You are on page 1of 7

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/335909625

Modeling the ergonomics of Goods-to-Man order picking

Conference Paper · May 2019

CITATIONS READS
0 61

4 authors, including:

Nelson King Jaby Mohammed


Khalifa University of Science and Technology Illinois State University
56 PUBLICATIONS   1,113 CITATIONS    29 PUBLICATIONS   90 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Educational Technology View project

Industrial Visualization and Modeling View project

All content following this page was uploaded by Nelson King on 19 September 2019.

The user has requested enhancement of the downloaded file.


Proceedings of the 2019 IISE Annual Conference
H.E. Romeijn, A. Schaefer, R. Thomas, eds.

Modeling the ergonomics of Goods-to-Man order picking

Shaikha Al Shehhi, Nelson King, Saed Amer, Jaby Mohammad


Khalifa University of Science and Technology
Abu Dhabi, United Arab Emirates

Abstract
The move towards a goods to man order picking strategy increases worker productivity by reducing travel time but
increases the possibility of repetitive motion injuries. An anthropometric simulation of the goods to man order
strategy was developed using Siemens Jack so the Task Analysis Toolkit could be used to study the effect of picking
hundreds of items an hour. The simulation assigns order pickers of different statures (Jack and Eva) ten items of
different shapes and weight to random bin locations on a shelf caddy for an hour. The preliminary analysis shows
that the lower back and metabolic energy thresholds are approached with significant flexion around the neck for
both Jack and Eva.

Keywords
Order picking, distribution center, human factors, repetitive motion injuries

1. Introduction
Warehouse order picking retrieves products from storage locations to fulfill a customer order. It is a labor-intensive
activity that traditionally required the order picker to walk many kilometers daily in a “man-to-goods” layout to pick
items scattered throughout the warehouse. While order picking software planned the optimal picking route, much of
the day was spent walking rather than picking. Amazon now uses a “goods-to-man” layout where the robots move
the storage bins to the workers at a dedicated picking station. By doing so, an Amazon picker is expected to retrieve
around 400 items in an hour [1]. While “goods-to-man” nearly eliminates walking time, the risk of repetitive motion
injuries increases due to the near constant order picking motion. Workers are working really fast in order to meet
their timed picking quotas reported anecdotally to be ten seconds [1].

2. Literature Review
Order picking has been identified as one of the costliest, labor-intensive, and time-consuming activity in every
warehouse. The literature on order picking has focused on the development of decision support models which can be
used to plan the picking process to reduce travel time [2]. This focus is mostly on one of the following problems; the
routing of the order pickers inside the warehouse, the layout of the warehouse, the pooling and batching of orders
and where each product is assigned and its specific location in the storage. One of the outcomes of this line of
research was the goods-to-man or parts to picker picking system [3]. However, these models neglect the importance
of human factors given the physical strain of manual order picking [4]. Grosse et al. [5] claimed that overlooking the
human characteristics while planning order picking results in poor system performance due to the conditions in
which workers must operate.

The tradeoff becomes one of minimizing walking distance, which accounts for high metabolic output [6] and
repetitive motion injuries of constant order picking [7] when the goods received directly at the order picking station
From the literature, it’s clear that a lot of order picking strategies neglect the human factors when modeling their
system and focus mainly in the performance aspects. Repetitive motion injuries from picking are masked with long
travel times allowing fatigue recovery. Fatigue recovery from a man-to-goods strategy focused on recovery from
walking long distances carrying or pushing picked items [8] More research and studies, as proposed here, should be
conducted to address the human factors aspect when designing order picking systems such as fatigue recovery,
lower back problems, metabolic energy expenditure and working posture. Following such a strategy will extend the
number of years that workers spend at their jobs and stay injury free without facing serious health concerns, and also
having a high-performance work practice.
Al Shehhi, King, Amer, Mohammad

3. Study Objective
The study focuses on the human performance of an order picker, especially the effects of repetitive motion on their
body. The simulation of picking motions may help in measuring the impact on the human body. Siemens Jack
software allows the picking process to be simulated to the activity level such as walking, grasping an item, and
carrying back to the order picking station. Different pickers will be modeled each with different height, weight,
gender, and age resulting in a different body orientation with respect to the shelf caddy. Two pickers are modeled in
this paper: Jack (male) and Eva (female). Siemens Jack provides both the motion simulation but also the analysis of
human performance based on the simulation through its Task Analysis Toolkit (TAT). The performance results
provide insight into human performance of an order picker.

The research seeks to answer the effects on a human from picking thousands of items in an eight hour shift. The
study also addresses the impact of size and gender of the worker on order picking performance. The simulation and
task analysis helps identify which motions (e.g., grabbing from bins located in different positions and heights) may
be problematic so that more focused analysis can be done on human subjects through interviews and mock order
picking scenarios using body sensors in future studies. The results may potentially identify some repetitive motion
risks that may require changes in the layout.

4. Methodology
The simulation of a goods-to-man order picking system will be modeled using Siemens Jack software. Jack is the
stand alone offering from Siemens PLM Software for human factors and ergonomic analysis. It has several add-on
modules. such as the 3D Body Scan that can be used to create humans using existing body scans and the Task
Analysis Toolkit (TAT) which is used in the manufacturing communities to design better workplaces and maximize
the safety of workers.

Modeling the ergonomics goods-to-man order picking will be done by following 4 main steps: (1) building the work
environment, (2) creating different virtual pickers and place them near the workstation, (3) assigning each one of
them to different picking order, and finally (4) analyzing their performance after they complete their tasks. The
simulation includes various scenarios applicable to picking from the robot-delivered storage bins including items
located at all bin levels. As well as, the different four sides of the bin which requires different body postures (sit,
squat, reach, bend over, and climb). The highest bins are over two meters high; to reach them, short workers need to
use a ladder while tall workers must crouch down. Some items are very light and others can be several kilograms.
All of these conditions are included in the simulation.

In Jack’s Task Simulation Builder (TSB), we can consider the impact of task demands that are performed over an
entire work shift (8~12 hours). Cumulative low back loading has been documented by numerous research groups to
avoid the risk of injury. In particular, the Ontario Universities Back Pain Study (OUBPS) Group, in cooperation
with General Motors Canada, found a relationship between cumulative loading exposure and the likelihood of a
worker reporting low back pain.

The study follows this sequence beginning with collecting data from the simulated environment (steps 1 and 2).
1- Perform 10 picks each last for 10-15 second.
2- Perform one hour run for the picks repeated in sequence with number of items picked equal 360 item.
3- Apply the 2 picks on both Jack and Eva separately.
4- Analyze the performance of both Jack and Eva.
5- Find possible solutions to address risk areas (e.g: change item location, picking posture, or use ladder)
6- Analyze the performance of both Jack and Eva after applying the solutions to constrains.
7- Compare the two strategies results.
8- Come-up with a conclusion and suggestions based on the results.

We can monitor the picker performance by using the TAT provided in Jack to see if there is a possibility that the
picker can be harmed during the duration of the shift. TAT has a lot of tools that can provide us with variable
information regarding the picker performance such as Lower Back Analysis (LBA) tools, National Institute for
Occupational Safety and Health Lifting Equation (NIOSH), Metabolic energy expenditure (MEE), fatigue recovery,
Ovako working posture analysis tool, Rapid upper limb assessment (RULA) and Manual handling limits. Only
LBA, MEE and static strength prediction reported here.
Al Shehhi, King, Amer, Mohammad

5. Analysis
The analysis is based upon a simulated order picking environment typical of a goods-to-man order picking station.

5.1 Simulated Environment


A manual simulation of a goods-to-man order picking station was created as shown below. A typical shelf caddy
was modeled to match the dimensions reported in the literature; roughly two meters high, one meter deep, and a
little over one meter wide [9]. Bins are placed on all four sides and a robot (i.e., Kiva) moves the shelf caddy from a
storage location to the order picking station at the required time. The order picker has a console that shows the order
to be picked and bins nearby to place the picked items of an order. The picker places the order bin on the conveyor
once all the items have been picked from the shelf caddy. Typically items are picked from more than one shelf
caddy. Except for common items like a conveyor or table, each object had to be acquired from published databases.
The shelf caddy has modeling complexity because the shelving unit, bins, and item in the bins all occupy the same
physical space.

Figure 1: Goods-to-man order picking station

5.2 Anthropometric Models


Jack and Eva, the names used by Siemens Jack in Asia, represent a virtual male and female human based on the
Asian population (ASIAN_INDIAN_NID97) database. The profile and anthropometric measurements of Jack and
Eva are found in Figure 2 which represent 95 th percentile measurements.

Figure 2: Jack and Eva measurements


Al Shehhi, King, Amer, Mohammad
A male and female are compared to not only reflect gender, but also to study height effect, as a female needs to use a
ladder for the upper bins.

5.3 Order Picking Strategy


The order picking strategy was implemented with the simulation builder which requires identifying each movement
(e.g., turn, walk, locate, grasp, walk back). The simulation uses the following picking and break pattern: Moreover,
based on comments from Amazon workers on indeed.com regarding shifts and breaks; we can estimate the working
patterns that order pickers are following while working. For example; for a 12 hour shift there are three main breaks:
30 minutes for lunch break and another two 15 minutes breaks. So; in total one hour of break time in a 12 hour
working shift. A worker picks approximately 360 items in one hour (an item every 10 seconds).

A picking strategy was developed to be representative of the random stocking strategy employed with the shelf
caddies. When new items arrive to the warehouse, they are stowed in any available free space in a bin since the bin
and caddy identifiers are scanned. This means that the items in an order will be randomly located on any side of the
shelf caddy. In addition, the shape and weight of an item will also be randomly chosen. Jack software requires
specific items to be picked from a designated location. Therefore, randomization was done by randomly assigning
an object by bin height and one of the four sides of the shelf caddy.

The following are the picking strategies that Jack and Eva will follow during the simulation.
Pick1: pick 1 low leveled (row 2) item w/o the basket (Sphere: 0.63 kg).
Pick2: pick 1 high leveled (row 5) item w/o the basket (Cube: 0.6 kg).
Pick3: pick 1 heavy medium leveled item (row 4) item w/o the basket (Dictionary: 2.25 kg).
Pick4: pick 1 low leveled (row 1 plane toy:1.154 kg) item and 1 high level item (row 5 Car toy:1.21 kg) with the
basket.
Pick5: pick 1 high leveled (row 5 Mug: 0.243 kg) item and 1 low level item (Row 1 pyramid: 0.58 kg) with the
basket.
Pick6: pick 1 high leveled (row 4) item w/o the basket (Adidas Shoe: 1.36078 kg).
Pick7: pick 2 light weighed medium leveled item (row 3 and 4 in two difference shelfs) w/o the basket (Mascara
stand: 0.039 kg and leather handbag: 0.9 kg).
Pick8: pick 2 light low leveled item (row 2 in two difference shelfs) item w/o the basket (iPhone: 0.172 kg &
headphone: 0.25 kg).
Pick9: pick 2 medium and high leveled items (row 3 and 5 from the same shelf) with the basket (Kettle: 0.9071 kg &
Guitar: 1.246 kg).
Pick10: pick 3 median leveled item (1 in row 3 and 2 in row 4 from the same shelf) item with the basket (safety
glasses: 0.093 kg, ratchet: 0.082 kg and screwdriver: 0.0739 kg).

Some of these picks are shown in Figure 3. The left frame shows Jack picking an item from a chest-high bin. The
right frame shows Eva picking from a bin that would be over her head so a ladder is required. All of these motions
must be pre-determined and built in the simulator. Since the picks represent randomly located items; the repetition
for an hour occurs by repeating the simulation of those 10 picks. The repetition feature is supported by the TSB,
where each of those 10 picks will be repeated for 36 times to complete the total of 360 picks in a 1 hour run.

Figure 3: Jack and Eva picking items


Al Shehhi, King, Amer, Mohammad

5.4 Analysis from Jack TAT


Three kinds of analysis are reported in this paper: Lower Back Analysis, Metabolic Energy Expenditure, and Static
Strength Prediction based on cumulative loading. Cumulative loading can be thought of as the accumulation of all
loads, both high and low, that are experienced over the course of a day, which consider the impact of task demands
that are performed over an entire work shift. Results calculated using this tool are compared against published
threshold limits to classify injury likelihood. The method used to calculate the Cumulative Exposure follow the
protocols used in the original study [10] and limits are based on analysis of data as presented by Frazer [11].

Lower Back Analysis (LBA) is an option that allows us to calculate the Threshold limit values for Cumulative
Compression and Cumulative Flexion/Extension Moment which are 22.5 and 0.6 MegaNewton-seconds (MN)
respectively. Those numbers were based on an analysis of the epidemiological data from Frazer study [11]. These
values are set as the default, and it is assumed that the shift duration represents a full work day (8~12 hours). The
data represents a single hour of the shift with the results reported in Table 1. The data shows that Eva approaches the
LBA threshold of 22.5 and 0.6 MN based on the weights of the objects picked. Jack has a greater margin remaining
than Eva. Both Jack and Eva get a good workout.

The Metabolic Energy Expenditure (MEE) tool considers the task demands as well as a person’s characteristics
(gender, anthropometry) to predict metabolic requirements for a job or activity. MEE is based on experiments run at
the University of Michigan to measure the energy costs associated with industrial tasks [6]. Prediction equations
from this study are in widespread use by practitioners to evaluate physiological task demands.
The integration of these equations into the Task Simulation Builder (TSB) offers the ability to proactively and
efficiently evaluate metabolic demands of job concepts early in the design phase. The tool automatically estimates
the number of kilocalories expended for each action performed within the simulation, taking into consideration the
posture and exertion demands. Overall energy expenditure rate (kcal/min) derived from the energy cost estimates are
compared against industry standard, or user specified, threshold limits to identify if a job may be associated with
increased risk of fatigue.

Table 1: Siemens Jack performance data (% of threshold) for 8-hour shift


Threshold Jack Eva
Lower Cumulative Compression (MN) 22.5 17.46 (77.6%) 18.93 (84%)
Back
Cumulative Moment (MN) 0.60 0.37 (61.67%) 0.42 (70%)
Analysis
Metabolic Energy Expenditure (kcal/min) 4.1 3.72 (90.73%) 3.91 (95.37%)
Energy Total Metabolic Cost (kcal) -- 1103.31 1247.43

Static Strength Prediction provides additional data about mobility (stand, walk, sit), joint Angles (Neck, back, wrist,
shoulders, elbow) and how much time spent on this posture and the category of the thresholds (mild, moderate or
significant). The analysis shows significant flexion around Jack’s neck and back. For Eva, significant flexion occurs
around the neck and shoulders. Due to space limitations, results for Eva are excluded and available upon request.
Results for Jack can be seen in Figure 4.

Figure 4: Jack’s static strength prediction


Al Shehhi, King, Amer, Mohammad

5. Discussion
A goods-to-man system reduces the need for planning the location of goods typically done in a man-to-goods
system. A goods-to-man system, which minimizes warehouse familiarization and the need to walk great distances,
has human factors implications. The possibility of repetitive motion injuries from picking thousands of items a shift
using a goods-to-man order picking strategy must be considered which prompted this research.

We expect to see variations in risk based on anthropometric dimensions. Jack, a 95th percentile Asian male, does not
approach the thresholds for Lower Back Loading and Metabolic Energy Expenditures. However, significant joint
concerns around the neck (flexion, extension, and rotation) and back (flexion) were evident. Eva showed similar
results for Lower Back Loading and Metabolic Energy Expenditures but more joint concerns presumably due to
reaching into the lower-leveled items. Her simulation doesn’t account for the ladder which will be done later.

This research required manually modeling the order picking strategy task by task with Jack. While time-consuming,
the simulated motion may not be completely representative of the way humans pick orders (e.g., picking more than
one item at a time. The fidelity of the model is limited by the Jack-driven motions (e.g., pick a task) and the need to
extrapolate the motions of picking a small number of objects to the entire shift. We suggest using inertial motion
capture, such as by Synertial, to build a Jack simulation using real subjects and their real motions in picking objects.
By doing so, the fidelity of the simulation increases especially of the flexion around joints. The simulation could be
even more accurate in hard to simulate tasks such as grapping the items and going up and down ladders. Synertial
simulation still requires manually placing the objects into the skeleton’s hands so TAT can be performed. Either
approach has limits to the analysis because only a few minutes of order picking can be realistically captured.

This research may be the first to apply anthropometric modeling to a new approach to order picking. The
preliminary analysis shows that a goods-to-man system does increase the loads and stresses on the order pickers.
While the threshold for LBA was not reached, the posture data needs to be more carefully examined.

Acknowledgements
The UAE Ministry of Education in partnership with Siemens LLC UAE provided a software grant for the
Tecnomatix software used in this research.

References
1. Correll, N., et al., Analysis and Observations From the First Amazon Picking Challenge. IEEE
Transactions on Automation Science and Engineering, 2018. 15(1): p. 172-188.
2. de Koster, R., T. Le-Duc, and K.J. Roodbergen, Design and control of warehouse order picking: A
literature review. European Journal of Operational Research, 2007. 182(2): p. 481-501.
3. Dallari, F., G. Marchet, and M. Melacini, Design of order picking system. The International Journal of
Advanced Manufacturing Technology, 2009. 42(1): p. 1-12.
4. Napolitano, M., 2012 warehouse/DC operations survey: mixed signals. Logistics Management, 2012.
51(11): p. 54-63.
5. Grosse, E.H., et al., Incorporating human factors in order picking planning models: framework and
research opportunities. International Journal of Production Research, 2015. 53(3): p. 695-717.
6. Garg, A., D.B. Chaffin, and G.D. Herrin, Prediction of metabolic rates for manual materials handling jobs.
American Industrial Hygiene Association Journal, 1978. 39(8): p. 661-674.
7. Battini, D., et al., Human energy expenditure in order picking storage assignment: A bi-objective method.
Computers & Industrial Engineering, 2016. 94: p. 147-157.
8. Mital, A., R.R. Bishu, and S.G. Manjunath, Review and evaluation of techniques for determining fatigue
allowances. International Journal of Industrial Ergonomics, 1991. 8(2): p. 165-178.
9. M. Wulfraat, “Is Kiva Systems a Good Fit for Your Distribution Center An Unbiased Distribution
Consultant Evaluation,” MWPVL. [Online]. Available: http://www.mwpvl.com/html/kiva_systems.html.
[Accessed: 06-Apr-2019].
10. Norman, R., et al., A comparison of peak vs cumulative physical work exposure risk factors for the
reporting of low back pain in the automotive industry. Clinical Biomechanics, 1998. 13(8): p. 561-573.
11. Frazer, M. Job Rotation as a Strategy to Reduce Risk of Musculoskeletal Injuries – Does it Really Work? .
in Association of Canadian Ergonomists 39th Annual Conference. 2008. Calgary, Alberta.

View publication stats

You might also like