Advanced Robot­ics – mis­sion and vision


2. October 2018


Robot­ics, what exactly does that mean? When does a machine become a robot? And what is the dif­fer­ence between con­ven­tional robot­ics and Advanced Robot­ics?

The term robot comes from the Slavic word “rob­ota”, mean­ing labor or drudgery. As a term for humanoid machines this term first appeared in the play Rossums Uni­ver­sal Robots, by the Czech writer Karel Čapek (1890–1938). Isaac Asi­mov, a sci­ence fic­tion author, first spoke of robots in his short story Runaround in 1942. The desire and idea to build auto­mated machines, how­ever, goes back much fur­ther, to antiq­uity. Leonardo Da Vinci made sketches and plans for the con­struc­tion of robots. Peo­ple have always been fas­ci­nated with cre­at­ing machines that can per­form tasks or inter­act with them. And there is another point of con­tention in the ter­mi­nol­ogy: are robot­ics and automa­tion syn­ony­mous? Accord­ing to Thomas Christaller, robots are “sen­so­ri­mo­tor machines for expand­ing the human capac­ity to act. They con­sist of mecha­tronic com­po­nents, sen­sors and com­puter-based con­trol func­tions. The com­plex­ity of a robot dif­fers sig­nif­i­cantly from other machines in the greater num­ber of degrees of free­dom and the vari­ety and scope of its behav­ioral forms”.

Tomorrow's robot is no longer deterministic, it is flexible, intelligent and autonomous.
The big dif­fer­ence: the abil­ity to act autonomously

So the biggest dif­fer­ence between a sim­ple machine and a robot is the abil­ity to act autonomously in cer­tain sit­u­a­tions. Look­ing more closely, many con­ven­tional robots today are, in the strictest sense, only very pow­er­ful machines. An exam­ple of this is the con­ven­tional indus­trial robot. Mod­ern pro­duc­tions are impos­si­ble to oper­ate with­out them. Cars are almost exclu­sively built by giant indus­trial robots, which per­form the same move­ments and the same work steps extremely pre­cisely and quickly over and over again. Although these robots are state of the art, they fall under the umbrella of con­ven­tional robot­ics. Their actions are com­pletely pre­de­ter­mined, mean­ing that they are deter­min­is­tic and not autonomous. Another exam­ple would be a wash­ing machine. Once the but­ton is pressed, it begins its stan­dard oper­at­ing rou­tine. There are no devi­a­tions. The robot­ics of the future looks dif­fer­ent. In a nut­shell, one could say that tomorrow’s robot­ics is no longer deter­min­is­tic, it is flex­i­ble, intel­li­gent and autonomous. A real quan­tum leap. In order to under­stand the extent of this change, we take another step back and look at a (still) imag­i­nary wash­ing machine: using var­i­ous sen­sors, the wash­ing machine rec­og­nizes which laun­dry it should wash today. For exam­ple, black sports­wear. Know­ing this, it selects the appro­pri­ate deter­gent and tem­per­a­ture. It uses the water level of the machine to cal­cu­late the dura­tion of the wash cycle. And if the deter­gent sup­ply is run­ning low, it sends an order to a pre­de­fined online retailer. This behav­iour is a lot clev­erer than a con­ven­tional indus­trial robot will ever be – but it’s a far cry from “advanced” because the wash­ing machine is not autonomous. Its behav­iour space is still extremely lim­ited, lack­ing the abil­ity to react to novel sit­u­a­tions. It has been well pro­grammed, its sen­sors work, but only in a spe­cific domain. If “black” and “sports­wear” do not appear in the pro­gram code, it is unable to oper­ate.

Advanced: a Mag­a­zino robot must make its own deci­sions

So let’s take a look at Mag­a­zino robots and their abil­i­ties. In a ware­house, these robots can not only take over indi­vid­ual steps, but actu­ally carry out all oper­a­tions that a human employee would in the same posi­tion, mean­ing they can com­plete the entire pick­ing process. Con­nected to the ware­house man­age­ment sys­tem, they receive their pick orders via WiFi. They then nav­i­gate autonomously to the cor­rect shelf, iden­tify the tar­get object with their cam­eras and sen­sors, grasp it, store it in their back­pack and trans­port it for fur­ther pro­cess­ing. Paths through the ware­house and grasp­ing motions are not pre­de­fined. In each sit­u­a­tion, the robot has to decide for itself the opti­mal route through the ware­house and where best to grasp the tar­get object. It is also con­fronted with obsta­cles and unex­pected dif­fi­cul­ties. In any envi­ron­ment where peo­ple work, order tends toward chaos. In other words, the envi­ron­ment is dynamic and sub­ject to change. Humans are not robots, their move­ments and grasp­ing motions are not com­puted in the same way. Humans don’t place objects with mil­lime­ter pre­ci­sion, aligned per­fectly with the shelf. A Mag­a­zino robot must be able to work with this human chaos – effi­ciently and trou­ble-free. And hav­ing over­come pre­cisely this chal­lenge makes Mag­a­zino robots “advanced” in the truest sense of the word. Mag­a­zino robots can not only adapt indi­vid­ual para­me­ters of their actions, but also decide to carry out com­pletely dif­fer­ent actions depend­ing on the sit­u­a­tion. The abil­ity to adapt allows Mag­a­zino robots to deal with com­plex real-world prob­lems.

An exam­ple: the Mag­a­zino robot TORU dri­ves along a cor­ri­dor in a ware­house and wants to turn left, only to find the pas­sage is blocked by a per­son. TORU doesn’t just stop, it imme­di­ately cal­cu­lates an alter­nate route to its des­ti­na­tion. Upon arrival, it does not find the tar­get pack­age at its expected loca­tion. With its cam­eras, it now searches the sur­round­ings to find the mis­placed object, grasp­ing it suc­cess­fully.

Advanced Cooperative Robot Operating System
ACROS – more intel­li­gent, inde­pen­dent and effec­tive robots

Job done, next one. This intel­li­gent behav­iour is based on the idea of not pre-pro­gram­ming every­thing, but rather to give the robot its own per­cep­tion through cam­eras and sen­sors, and the abil­ity to process this data in order to make deci­sions. The basis for this is the “Advanced Coop­er­a­tive Robot Oper­at­ing Sys­tem” or ACROS for short: the oper­at­ing sys­tem for our per­cep­tion-dri­ven robots. ACROS is our frame­work that will make the devel­op­ment of intel­li­gent, per­cep­tion-dri­ven robots much eas­ier and more effi­cient in the future. The knowl­edge that our robots col­lect dur­ing oper­a­tion makes them smarter in real time, because they work cloud-based and share their knowl­edge. Our goal is to make our robots more intel­li­gent, more inde­pen­dent and thus more effec­tive. For exam­ple, with a global data­base of objects and cor­re­spond­ing grasp­ing strate­gies, each robot could access this data when­ever it is con­fronted with an object it does not yet know. This is just one exam­ple of many sce­nar­ios. Per­cep­tion-dri­ven con­trol, machine learn­ing and cloud-con­nec­tiv­ity have great poten­tial in robot­ics. Every day we work on fur­ther devel­op­ing our robots and their capa­bil­i­ties. And every day we work on the future of Advanced Robot­ics – push­ing the bound­aries is what we do.