Fast Deployable Autonomous Systems for Order Picking

How Small and Medium Size Enterprises Can Benefit from the Automation of the

Picking Process

Marco Bonini, Augusto Urru and Wolfgang Echelmeyer

Forschungszentrum Logistik, Reutlingen University, Alteburgstrasse 150, Reutlingen, Germany

Keywords: Logistics, Automation, Robot-to-Goods, Economic Efficiency, Picking.

Abstract: The paper focuses on a recently introduced paradigm for the logistic process of picking, with respect to the

man-to-goods and goods-to-man concept: the robot-to-goods. First the task and system architecture of the

fast deployable autonomous commissioning system are described, then the economic efficiency of the

system is analysed in a real business case scenario using a simplified method, which is explained and

discussed. The clearly positive Net Present Value of the investment and the short Payback Period obtained

in the business case prove how the robot-to-goods paradigm for the commissioning process, implemented

through the automation of the forklift platform, is economically attractive for small and medium size

enterprises.

1 INTRODUCTION

The main logistic processes in which automation

through robotic still has a great potential are three:

(1) loading and unloading of unit loads, (2)

palletizing and de-palletizing and (3) commissioning

(or picking). Flexibility is the first challenge,

because far too often application scenarios in

logistics are unconstrained and dynamic. High

performances and reliability are also required in

order to make investments sustainable. Technologies

for each of the three aforementioned logistic

processes are available, but either they are partially

incomplete or they don’t properly fulfil each of the

three criteria: flexibility, performance and cost

(Bonini et al., 2015). For the (1) container loading

and unloading process (Bortfeld and Wäscher, 2012)

the only fully autonomous industrial solution

available is the Parcel Robot (robotics logistics

solutions). Other available technologies like the

Automatic Truck Unloader (Wynright), the RobLog

Industrial Demonstrator (RobLog Consortium), and

the PIQR1 (TEUN) are either prototypes or only at

an early stage of the development process.

For the (2) palletizing and de-palletizing process

(Bischoff et al., 1995) the complexity level of the

problem increases with the heterogeneity and lack of

standardization of the items. The more the items are

heterogeneous the more complex is the algorithm

that calculates the pattern of each layer of the pallet

(Wäscher et al., 2007). Not only the pallet needs to

be stable, but also efficiently loaded (Terno et al.,

1997), minimizing the free space between different

items and different pallet layers (Kocjan and

Holmström, 2008). Industrial solutions for sensors

and actuators aiming at overcoming the needs of

flexibility are available and mainly focused on the

recognition and picking of non-identical items

(RSW, Kuka Robotics, Qubiqa, Grenzebach

Automation and Fimec Technologies).

For the process of (3) commissioning the answer

of automation has been the transformation of man-

to-goods systems in goods-to-man systems

(Hamberg and Verriet, 2012). The most popular

systems available are developed by Kiva, Autostore

and Magazino. The first two follow the goods-to-

man paradigm, the last one, still at conceptual and

prototype stage, the robot-to-goods process. All of

these systems however require high standardization

of items and research is currently focusing on

solving the problem of flexibility. Recently

demonstrators from the best universities in the world

were tested within the Amazon picking challenge

held at the ICRA 2015 (Amazon) under the

following constraints: the shelves were prototypical

pods from Kiva Systems, and the robot (picker) had

to be fully autonomous. Therefore the scenario was

Bonini, M., Urru, A. and Echelmeyer, W.

Fast Deployable Autonomous Systems for Order Picking - How Small and Medium Size Enterprises Can Beneﬁt from the Automation of the Picking Process.

DOI: 10.5220/0005997804790484

In Proceedings of the 13th International Conference on Informatics in Control, Automation and Robotics (ICINCO 2016) - Volume 2, pages 479-484

ISBN: 978-989-758-198-4

479

still not industrial, but sort of simplified. Besides the

easy task of handling cuboids goods, some other

tasks were included in the challenge: grasping of

easy to damage goods (Oreo cookies, soft cover

books) and objects hard to perceive with traditional

vision systems and algorithm (Wurman, 2015). The

transformation of a man-to-good system in goods-to-

man is however not always possible and even when

it is, it comes with large impact on the existing

layout and with considerable investments.

Small and Medium size Enterprises (SMEs)

seeking cost effectiveness of the commissioning

process through automation are mostly based on the

man-to-good paradigm; a complete re-design of both

process and layout would result in a slow return of

investment, increasing the risk.

This paper is focussed on the recently introduced

robot-to-goods paradigm (Huanga, 2015) with

respect to the traditional man-to-goods and goods-

to-man concepts. First the task and architecture of

the fast deployable autonomous commissioning

system are described, then the economic efficiency

of the system is analysed in a real business case

scenario. As a result the automation of the forklift

platform for the commissioning process is proved to

be economically attractive for SMEs.

2 TASK AND SYSTEM

ARCHITECTURE

In the following the task of the fast deployable

autonomous system for warehouse commissioning

automation is described, the challenges are

explained and its potential architecture is outlined.

2.1 Navigation and Fast Deployment

The fast deployable autonomous system for

warehouse commissioning automation is based on

the hardware platform such as the EK-X, vertical

order picker from Still GmbH (Figure 1.a). This

forklift vehicle is currently used to navigate in high

rack warehouses enabling the safe picking of items

up to 12m high, all operations being manually

accomplished by an on-board operator (Figure 1, b).

The first challenge consists of transforming the EK-

X into an Automated Guided Vehicle (AGV), using

Simultaneous Localization and Mapping (SLaM)

methods, which are not based on localization

infrastructure (markers, reflectors etc.), while still

being capable of providing highly accurate pose

estimates in a changing dynamic environment. Being

infrastructure-independent (no markers or reflectors,

no a priori knowledge of the warehouse) these

SLaM methods are essential for the system to be

quickly deployable, reducing the barrier to the test

and to the set-up investment.

2.2 Picking and Mixed Palletizing

Advanced navigation technologies are however not

enough for accomplishing the commissioning task:

when the location of the target item has been

reached, identifying the exact position of the item,

grasping it and placing it safely onto a pallet are all

challenges that still need to be tackled. These

operations involve robots abilities such as object

recognition, motion, manipulation and decisional

autonomy.

Figure 1: overview of the starting hardware platform and

the current manual picking process.

The fast deployable autonomous system for

warehouse commissioning automation is set to

operate in a high rack warehouse where pallets with

homogeneous parcels (same shape and dimensions)

are stored: before the in-feed in the high rack stock

these pallets are often wrapped in plastic foil for

safety and security reasons. After the AGV has

navigated to the right location, (1) the manipulator

has to be positioned at the right height for picking

the desired parcel(s), (2) the plastic wrapping

surrounding the pallet (if any) and its parcels must

be cut (in case the pallet is still complete, using

principles like those implemented in the DefoCube,

by LMS Development Concept, the Robotic

Unwrapper, by BW Container System, or the MSK

Defotech, by MSK Covertech Group) and finally (3)

the right number of parcels has to be picked and

placed in the commissioning pallet on the AGV

(placed on the adjustable front forklift of the EK-X,

moving at the same height as the manipulator) in a

way that the resulting parcel staple is space-

optimized and stable. An overview of the

commissioning scenario, including a sketch of the

ICINCO 2016 - 13th International Conference on Informatics in Control, Automation and Robotics

480

hardware, is shown in Figure 2: overview of the

industrial demonstrator application scenario.,

together with the section of the mobile robot's

workspace needed in each corridor of the

warehouse.

Figure 2: overview of the industrial demonstrator

application scenario.

2.3 IT Warehouse Integration

Two additional challenges need to be tackled after

the picking process and in-between two picks

respectively: (1) the interface with the Warehouse

Management System (WMS) and (2) the

Collaborative Mission Planning and Fleet

Management. The stock in warehouses is managed

by the WMS and an interface with this module is

necessary in order to automate the task of

forwarding to the mission planner the

commissioning list and to update the stock status.

However, items in the commissioning list must be

prioritized according to specific rules, such as: pallet

stability, route optimization, traffic, and possible

synergies among AGVs in the picking strategies.

The mission planner and fleet manager module will

satisfy these requirements through compromises

between centralized and decentralized decisions.

3 EVALUATION OF ECONOMIC

IMPACT

In this section the potential economic impact of the

autonomous picking system previously introduced is

quantified and evaluated in a potential business case,

in order to give a hint of the economic benefits a

SME could have by investing in an in such a system.

The study case considers facts and figures provided

by a SME (Small Medium Enterprise), located in

Hohenstein-Ernstthal (Saxony, Germany), dealing

with transport and in-house logistics. First, data

concerning the current manual process are collected

and analysed, then possible characteristics and

performances of the future fully automated

warehouse are discussed and a comparison between

the current and future scenarios is carried out.

Finally the estimation of the differential cash flows

enables the calculation of the Net Present Value

(NPV) function which is a key factor in investment

decision-making.

3.1 Evaluation Scenario

Nowadays the company on which the analysis is

based is active 250 days per year and is operating

two (2) shifts per day. Each shift, seven (7) manually

operated vehicles similar to the EK-X, with

respective operators on board, are driven through the

shelves of the warehouse in order to compile the

orders. The number of picks per year is 500.000 on

average and each order includes averagely 16 items.

From this data the performance of the current overall

picking process can be calculated.

Considering the same application, with the

implementation of the autonomous picking system,

the following reasonable hypotheses can be

formulated:

 The on board operator is replaced by a robot

mounted on the AGV

 The fleet of 7 autonomous vehicles is able to

work two (2) shifts without any additional cost

(only some minor direct cost for additional

energy, not comparable to the additional cost of

additional workers for additional shifts).

 The automated commissioning task will be

performed in comparable time to the manual one

(maybe the picking operation will be slower than

the manual one, but the navigation time can be

improved due to collaborative picking

strategies).

 The only cost considered for the new vehicle is

arising from the additional modules needed in

order for the vehicle to be autonomous in the

navigation and in the picking, such as: vision

system, specific gripper and manipulator and

other necessary hardware modifications from the

existing forklift hardware. All these costs are in

this business case estimated for a total amount of

110.000€. In the warehouse, seven (7) EK-X or

similar would already be in use for the manual

process. Thus, no additional cost for buying

forklift equipment has been considered.

 The additional costs for service, support and

maintenance are not deemed to be relevant for a

first simplified analysis and therefore are

neglected in the following.

Fast Deployable Autonomous Systems for Order Picking - How Small and Medium Size Enterprises Can Beneﬁt from the Automation of

the Picking Process

481

 The cost of the autonomous fleet supervisor

needed for managing exceptions is evened out

with the cost of the supervisor and warehouse

manager of the current manual process, which is

no longer needed.

 The power consumption of the new autonomous

system will be higher than the one in the

currently deployed devices, but the difference is

for the first analysis considered to be negligible.

 Since the system will be rapidly deployable it is

assumed that the additional costs for the system

set-up and installation at the user's site are

negligible or included in the 110.000€.

The following summary table shows the

parameters, grouped by time, operation,

performance and cost, considering first the current,

then the future commissioning process (after the

implementation of the autonomous system).

Table 1: Parameters for the current and the future

commissioning process.

Parameter

Curren

system

Autonom

ous

system

Time

Working days per

year

250 250

Shift per day 2 2

Operati

# of employees per

shift

# of picks per year 500.000 500.000

# of item per order 16 16

Perfor

mance

Mean time fulfilling

1 order [min]

53,76 53,76

Mean time per

picking [min]

3,36 3,36

Cost

Cost of the operator

per year (gross) [€]

28.105

Cost for additional

modules on 1

vehicle [€]

110.000

3.2 Investment Analysis

Implementing the autonomous system at the year 0,

the company should invest an amount of 770.000 €,

necessary for 7 autonomous vehicles. However from

the moment the system is deployed and in action, the

cost of the operators (7 by 2 shifts, so 14) ceases to

exist, enabling constant savings for 393.470€ per

year.

The cash flows estimation allows the calculation

of the NPV, exploiting the following formula:





,







0







1 







(1)

Where:

 r is the interest rate, considered to be equal to

10% in this calculation;

 N is the service life of the autonomous system

set to 10 years

 NCFi is the net cash flow at the year i of the life

of the systems: this parameter describes the

savings that the autonomous system enables

every year in comparison to the current manual

process.

This NPV approach considers the greater risk of

the cash flows which are further in the future, thanks

to the actualization factor (denominator), penalizing

cash flows which lie further ahead (and therefore are

not certain) and considering instead the certainty of

investing in year 0 (this amount is not reduced by

the actualization factor).

Table 2: Expected cash flows of the differential savings.

Year Cash flow NPV [€]

0 -770.000 € -770.000 €

1 393.470 € -412.300 €

2 393.470 € -87.118 €

3 393.470 € 208.502 €

4 393.470 € 477.247 €

5 393.470 € 721.561 €

6 393.470 € 943.664 €

7 393.470 € 1.145.577 €

8 393.470 € 1.329.133 €

9 393.470 € 1.496.003 €

10 393.470 € 1.647.703 €

Figure 3: NPV Function.

3.3 Results

The result of the analysis highlights two points:

 The NPV at the end of the life of the system is

positive (year 10, 1.647.703€), which generally

ICINCO 2016 - 13th International Conference on Informatics in Control, Automation and Robotics

482

is the first factor for a positive decision whether

to invest;

 The NPV becomes positive in slightly more than

2 years, minimizing the risks linked to the

uncertainty (namely the Discounted Payback

Period), which is an additional factor for a

positive decision of investment.

4 DISCUSSION

In this first rough analysis some possible synergies

and factors that could increase the NPV and

decrease the Payback Time have not been

considered, such as:

 Possibility of exploiting the night shift and the

week-end without additional costs, increasing the

capacity;

 Thanks to the AGV fleet coordinator and mission

planner, collaborative planning and picking

techniques could be enabled. This directly

impacts on shorter routes for AGVs and thus

results in time savings and improvement of the

overall performance of the system. In turn this

could mean either that less vehicle are necessary

or that the capacity can be increased. This would

further reduce the payback time under 2 years;

 With the autonomous systems, the route and

picking sequence are chosen for the pallet to be

stable: this means that no additional time at the

picking place must be spent by the system for

moving parcels around in order to achieve the

pallet stability. This can be translated in time

saving with respect to the current manual

process: currently, while picking, the operator

needs to identify on the fly the most stable

position for the parcels taking into account the

overall stability and that there will be more items

on the same pallet. Even though the result of a

human operator compiling a mixed pallet will

most likely be better in stability and volume

optimization, the robot is able to pre-plan this,

which avoids re-palletization that even the most

experienced human operator needs to accomplish

in order to reach a satisfactory result;

 The depreciation tax shield, not included in this

model, will have a further positive impact on the

NPV, further pushing towards a positive decision

of investment.

In the presented business case, for the purpose of

demonstrating its utility and convenience, the fast

deployable autonomous system for order picking is

considered as a product, hence with a technology

readiness level (TRL) of 9, namely as an “actual

system proven in operational environment”

(European Commission). As a matter of fact, even if

most of its sub-systems - such as the SLaM module

or the feet management - have a high TRL (7 or

more), the system as a whole has a current TRL of 2

(“technology concept formulated”), because its sub-

systems have not yet been integrated, tested and

optimized in their potential synergies.

The future work in this regard is twofold. First

(1) step changes in TRL of the single modules need

to be achieved; in particular the manipulator and the

technology to cut the pallet’s wrapping need to be

optimized in order to be lightweight, since they are

to be in operation at 10 meters height and suitable

for the few cluttered workspace available between

the shelves of the warehouse. Then (2) work needs

to focus on the integration of each module, proofing

the effectiveness and efficiency of the whole system.

Only in this second phase it will be possible to

assess open points concerning, for instance, the

overall system positioning accuracy or the autonomy

and efficiency of its power supply.

5 CONCLUSIONS

This paper shows how the robot-to-goods paradigm,

implemented thanks to a fast deployable

autonomous commissioning system, can enable

savings for small and medium size logistic

enterprises. First the task and system architecture

have been described, then the economic efficiency

of fast deployable autonomous commissioning

systems has been analysed in a real business case

scenario. The simplified method used for the

business case analysis has been explained and

discussed. The clearly positive Net Present Value of

the investment and the short Payback Period, proved

how the automation of the forklift platform for the

commissioning process is economically attractive

for SMEs.

REFERENCES

Amazon, picking challenge, https://www.youtube.com/w

atch?v=UrpMfdj-Mpc.

Autostore https://www.youtube.com/watch?v=iyVDMp2b

L9c.

Bischoff, E., Janetz, F., and Ratcliff, M., 1995. Loading

pallets with nonidentical items, European Journal of

Operational Research 84.

Bonini, M., Prenesti, D., Urru, A. and Echelmeyer, W.,

2015. Towards the full automation of distribution

Fast Deployable Autonomous Systems for Order Picking - How Small and Medium Size Enterprises Can Beneﬁt from the Automation of

the Picking Process

483

centers. In proceedings of 4th International

Conference on Advanced Logistics and Transport

(ICALT), pp.47-52, 20-22 May 2015 doi:

10.1109/ICAdLT.2015.7136589.

Bortfeld, A.; Wäscher, G., 2012. Container Loading

Problems - A state of the art overview, Otto von

Guericke University Magdeburg, ISSN 1615-4274.

BW Container Systems – Robotic Unwrapper.

http://www.bwcontainersystems.com/docs/default-sou

rce/fleetwoodliterature/robotic-unwrapper-unbagger-e

nglish.pdf?sfvrsn=12.

European Commission. href=http://ec.europa.eu/resea

rch/participants/data/ref/h2020/wp/2014_2015/annexe

s/h2020-wp1415-annex-g-trl_en.pdf.

Fimec Technologies, Packtris, http://www.youtube.com/

watch?v=KzopQMvbLMk.

Grenzebach Automation, MXRP, http://www.youtube.

com/watch?v=pcH3zy8ib7M.

Hamberg, R., Verriet, J., 2012, Automation in Warehouse

Development, Springer London, print ISBN 978-0-

85729-967-3, online ISBN 978-0-85729-968-0, DOI

10.1007/978-0-85729-968-0_2.

Huanga G. Q., Chenb M. Z. Q., Panc J., 2015 Robotics in

ecommerce logistics, The Hong Kong Institution of

Engineers HKIE Transactions, vol. 22, issue 2, pp. 68-

77 DOI:10.1080/1023697X.2015.104396.

Kiva. https://www.youtube.com/watch?v=3UxZDJ1HiPE.

Kocjan, W., Holmström, K., 2008, Generating stable

loading patterns for pallet loading problems, The Fifth

International Conference on Integration of AI and OR

Techniques in Constraint Programming for

Combinatorial Optimization Problems CPAIOR08.

Kuka Robotics, OrderAssembly Lösung zur gemischten

Palettierung, http://www.youtube.com/watch?v=zusk5

KtVUYs.

LMS Development Concept - DefoCube. http://lms-devel

opment-concept.de/produkte/defocube.html.

Magazino. http://www.magazino.eu/toru/?lang=en.

MSK Covertech Group - MSK Defotech. http://www

.msk.de/branchen/getraenke/defolieren.html.

Qubiqa, layer picker, http://www.qubiqa.com/default.asp?

Action=Details&Item=561.

RobLog Consortium, Industrial Demonstrator,

https://www.youtube.com/watch?v=XmeuwFp3eF.

Robotics Logistics Solutions, Parcel Robot,

https://www.youtube.com/watch?v=MqhRMudPXLE.

RSW mixed case pallettitzing, http://www.youtube.com/

watch?v=rWT7JJbfhvo.

Terno, J., Scheithauer, G., Sommerwei, U. and Riehme J.,

1997, An efficient approach for the multi-pallet

loading problem, European Journal of Operational

Research, vol. 123, pp. 372–381.

TEUN, PIQR1, https://www.youtube.com/watch?v=A5n-

sTrMtO8.

Wäscher, G.; Haußner, H. and Scurmann, H., 2007. An

improved typology of cutting and packing problems.

European Journal of operational research ,183, 1109-

1130.

Wurman, P., Romano, J., 2015, The Amazon Picking

Challenge 2015 [Competitions], in

Robotics &

Automation Magazine, IEEE , vol.22, no.3, pp.10-12,

doi: 10.1109/MRA.2015.2452071.

Wynright, Automatic Truck Unloader, https://www.you

tube.com/watch?v=WnxudJbeJW4.

ICINCO 2016 - 13th International Conference on Informatics in Control, Automation and Robotics

484