robotics of the future in the pharmaceutical industry

Robotics of the future in thepharmaceutical industryDr. Carl-Helmut Coulon l INVITE GmbH, Leverkusen

Correspondence: Dr. Carl-Helmut Coulon, Head of Future ManufacturingConcepts, INVITE GmbH, CHEMPARK, Geb. W32, 51368 Leverkusen;e-mail: [email protected]

Abstract

This article differentiates applications of "modern robotics" from classical automation inorder to illustrate that technical advances in recent years have made a new type ofautomation possible. It discusses the drivers and applications within the pharmaceuticalindustry which will benefit from this modern robotics and gives an indication of theindustry-specific requirements. Section 5 outlines how machine learning will simplify andspeed-up developments in future. Finally, the article envisions major opportunities for thepharmaceutical and robotics industries justified by the availability of the described newtype of applications in combination with the outlined demands.

1. Introduction

Requirements regarding quality, er-gonomics and efficiency reveal thenecessity to a much higher degree ofautomation for the production ofpharmaceuticals. In addition, forsafety reasons the industry is askedto technically separate humans andhigh potent substances whereverpossible [1, 2]. However, looking at

production, many activities are cur-rently still carried out by humans,even though there are strong rea-sons to avoid it.

The main message of this articleis the hypothesis that modern ro-botics will provide the pharmaceu-tical industry with cost-effectiveand easy-to-use flexible applica-tions as soon as the pharmaceuti-cal industry specifies these applica-tions and their requirements.Phrased the other way round:There will be a relevant market forrobots in the pharmaceutical in-dustry in future.

2. Modern Robotics: Newabilities by newtechnologies

What distinguishes modern roboticsfrom traditional automation? Thissection names the differences andwill show experiments and previousimplementations as examples. Theaim is to prove that modern roboticscan often provide solutions that aresuitable for the challenges men-tioned above in terms of flexibilityand cost-effectiveness.

2.1 New technologies enableflexibility and cost-effectivenessIndustrial robots are used world-wide in large numbers at corre-spondingly attractive conditions,especially regarding the ratio ofcosts to functionality. If flexibilityand cost efficiency at low through-put are important, such a "univer-sal machine" is superior to a spe-cialized machine. Figure 1 showshow developments in recent yearshave contributed to the necessaryfeatures which support cost-effec-tiveness.

Today's development environ-ments simplify programming signifi-cantly. In particular, the increasinglysimple integration of sensors andgrippers in integrated developmentenvironments speeds up develop-ment considerably. This means thatsimple applications such as remov-ing objects from a conveyor belt inpackaging can be taught to the ro-bot in minutes.

In contrast, modern sensor tech-nology such as image recognitionand touch sensors is not simple perse. Anyhow these technologies are

Key Words

• Robotics

• Pharmaceutical industry

• Automation

• Production

• Laboratory

Author

Dr. Carl-Helmut CoulonSince 2015 Dr. Carl-Helmut Coulon heads the“Future Manufacturing Concepts”group at theResearch Institute INVITE including the subarea“application-oriented robotics research”. After hisPhD in artificial intelligence in 1997 he started atBayer AG in the automation department followedby various functions around the production tech-nology up to his switch to INVITE 2015.

TechnoPharm 10, Nr. 3, 150–157 (2020)© ECV • Editio Cantor Verlag, Aulendorf (Germany) Coulon l Robotics of the future 1

now available and integrable in highquality and variety for the engineer.Recognizing an object with an accu-racy of a tenth of a mm or monitor-ing the correctness of the performedaction via the tactile sensor systemcan now be integrated into roboticsapplications more easily than be-fore. This makes it increasingly ea-sier to develop highly flexible andcomplex applications like the onedescribed in the next section. Thenecessity to apply artificial intelli-gence and its impact to flexibilityand cost-effectiveness will be ex-plained in section 5.

In addition, new safety conceptscan reduce the space requirementsof solutions and integrate them bet-ter into the workflow if, for example,they can do without a separatingsafety fence by utilization of safety-sensors ( fig. 2).

2.2 Modern robotics as analternative to classicautomation based on processadaptationIn classical automation, it is com-mon practice to standardize theworld so that it is easier to auto-mate. However, if the existing stan-dard cannot be handled by classicalautomation or if it would be verycostly to adapt existing processes toautomation due to regulatory re-quirements restricting changeswithin the process, no automatedsolutions have been developed sofar.

A relevant example is the hand-ling of solid substances in contain-ers of all kinds in the incominggoods department of companies( fig. 3). Inliners in drums are theworldwide standard, especially foractive substances. Millions of them

are produced annually in the chemi-cal, pharmaceutical and food indus-tries. Since there is no automationavailable to date that can handle thevariety of hundreds of differenttypes of these containers in the in-coming goods department of thesites, the vast majority of containersof this type are now handled by hu-mans.

Obviously, the processes wouldbe much easier to automate at thispoint and also more secure if a newstandard of containers could be in-troduced which would not only beeasier to handle automatically butwould also allow a closed handlingof the substances so that contactwith the environment is avoidedduring handling.

Despite these strong drivers forstandardization it is not to be ex-pected that the task will become ea-

Figure 1: Overview on new abilities supporting cost-effective applications enabled by new technologies (source of all figures: INVITE GmbH).

Figure 2: End-of-line-Palletizing without safety fence.

2 Coulon l Robotics of the futureTechnoPharm 10, Nr. 3, 150–157 (2020)

© ECV • Editio Cantor Verlag, Aulendorf (Germany)

sier for automation in the near fu-ture either.

For these reasons, there was anopen innovation challenge in 2016/2017 that asked to develop proto-types of an automated solution forthe following process:• Open drum• Open inner bag/inliner• Take sample and place into sam-ple holder

• Close inner bag/inliner• Close drum and attach new seal.Further details of the challenge aredescribed within [3].

This challenge has shown thatmodern robotics offers new flexibleapproaches to solve a complex tasklike this. 2 characteristics of the pro-totypes should be described in de-tail:• Image recognition: The state ofthe art has evolved in the last fewyears in such a way that todaythere are not only many systemson the market that can recognize3-dimensional point clouds withsufficient accuracy, but the costsof these systems have decreasedto such an extent that it does notsignificantly increase the systemcosts to use several cameras. As aresult, clearly defined objects suchas the barrel, the closure or eventhe cable tie can be detected with-out any problems, even if the col-ors of the objects vary. After all,even a white cable tie on a whitebag must and can only be recog-nized by its shape.

• Tactile sensors: There is hardly asimpler, more reliable andmoreprecise way of determining a posi-tion than by touch. Meanwhile,force feedback sensors are builtinto many robot arms and are al-ternatively available as standard

modules for the wrist of the robot.If, for example, the robot arm (sup-ported by force feedback sensors)drives against the edge or closureof the barrel or hooks under thecable tie with a tool, the situationis perfectly detected to reliably en-able the desired robot action.

Finally, the combination of image re-cognition and tactile sense in a "hu-man-like" way is very effective. Forexample, the camera can detect theposition of the cable tie and theforce feedback sensor can then veri-fy the position during the openingprocess and monitor the success ofthe action.

Based on the results of the com-petition and further research, an in-dustrial solution for this task is cur-rently being developed.

The insight from this example isthat even without adapting the pro-cesses, modern robotics can make itpossible to automate processes thatwere previously seen as not automat-able. The next section will line out thedrivers behind applications like thiswith focus on different areas withinthe pharmaceutical production.

3. Drivers and futureapplications of modernrobotics

Some tasks have a high attractive-ness for automation but cannot beautomated in an economically vi-able way by "classical" automationbecause it is not about high-fre-quency and big business cases. Theattractiveness is given by other driv-ers ( fig. 4), such as• separation of product and human,• work ergonomics,• automatic "perfect"documenta-tion,

• high repetitive accuracy or• 24/7 availability of the sub-pro-cess.Efficiency and thus increased

profitability are of course goals aswell, but rather as a consequence ofreaching the mentioned goals andas boundary conditions for any in-vestment.

Even though the drivers are ob-viously relevant as drivers for auto-mation operational units have diffi-culties to identify the correspondingopportunities. In order to revealthese potentials, an assessment forthe individual companies and orga-nizational units is helpful:• Ergonomics: Where is the highestsickness-rate?

• Monotony: Job "hated most"…• Safety/Separation of Product & Hu-man: Where is the highest need toprotect human and/or product?

• Efficiency: Which is the single pro-cess step with the highest demandof manual work?

• Quality: Which is the most "trick-iest" manual process causing var-iances in the quality?Where dohuman errors cause deviations?

• Documentation: In which processis the documentation effort forthe manual work the highest incomparison to the work itself?Where are higher GMP require-ments on data integrity?

• Missing 24/7: Which functioncauses the highest "impact"due tomissing 24/7 availability?

Following these questions, the nextsections name drivers and concreteapplications for different segmentsof the pharmaceutical industry.Some of these applications are al-ready the subject of current indivi-dual developments, others are theresult of the first meeting of the In-

Figure 3: Variety of incoming containers with solids.


ternational Society for Pharmaceuti-cal Engineering (ISPE) Special Inter-est Group Robotics (DACH) [4]. ThisSpecial Interest Group (SIG) willcontinue to detail and prioritize themost attractive applications. The fi-nal goal is to get access to these ap-plications at best conditions (withrespect to implementation effortand costs compare section 5).

3.1 ProductionChangeover, cleaning and prepara-tion of material are physical activ-ities for people working in produc-tion in the process industry. Theseactivities are characterized by highdemands on flexibility, documenta-tion and/or the ability to mastercomplexity. Automation that goesbeyond the previous level must beparticularly flexible here.

Examples of such applicationsare:• Depalletizing and re-palletizing(e.g. order picking or from woodento aluminum pallets)

• Opening of containers for solids,sampling of solids for Quality Con-trol (QC) and dosing of solids set-up and cleaning of standardizedproduction equipment such astableting turrets or format parts/handling parts

Within the first session of the ISPESIG Robotics the participating com-

panies added many demands forautomation within production witha special focus to the productioncore process, transportation andcleaning (table 1).

3.2 PackagingFor large-volume standard packag-ing, classic packaging systems areestablished. They• have a high performance,• require a lot of space, as each "op-eration" is carried out in a differ-ent part of the line,

• are profitable for large quantities,since they consist of many compo-nents which leads to high costs inconstruction and system integra-tion, and

• have to be extensively changedover for handling different prod-ucts and are therefore unsuitableif the products change frequently.

However, as soon as the number ofpieces is low and/or the variants arefrequent, one sees people perform-ing monotone and ergonomicallyquestionable tasks. In contrast totraditional automation, humans aremuch slower, but much more flex-ible – and they require much lessspace.

Examples of such applicationsare: (all for small batches and differ-ent products)• Palletization at packaging lines

• Simple pick-and-place applica-tions, e.g. from the belt into thebox

• Complex packaging, e.g. multipledifferent components in one pack-age

• Closing and labelling the cartonTable 2 lists further demands dis-cussed within the SIG Robotics.

3.3 Quality Control/Research& development laboratoriesIn QC, many people perform routinetasks – if QC cannot be integratedinto production inline. Complexanalyses are often carried out byanalytical equipment, but the com-plete material flow and the partiallynecessary sample preparation isdone by humans.

Due to the variety of tasks andnecessary steps, traditional labora-tory automation is only suitable ifits use is worthwhile due to highthroughput and low sample andprocess variance, e.g. preparation ofmicrotiter plates.

Examples of such applicationsunmatched by classical automationare:• Sampling from containers (drums,bags, barrels, buckets, Big Bags)

• Sample preparation in the micro-biological laboratory

• Separation and dosing of samplesfor the different analysis needs

Figure 4: Overview of drivers of modern robotics in the pharmaceutical industry (HSE: Health, Safety and Environment targets).



• Transport of samples betweenspecialized analytical instruments

Table 3 lists many more processesnamed within the SIG Roboticswhich are executed by humans to-day and are meant to be attractivefor automation.

4. Industry-specificrequirements

Robotics applications must fulfillsimilar requirements in differentindustries – mainly regardingmaintainability, usability, reliabilityand safety [1, 2]. By targeting ap-plications which are applied not

only in one plant but within thewhole industry, the companieswithin the pharmaceutical industrymust align their requirements. To-day it is quite common that e.g.the safety regulations are applieddifferently by different companiesleading to unnecessary develop-ment and costs.

Within the pharmaceutical in-dustry there are many require-ments on top. The most specificrequirements are the necessity toavoid cross contamination and therequired ability to qualify an appli-cation with respect to regulations.Table 4 lists concrete features dis-cussed within the SIG Robotics

workshop and related to these2 major requirements.

5. Machine learning willfurther increase flexibilityand cost-effectiveness

Looking at the technical advancesin automation it is obvious thatthe pharmaceutical industry isnot anyhow close to the limit.Looking at the sample of incom-ing goods it is also obvious that itstill costs an enormous effort tohandle the variety of real life byautomation.

The higher the variance of pro-cesses and objects to be handled inan application, the higher the devel-opment effort increases accordingly.

In the coming years, the use ofmachine learning will reduce thisdevelopment effort. Today, ma-chine learning is already used in-tensively in image recognition anddata analysis. However, the "ma-nipulation" of an object to behandled is usually engineered "byhumans". Even though it is alreadypossible to teach a robot relativelyeasily by demonstration, this israrely used for development of in-dustrial applications. Inserting an

Table 1

Selection of needed applications for production named within the first meetingof SIG Robotics.

Process Transport Cleaning

Homogenizing/Mixing Package handling between processsteps

Format parts, punches in tableting,complex shaped components, alsowash of organic substances

Filling Handling in interim storage: stacking,sorting

Cleaning equipment within cam-paigns

Unpacking/dosing of excipients andactive ingredients

Trolley transport Inspect cleaning

Mixing of raw materials Change of equipment (e.g. replacing afull container with an empty one)

Room cleaning

Unpacking/dosing active ingredients

Unpack and supply consumables (sterile)

Automatic loading, docking, undocking ofcontainers

Table 2

Selection of needed applications for packaging namedwithin the first meeting of SIG Robotics.

PackagingPallet box loading/unloading with collaborating robot (FS/bundles/tray)

Gluing, filling and regluing folding boxes

Packaging of bottles and blisters for current manual packaging(very small batches -> high flexibility)

Repacking (de-bottling, de-blistering) under a protective atmosphere

Removing labels from boxes


object into another object with anaccuracy of a tenth of a mm,pouring 100 ml from a bottle or re-

acting to tactile sensors or imagerecognition is still done by pro-gramming today.

Future systems will learn bythemselves to deal with new situa-tions in a targeted manner. In 2016Google published a video where ro-bots learn from more than800 000 attempts to pick differentunknown objects out of a drawer[5]. Although learning robotics iscurrently still primarily a researchtopic, the video demonstrates theopportunities to increase the flex-ibility and cost-effectiveness of ro-botics applications in future.

6. Major opportunitiesfor the pharmaceuticaland robotics industries

By screening the current roboticsapplications at more than a dozenpharmaceutical production sites thenamed palletizing application is theonly known off-the-shelf solution.All other robotics applications with-in the pharmaceutical productionare custom-built. This results in

Table 3

Selection of attractive applications for automation in laboratories named withinthe first meeting of SIG Robotics.

Preparation and samplemanagement

Assay/Process Performance Finishing, disposal, cleaning,rennet maintenance

Sample transport and distribution Titration process (Sterile) cleaning: Wiping, suction andrinsing of robots

Buffer and solution preparation Determination of drying loss and sulfateash

Ample storage (stability sample supply,retained samples, samples formicrobiological analyses)

Distribution and provision ofconsumables, materials to the work-

places

Passenger process of a cell culture Disposal

Sample Weighing and Dilution Cutting off blood tubes Preliminary cleaning of (glass) vessels

Preparation of standards and testsolutions

Performance of thin layer chromatogra-phy

Goods receipt for samples/SampleSeparation

Bacterial Endotoxin Test (LAL)

Quality Tests: particles, tightness,level/weight/volume

Microbial sterility testing

Wet chemical analysis

High Pressure Liquid Chromatography(HPLC) preparation and measurement

Table 4

Selection of industry specific requirements namedwithin the first meeting of SIG Robotics (RRKL: cleanroom class).

Avoid (cross) contamination Capable of CSV (computationalsystems validation)

Cleanability of the robot in contactwith OEB 4/5 materials (effectiveness& solvent resistance)

Qualifiable

Surface condition (roughness requiredby cleanroom-class (RRKL))

Data Integrity/Audit Trail

Cleaning validation concept Implementation of pharmacopoeiamethods

Hygienic Design

Suitable IP class for each cleaningspecification

Limited emission of particles (accord-ing to RRKL)



high effort in engineering, develop-ment, qualification and implemen-tation.

Especially in processes that are si-milar in all companies, the flexibilityof modern robotics can enable stan-dard solutions that are not only re-warding for suppliers due to theirnumber of units but are also highlyattractive for the applying industry interms of implementation effort andcosts. What is necessary to get there?

First, the pharmaceutical indus-try must identify the applicationsand their requirements. Looking atorganizations like ISPE and othersthis topic gets more and moreawareness. It is just a matter of timeuntil the application targets arepublished.

Second, robotics companies areneeded which are engaged enoughto envision the selling potential ofsuch applications and willing to de-liver the most flexible, most easy toqualify, most easy to install andtherefore most competitive roboticsapplication.

The palletizer in fig. 2 mightnot be the most typical pharma-ceutical application, especially be-cause it can be applied aside fromregulative requirements. Anyhowit is flexible enough so serve thedifferent clients, is already sold inhundreds, is set up in minutesinstead of weeks and thereforea good example where "industry"can get if "industry" really wantsto!

Literature[1] European Council Directive 89/391/EEC

of 12th June 1989 on the introduction ofmeasures to encourage improvements inthe safety and health of workers at work

[2] Information der Deutschen GesetzlicheUnfallversicherung (DGUV) 213-083;Section 6.2

[3] Open Innovation Challenge: https://grants4tech.bayer.com/robotic-competition/; Bayer AG Corporate Technology &Manufacturing – Innovation Manage-ment

[4] https://ispe-dach.org/arbeitsgruppen/sig-robotic/; Results of first constitutingmeeting 18th Dec 2019

[5] Google Develops Robot Arms that Learnto Pick Up Objects: https://ai.googleblog.com/2016/03/deep-learning-for-robots-learning-from.html

All links were checked most recently on14th Apr 2020.


robotics of the future in the pharmaceutical industry

Documents