. The system was evaluated througha two week field trial that took place with the collaborationof a major UK construction and maintenance company.

2. DOMAIN ANALYSIS

2.1 The Hand Arm Vibration ProblemLong-term exposure to hand arm vibration can lead to

serious health conditions such as “vibration white finger”(VWF) and in extreme cases to life-long disability. VWF istriggered by excessive use of vibrating machinery such as hy-draulic drills and breakers, and causes the fingers to becomenumb and begin turning white. As the disease progresses itcan become irreversible; the person su!ers increasingly fre-quent painful attacks at any time and may even lose theirfingers (typically this only happens in extreme cases, for ex-ample, when people are working with vibrating machineryin very cold conditions, as is the case in the forestry industryamong people working with chainsaws).

Extensive health and safety regulations exist to limit work-ers’ exposure to HAV. For example, in 2005 the UK Parlia-ment introduced the Control of Vibration at Work Regula-tions [10]. These regulations set limits for daily exposureto vibration, define methods for their calculation and for-mulate guidelines for monitoring an employee’s exposure tovibration. The guidelines place a responsibility on employ-ers to assess each individual employee’s exposure risk and toconsider the specific working conditions of each employee.When it is not possible to eliminate a worker’s exposure tovibration then suitable health surveillance must be put inplace and immediate action be taken if specific vibrationlimits are exceeded.

The damage caused by exposure to vibrations is a combi-nation of both the frequency of the vibrating tool and theduration of the exposure. Using a tool that vibrates at alow frequency for a long time can be as damaging as usinga heavily vibrating tool for a short time. Thus regulationsspecify two figures for exposure duration, a limit to the over-all daily trigger time (Daily exposure limit), and a limit onshort-term exposure to very high levels of vibration (Dailyexposure action). These limits are defined in terms of theaverage daily exposure dose A(8):

• Daily exposure limit value = A(8) of 2.5m/s2

• Daily exposure action value = A(8) of 5.0m/s2

A(8) is defined as ahv

p

T/8 where

ahv = actual vibration acceleration rate expressed inm/s2

T = actual exposure duration expressed in hours

In turn, ahv is composed of the root–mean–square accel-eration magnitudes in three orthogonal directions, x, y andz, at the vibrating surface in contact with the hand. Wheredaily vibration exposure A(8) is above 2.5m/s2 but below5.0m/s2, the following steps should be taken: Inform work-ers of risk, carry out regular health surveillance of worker,record assessment. Where daily vibration exposure exceeds5.0m/s2, the following actions must be taken: Limit usageimmediately, rotate workers, introduce other working meth-ods to reduce HAV. Since in practice ahv is not known fora specific piece of work, equipment manufacturers publishestimated ahv values for each individual tool, which can beused to estimate an operator’s exposure.

There is a distinct lack of automated solutions for assistingwith HAV H&S rule compliance. For example, vibration ex-posure data is typically manually recorded by operatives onpaper sheets, which are then entered by hand into a healthand safety information system. Moreover, key data such astrigger time must be estimated by operatives, most oftenhours after work has been completed. This of course raisesserious concerns with respect to completeness, accuracy andconsistency of captured data. The current practice can beimproved using mobile data entry solutions based on hand-held wireless computers. Yet, while mobile solutions reducethe need for paper forms they still su!er from the fact thatthey rely on human information gathering and recordingin the field. In addition, current solutions are tailored foro!-line processing of data in the back o"ce, ignoring thepotential benefits for real-time information in the field.

2.2 Requirements CaptureIn close collaboration with a major UK company that car-

ries out road maintenance and construction operations, weset out to design, implement and evaluate a Hand Arm Vi-bration (HAV) monitoring system, based on our approach ofaugmenting work artefacts. Our team consisted of computerscientists, psychologists, ethnographers and organisationalmanagement experts. Together, we engaged with the com-pany on various levels, ranging from high-level managementto workers in the filed, to investigate current work practices,identify technical and usability requirements, and anticipatedeployment challenges. The following gives an outline of ourapproach and introduces the requirements that we gathered.

2.2.1 MethodologyTo inform the design and elicit requirements we followed

a two-pronged approach. On the one hand, we undertookextensive work place studies and interviews to understandcurrent work practices. On the other hand, we generateddesign sketches to elicit concrete feedback.

In total we conducted over twenty formal interviews withmanagers and employees and 16 full days of work obser-vation. Field observation days also involved informal dis-cussions with upwards of 25 operatives going about theirdaily work, with the fieldwork being conducted across threedi!erent regional sections of the company, each of whichwith its own safety culture and traditions. An importantresult with respect to risks from vibration and noise expo-sure (“invisible risks”) was that it emerged that these wereappropriately assessed and their mitigation documented atan organisational level, yet were often underestimated orviewed as insignificant by operatives due to the apparentlyloose linkage between the risk and its outcomes. For exam-

Actual Exposure

Maximum Legal Exposure

?




Drill-Dosimeter pairing determined by proximity




In -Vehicle System

EnterpriseSystem

Network node

Actuator component ( including displays )

Sensor component

Intermittent communication

Figure 2: HAV monitoring system

A key issue for the NEMO system is to ensure that vibra-tion data is always recorded by the user device of the personwho is actually using the equipment. While in practice onlyone person operates each piece of equipment at each pointin time, there may be several ‘bystanders’, i.e. workers whoare in close proximity of the equipment without operating it.In this case, the NEMO system must ensure that data fromthe sensor node is recorded only on the operators user deviceand not on the bystanders’ devices. We call this the asso-ciation problem. This problem arises in any systems whereit is necessary to determine, for example, which worker isusing a specific tool.

We wished to design the system such that the NEMO userdevice could function as a general purpose H&S tag. Fromthis perspective, it made little sense to develop a custom so-lution for solving the association problem between a workerand a drill. For example, the particular characteristics ofthe drilling scenario, where the tool in operation and theoperative experience correlated vibrations, would suggest asolution similar to the one presented in [15] — identifyingthe tool user, through the correlation of accelerometer read-ings on both the tool and the operative. Another approachwould be to augment the drill handle with a short rangeRFID reader in order to read an RFID tag from the opera-tive’s glove. Although both these approaches would providea reliable solution to the association problem for this specificapplication, it would not generalise to other H&S scenarios.

We decided to implement two distinct approaches to theassociation problem in order to explore design alternatives.Firstly, we built a solution that required involvement of thedrill operator. Specifically, the system required the drill op-erator to press a button on their personal NEMO devicebefore they started using the drill. The button press wasonly required when a new worker was about to use the samedrill — i.e. on a change of operator. This approach su!ersfrom a number of obvious shortcomings. Most importantly,the requirement for minimal impact on users is not met —the workers have to change their work practice due to thenew system. In addition, the fact that the association relies

Figure 3: NEMO Devices

on user input makes the system susceptible to human er-ror. However, the approach has the benefit that when usedcorrectly, it provides an unambiguous indication of the tooloperator.

The second approach that we implemented was to mea-sure the proximity between the drill and all personal NEMOdevices using the wireless interface on each node. This ap-proach was developed on the assumption that over a periodof time, the worker operating the drill would spend moretime closer to the drill sensor node than all the other work-ers in the field. This approach appears to satisfy all require-ments established by the requirements capture phase, if itcan be shown to work reliably in the field.

In addition to the association problem, our abstract designalso raises issues with respect to support for “disconnectedoperation”. One of the characteristics of the HAV monitor-ing systems is that the operating environment is unknown,with minimal or no infrastructural support. For example, apatching gang is normally sent out to perform a task in acompany owned vehicle. However, the actual location wherethe tarmac is to be patched may be quite far from the parkedvehicle, depending on tra"c conditions and availability ofparking spaces. This implies that any infrastructure carriedwithin the vehicle may not always be available to the fieldcomponents. These characteristics have direct implicationson the communication requirements as well as the design ofthe overall system. Indeed, the mobile nodes that are de-ployed in the field (i.e. the drill sensor node and the personalNEMO devices) are required to operate without any accessto a back-end infrastructure. This implies that the fieldcomponents should be able to fully realise the H&S regula-tions specified by the company, in complete isolation. Thisindicates a departure from the traditional design of a wire-less sensor network where the sensor nodes push their sensorreadings to a back-end infrastructure that implements theapplication logic.

In the developed HAV monitoring system the functional-ity of monitoring the use of drills in the field, calculatingthe exposure to HAV, and reporting violations to HAV reg-




Technology

‣ Design process

Challenges & lessons




Design process

Google Maps http://maps.google.co.uk/maps/ms?f=q&hl=en&geocode=&ie=UTF8&om=1&msa=0&msid...

2 of 3 18/9/07 17:44


Technology

Design process

‣ Challenges & lessons




Invisible RisksRisk perception is poor Invisible risks

Lesson 1: Shift from H&S compliance to increasing awareness




Self-Image and Social Pressure Safety equipment often not used

Lesson 2: Personal and social “invisibility” becomes key design goal




Past Experiences & NarrativesGPS had been experienced as surveillance technology

Lesson 3: Technology narrative should be part of the design




Developers’ Blind-spot

Günter, H., Grote, G. & Boos, D. (2006): Organizational issues in ubiquitous computing, Paper presented at 22nd EGOS Colloquium, Bergen, July 06-08.Boos, D., Günter., H., & Grote, G. (2007): Organizational issues, technological frames and the development of a new ubiquitous computing prototype, Paper presented at 23nd EGOS Colloquium, Vienna, Austria, July 05-07.

(a) Wider organizational aspects are perceived to be of low importance (organizational culture, re-distribution of power, organizational disruption, societal issues, privacy and security concerns)

(b) Even organizational issues that are considered to be of importance often do not actually influence the design process




Technology Archetypes

Sensor network(sensing + data)

Smart artifacts(interaction)

Developers view problem in light of well-known system models




Technology Archetypes

Management

Operatives' behavior

Data logs

Exception reports

Health & Safety Rules

Operatives

Feedback

Automatic data capture

Enterprise

System

Personal Health & Safety Record

Operative'sbehavior

Context-sensitive notices

Sensor-network inspired H&S system

(top-down)

Smart-artifact inspired H&S system

(bottom-up)

Different system models imply different H&S solutions




Sensor Network Smart Artifact

Risk perception unchanged increased

Trust in rules unchanged increased

Intelligibility (information awareness) low high

Surveillance increased unchanged

Provable compliance supported not supported

Well intentioned rule violations more difficult unchanged

Rule formulation more rigid more rigid

Behavior-based safety not supported supported




Display Matters

Sensor-network inspired H&S system

Smart-artifact inspired H&S system

Unit w/o display was perceived as surveillance technology




Behavior-based Safety


Operative'sbehavior



Operative'sbehavior


Operatives as source of expertise and behavioral change

Collaborative safety




Points to Take Home

• Industrial workplace is opportunity for ubicomp• Design process must investigate organizational context• Architecture is not neutral




Thank You

Gerd Kortuem

[email protected]

http://www.comp.lancs.ac.uk/nemo








• James Brown, Joe Finney, Christos Efstratiou, Ben Green, Nigel Davies, Mark Lowton and Gerd Kortuem. Network Interrupts: Supporting Delay Sensitive Applications in Low Power Wireless Control Networks. CHANTS 2007.

• Efstratiou, C., Davies, N., Kortuem, G. Finney, J., Hooper, R., and Lowton, M. Experiences of Designing and Deploying Intellignent Sensor Nodes to Monitor Hand-Arm Vibrations in the Field. Proceedings of MobiSys 2007, San Juan, Puerto Rico, June, 2007

• Iszatt White, M. Catching Them At It? An Ethnography of Rule Violation. Symposium on Current Developments in Ethnographic Research in the Social and Management Sciences. 13th-14th September 2006. Liverpool, UK.

• Busby, J. and Iszatt-White, M. Pushing the Boundaries of HRO. Thinking: Non-complex and Uncoupled but still Deadly. SRA Annual Meeting - Risk Analysis in a Dynamic World: Making a Difference, 3rd-4th December 2006. Baltimore, Maryland

• Davies, N., Efstratiou, C., Finney, J., Hooper, R., Kortuem G., Lowton, M. Sensing Danger – Challenges in Supporting Compliance in the Field. 8th IEEE Workshop on Mobile Computing Systems and Applications (HotMobile 2007), Tucson, Arizona, February, 2007

• Lowton M. and Finney, J. Finding NEMO: On the Accuracy of Inferring Location in IEEE 802.15.4 Networks. Proceedings Workshop on Real-World Wireless Sensor Networks (RealWSN 2006), Uppsala, Sweden June 2006.



sensor networks or smart artifacts? - ubicomp 2007

Technology

nemo system

industrial health

displays enterprise

exposure exposu actualdaily

sensor networks

drilldosimeter pairing

hs compliance

toa key issue