The Future of Automated Random Bin Picking

Robot­ic bin pick­ing is incred­i­bly impor­tant as it is able to take the tedious, monot­o­nous tasks off of work­ers hands. It is able to effec­tive­ly han­dle bulk parts sort­ing, dan­ger­ous oper­a­tions, and/​or labor-inten­sive order ful­fill­ment. Because of its com­plex­i­ty, ran­dom bin pick­ing has yet to be per­fect­ed or com­plete­ly grasped by automa­tion sys­tems. How­ev­er, vision­ar­ies are pio­neer­ing their way into that unchart­ed ter­ri­to­ry. Many experts pre­dict that robot­ic ran­dom bin pick­ing will become main­stream by 2020 as there are a vari­ety of sub­sets of bin pick­ing are already commonplace. 

A Staff Engi­neer at Fanuc Amer­i­ca Cor­po­ra­tion in Michi­gan, David Dechow, says that the ​“Mar­ket­place is hyped up about robot­ic bin pick­ing. The real­i­ty is that machine vision is suit­able, robust, and reli­able for a sub­set of that entire world of things we would like it to do, whether it be inspec­tion, 2D guid­ance, or 3D guid­ance. From my hands-on expe­ri­ence over a num­ber of years, bin pick­ing is in the same cat­e­go­ry now. There is real-world capa­bil­i­ty, but it is still a sub­set of the whole world of things we would love it to do.”

In gen­er­al, bin pick­ing can be divid­ed into three dif­fer­ent cat­e­gories. Each con­tains a spe­cial amount of con­sid­er­a­tions based on the part posi­tion that results in is a dif­fer­ent lev­el of com­plex­i­ty, cost, and time. As one may guess, the char­ac­ter­is­tics of parts often deter­mine how eas­i­ly they will be picked from a bin. The parts that have the great­est suc­cess in Bin Pick­ing are geo­met­ri­cal­ly sym­met­ric parts. These parts don’t have strange fea­tures, are not too heavy, and have some sort of suf­fi­cient pla­nar sur­face in all of their ran­dom ori­en­ta­tions. These few geo­met­ric sur­faces are very easy to pick up with a robot and very easy to grip.

The three cat­e­gories are as fol­lows: Struc­tured Bin Pick­ing is when parts are sit­u­at­ed in an orga­nized pat­tern in which they can eas­i­ly be iden­ti­fied and picked. This can be auto­mat­ed using 2D Vision, imag­ing, and analy­sis. Semi-struc­tured Bin Pick­ing is when the parts are posi­tioned with mild amounts of orga­ni­za­tion to aid in pick­ing. Ran­dom Bin Pick­ing is when the parts have com­plete­ly ran­dom posi­tions, can be over­lap­ping, and have mul­ti­ple ori­en­ta­tions, mak­ing this the most com­pli­cat­ed ver­sion. Advanced tech­nolo­gies con­sist­ing of 3D imag­ing and 3D analy­sis will have to be cre­at­ed to tack­le the most chal­leng­ing parts yet; the shin­gled, pack­aged, or deformable parts that are dif­fi­cult to image with machine vision. The height vari­a­tion, thin nature, and par­tic­u­lar geom­e­try make it very con­fus­ing, even with 3D imag­ing tech­niques.

In order to con­tin­ue to work towards suc­cess with 100% accu­ra­cy for ran­dom bin pick­ing, there are three big com­po­nents that need to work togeth­er. These are the sen­sors, soft­ware, and EOAT. Many com­pa­nies are aware of this and are cur­rent­ly rac­ing to cre­ate the best, most effi­cient soft­ware to pro­vide the per­fect ran­dom bin pick­ing solu­tion. They are in the works of ensur­ing the per­fect algo­rithms to pre­vent things such as the robot and EOAT from col­lid­ing with the sides of the bin and oth­er objects. Fanuc has already stepped up to the plate and cre­at­ed an Inter­fer­ence Avoid­ance fea­ture that is soft­ware based and comes stan­dard with the 3D area sensor. 

So far, there has been a suc­cess­ful turnkey (cell, robot, process automa­tion equip­ment, vision, arti­fi­cial intel­li­gence, and safe­ty equip­ment. etc.) solu­tion for robot­ic phar­ma­cy order ful­fill­ment sys­tems that can ensure 100% guar­an­tee. The RX Unit Pick Work­cell is a com­plete­ly auto­mat­ed robot­ic phar­ma­cy order ful­fill­ment sys­tem that can han­dle thou­sands of spe­cif­ic, pre-pack­aged, med­i­cine for mail order deliv­ery or to fill for cen­tral loca­tions. This work­cell can make dis­pens­ing of pre-pack­aged, high vol­ume units-of-use, look easy.

Anoth­er big area of growth for robot­ic bin pick­ing is E‑commerce and order ful­fill­ment. Stores such as Wal­mart, Cost­co, or Tar­get may even­tu­al­ly look to automa­tion for their online order­ing. The chal­lenge right now is the end effec­tor that needs to be able to grab a bot­tle of deter­gent and/​or tube of lip balm. Soft Robot­ics is work­ing on adap­tive grip­ping tech­nolo­gies, seen with their flex­i­ble grip­pers com­bined with Uni­ver­sal Robot­ics Spa­tial Vision Soft­ware.

As you can see, it is so impor­tant for researchers to con­tin­ue think­ing, dream­ing, and ulti­mate­ly cre­at­ing solu­tions to these every­day, ran­dom bin pick­ing appli­ca­tions. At Robots​.com, we believe that the remain­ing appli­ca­tions that pose spe­cif­ic chal­lenges will be solved in the near future. We are excit­ed for what is to come! For any ques­tions about automa­tion or bin pick­ing robots, con­tact us online or give us a call at 740−251−4321.