Robot picking: how intelligent systems are transforming intralogistics

In the ware­house of the future, humans and robots will work side by side. Autonomous vehi­cles nav­i­gate inde­pen­dent­ly through the aisles, while AI sys­tems opti­mise pick­ing orders — with­out any human inter­ven­tion. But how far away are we real­ly from this?

We are actu­al­ly already there. As indus­tri­al plan­ners, we are con­stant­ly deal­ing with far-reach­ing tech­no­log­i­cal rev­o­lu­tions in intral­o­gis­tics. The con­stant mar­ket pres­sure for greater speed, pre­ci­sion, effi­cien­cy and flex­i­bil­i­ty makes automa­tion a strate­gic neces­si­ty for our cus­tomers. The fusion of hard­ware and intel­li­gent soft­ware is redefin­ing the rules of the game in the field of robot­ic order pick­ing.

Let’s first take a look at the stages of robot-assist­ed order pick­ing.

Robot-assisted order picking

Automation levels for the use of picking robots

Diagram of automation levels of picking robots
Diagram of automation levels of picking robots

This is the first devel­op­ment stage on the way to ful­ly auto­mat­ed order pick­ing. It com­bines human flex­i­bil­i­ty with robot­ic pre­ci­sion. The human remains flex­i­ble and capa­ble of mak­ing deci­sions, while the robot acts as a “cobot” and ensures pre­ci­sion and repeat accu­ra­cy.

In this approach, humans and robots work close­ly togeth­er: The robot per­forms the repet­i­tive or ergonom­i­cal­ly stress­ful move­ments — such as grip­ping, lift­ing or posi­tion­ing items — while the human con­tributes the deci­sion-mak­ing log­ic or qual­i­ty con­trol.

Tech­nol­o­gy:

Semi-auto­mat­ed pick­ing robots are based on sim­ple image pro­cess­ing or bar­code scan­ning. The robot exe­cutes the intend­ed move­ment pre­cise­ly — either by direct autho­ri­sa­tion from the employ­ee or on the basis of pre­de­fined para­me­ters. The pick­ing robot ori­ents itself to fixed grip­ping points and pre-struc­tured objects (e.g. in box­es), but there is no adap­tive han­dling. The human remains in the con­trol loop at all times: he selects, checks and cor­rects if nec­es­sary.

What is a cobot?
A cobot is a col­lab­o­ra­tive robot for work­ing safe­ly and direct­ly with peo­ple in a shared work­space.

Advan­tages:

  • Short­er pick­ing times
  • Con­sis­tent process qual­i­ty with reduced phys­i­cal strain for staff

Field of appli­ca­tion:

  • Small to medi­um-sized ware­hous­es with stan­dard­ised prod­ucts

This inter­ac­tion offers an ide­al intro­duc­tion to automa­tion because it hard­ly changes exist­ing process­es, but sig­nif­i­cant­ly improves effi­cien­cy and ergonom­ics.

In this sec­ond automa­tion stage, the pick­ing robot works large­ly autonomous­ly. Humans only inter­vene in excep­tion­al cas­es, faults or for high­er-lev­el process mon­i­tor­ing.

A struc­tured work­ing envi­ron­ment is a pre­req­ui­site at this stage: stan­dard­ised prod­ucts are placed in defined posi­tions, uni­form con­tain­ers or load units. This clear struc­ture enables the pick­ing robot to recog­nise the posi­tion and shape of the pre­de­fined items, plan its move­ments pre­cise­ly and exe­cute them in a repeat­able man­ner.

Tech­nol­o­gy:

These sys­tems utilise advanced image pro­cess­ing (2D/3D) and sim­ple AI mod­els. Sen­sors and cam­era sys­tems record the envi­ron­ment and pre­de­fined algo­rithms cal­cu­late the grip­ping points. This enables the robot to remove the items safe­ly.

Advan­tages:

  • Max­i­mum process reli­a­bil­i­ty thanks to a com­bi­na­tion of pre­cise mechan­ics, pre­de­fined lay­outs and intel­li­gent con­trol soft­ware
  • Avoid­ance of man­u­al­ly caused errors, con­sis­tent qual­i­ty lev­el
  • Keep through­put at a high lev­el thanks to high speed

Field of appli­ca­tion:

  • Assort­ments with high vol­umes, low vari­ant diver­si­ty and clear­ly defined prod­uct dimen­sions
  • Auto­mat­ed small parts ware­hous­es, ALB sys­tems with fixed pick­ing sta­tions

This is where the tran­si­tion to ful­ly autonomous sys­tems takes place. This stage is char­ac­terised by a degree of auton­o­my that was pre­vi­ous­ly only pos­si­ble through human expe­ri­ence and intu­ition.

In con­trast to the pre­vi­ous stages, the work­ing envi­ron­ment is no longer pre-struc­tured here: Items are avail­able in dif­fer­ent posi­tions, shapes or pack­ag­ing. They can vary in terms of mate­r­i­al prop­er­ties and sur­faces and even over­lap.

Tech­nol­o­gy:

The pick­ing robot mas­ters this com­plex­i­ty inde­pen­dent­ly with the help of arti­fi­cial intel­li­gence, 3D cam­era technology/image recog­ni­tion and adap­tive grip­ping tech­nol­o­gy. Mod­ern vision sys­tems record the sce­nario in real time, analyse the size, posi­tion and ori­en­ta­tion of the objects and auto­mat­i­cal­ly select the opti­mum grip­ping point. The sys­tem can han­dle an almost unlim­it­ed num­ber of vari­ants.

The use of machine learn­ing and deep learn­ing algo­rithms enables the sys­tem to learn from count­less image data in order to pre­cise­ly recog­nise and dif­fer­en­ti­ate between objects. For intel­li­gent plan­ning of the grip­ping process (Grasp Plan­ning), the soft­ware eval­u­ates var­i­ous pos­si­ble grip­ping points and move­ment paths. It sim­u­lates their chances of suc­cess and decides on the most sta­ble and safest option.

Advan­tages:

  • High­er speed and process sta­bil­i­ty through the use of intel­li­gent, self-learn­ing robot sys­tems
  • High adapt­abil­i­ty

Field of appli­ca­tion:

  • Dynam­ic pro­duc­tion and logis­tics envi­ron­ments in which prod­ucts, orders and pack­ag­ing for­mats change fre­quent­ly
  • Mixed items on con­vey­or belts, in chaot­ic stor­age or returns
  • Large ware­hous­es with a wide vari­ety of items, e.g. e‑commerce, phar­ma­ceu­ti­cals, food

Intel­li­gent, learn­ing robot pick­ing describes the fourth and cur­rent­ly most advanced devel­op­ment stage of order pick­ing automa­tion. It cen­tres on the robot’s abil­i­ty to con­tin­u­ous­ly improve itself. It no longer just acts autonomous­ly, but adap­tive­ly and self-opti­mis­ing.

The pick­ing robot eval­u­ates its actions in real time, analy­ses suc­cess and fail­ure and auto­mat­i­cal­ly adapts its strate­gies accord­ing­ly. It devel­ops its own under­stand­ing of which grip­ping pat­terns, speeds and move­ment sequences promise the great­est suc­cess — even under chang­ing con­di­tions. The result is robots that are no longer pro­grammed, but trained.

Tech­nol­o­gy:

Tech­ni­cal­ly speak­ing, the sys­tem com­bines vision, sen­sor tech­nol­o­gy, grip­ping force con­trol and real-time feed­back. It uses advanced deep learn­ing mod­els, sen­sor fusion and data-dri­ven deci­sion log­ic to learn from every move­ment, every grip­ping process and every envi­ron­men­tal sit­u­a­tion. It recog­nis­es prod­uct types, pack­ag­ing mate­ri­als or pri­or­i­ties in order pro­cess­ing and adapts its behav­iour accord­ing­ly. In com­bi­na­tion with pre­dic­tive ana­lyt­ics and edge com­put­ing, these sys­tems make deci­sions them­selves and almost instan­ta­neous­ly.

Advan­tages:

  • React flex­i­bly to new arti­cles, lay­outs or process changes with­out the need for inter­ven­tion by spe­cialised per­son­nel
  • Con­sis­tent, pre­cise and eco­nom­i­cal, even with a high num­ber of vari­ants
  • Con­tin­u­ous process improve­ment, pick rate opti­mi­sa­tion and error detec­tion (e.g. dam­aged goods)
  • Low down­times
  • Short teach-in times for the robots

Field of appli­ca­tion:

  • High­ly dynam­ic bear­ings
  • Large dis­tri­b­u­tion cen­tres
  • Auto­mat­ed ful­fil­ment cen­tres

Practical example: Automated robot picking at Dr. Falk Pharma GmbH

At SOLTIC, we are cur­rent­ly work­ing on the automa­tion of the order pick­ing process at Dr Falk Phar­ma GmbH. The com­pa­ny spe­cialis­es in diges­tive and meta­bol­ic med­i­cine. As a fam­i­ly-owned com­pa­ny with a glob­al net­work, Dr Falk Phar­ma GmbH focus­es on the devel­op­ment and dis­tri­b­u­tion of inno­v­a­tive med­i­cines.

SOLTIC is work­ing with Dr Falk Phar­ma GmbH to inves­ti­gate the use of auto­mat­ed pick­ing robots as part of a project for effi­cient group logis­tics. The focus is on test­ing the fea­si­bil­i­ty of auto­mat­ed robot pick­ing at the com­pact con­tain­er ware­house (Auto­Store). The objec­tives of the study, which is being con­duct­ed in coop­er­a­tion with a poten­tial sup­pli­er, are

  1. Clar­i­fi­ca­tion of the tech­ni­cal fea­si­bil­i­ty of the robot solu­tion, tak­ing into account the prod­uct range of Dr Falk Phar­ma GmbH
  2. Detailed analy­sis of the impact on upstream and down­stream process­es (such as pack­ing and ship­ping process­es)
  3. Final eval­u­a­tion of the result­ing busi­ness case

Get an impres­sion of the project in the fol­low­ing video. 

Robot-assisted order picking

Picking robots: an overview of the technology behind them

Robot­ic pick­ing is far more than just a robot­ic arm. It is a com­plex sys­tem based on the inter­ac­tion of sev­er­al tech­nolo­gies:

Arti­fi­cial intel­li­gence (AI) & machine learn­ing (ML)

  • Object recog­ni­tion
    AI-sup­port­ed algo­rithms enable the robots to iden­ti­fy items of dif­fer­ent shapes, sizes, tex­tures and posi­tions (bin-pick­ing) in the stor­age con­tain­er. A pre-trained deep learn­ing algo­rithm iden­ti­fies opti­mum grip­ping points for any item.
  • Grip­ping point deter­mi­na­tion
    ML con­tin­u­ous­ly opti­mis­es the pick­ing strat­e­gy to pre­vent dam­age and ensure a high pick rate, even with vary­ing prod­uct ranges.

Vision sys­tems (image pro­cess­ing)

  • 3D cam­eras and sen­sors
    High-res­o­lu­tion 3D image pro­cess­ing sys­tems record the exact posi­tion and ori­en­ta­tion of the items in the room in order to con­trol the move­ments of the robot arm with mil­lime­tre pre­ci­sion.

Grip­per tech­nol­o­gy

  • Mul­ti­func­tion­al grip­pers
    Flex­i­ble grip­pers are used that can adapt to dif­fer­ent prod­uct types, such as vac­u­um grip­pers (for flat sur­faces), fin­ger grip­pers or adap­tive grip­ping sys­tems that can also change their grip­ping sys­tem auto­mat­i­cal­ly depend­ing on the prod­uct.

Conclusion

The sce­nario described at the begin­ning is no longer a dis­tant vision. Many of these tech­nolo­gies are already being used suc­cess­ful­ly in every­day indus­tri­al appli­ca­tions. The bound­aries between man, machine and dig­i­tal intel­li­gence are becom­ing increas­ing­ly blurred. Mod­ern intral­o­gis­tics sys­tems are no longer reac­tive, but proac­tive, adap­tive and net­worked. For com­pa­nies, this means that the future of logis­tics is not tomor­row, but today — in every facil­i­ty that relies on flex­i­ble automa­tion, adap­tive sys­tems and seam­less inte­gra­tion.

With SOLTIC, your intralogistics of the future becomes a reality

We see your order pick­ing as a place where tech­nol­o­gy and human exper­tise are not in com­pe­ti­tion, but rather rein­force each oth­er. The result is a new qual­i­ty of pro­duc­tiv­i­ty, trans­paren­cy and flex­i­bil­i­ty — and a deci­sive build­ing block for the indus­tri­al val­ue cre­ation of tomor­row.

Find out more about our ser­vices

Call up ref­er­ences

Would you like to shape the future with us?

Portrait Bastian Wenz

We look forward to hearing from you.

Bas­t­ian Wenz
Senior Project Man­ag­er, Expert

bastian.wenz@soltic.ch

+41 61 551 74 35

Call up arti­cle overview

Speak about us