Harpy – Sumo Bots 2009

This will be a col­lec­tion of design infor­ma­tion, videos, and pho­tos of Harpy, the robot­ic sumo plat­form I built for the Sci­ence Olympiad tour­na­ments in New York State.

Harpy’s dri­ve­train uses the motors that were used in La Cucaracha, only paired with larg­er, stick­ier wheels. It also has more capa­ble, robust 4-cell 2.3Ah A123 lithi­um-iron-phos­phate bat­ter­ies capa­ble of 70A con­tin­u­ous, but has a motor con­troller (Dimen­sion Engi­neer­ing Saber­tooth 2×25A) that is akin to com­plete crap. Videos on La Cucaracha‘s page will give you an idea of what Harpy would be like with a bet­ter motor con­troller.

Harpy took a lot of cues from lessons learned build­ing and com­pet­ing with pre­vi­ous sumo bots. Its design goals were to be as sim­ple and com­pact as pos­si­ble, min­i­miz­ing espe­cial­ly height and the length between the wheels and the front edge of the scoop. The first is to allow the bot to fit inside a foam-lined Plas­ti­case 628 pro­case, which were offered at a dis­count to Tech­ton­ics, and to low­er the cen­ter of grav­i­ty. The sec­ond was to max­i­mize robot maneuverability—Harpy does not have swing around a large front “head” like many oth­er robots do, but instead piv­ots about a point rel­a­tive­ly fore to be bot. Its over­all length was reduced, but if more length was require for space, it would have been added to the rear.

Harpy’s weight dis­tri­b­u­tion (which is to place a max­i­mal amount of weight upon tires and to min­i­mize weight on ground con­tact points that do not con­tribute to push­ing force) was heav­i­ly con­sid­ered dur­ing design, but it was not a major design goal. Sim­plic­i­ty and robust­ness of the chas­sis were seen as more impor­tant and could not be com­pro­mised by design choic­es that improved weight dis­tri­b­u­tion. Thus, a four-wheel design (dri­ven by belts or gears from the two motors) was reject­ed, as well as a scoop with adjustable “ride height.”

Pur­su­ing the goal of robust­ness and sim­plic­i­ty, Harpy fea­tures no flat hor­i­zon­tal load-bear­ing struc­tures in its chas­sis. That is, con­trary to most “stan­dard” robot designs, there is no “base plate” or plat­form on which oth­er com­po­nents are mount­ed. Instead, Harpy has four one-quar­ter inch PETG plates stand­ing ver­ti­cal­ly. They clamp onto the motors and clip onto the ends of the bat­tery pack. In this way, the stur­dy gear­mo­tors and “shot­gun bar­rel” style A123 pack (built from kit from the now-defunct Mod­el Elec­tron­ics Corp.) serve as the pri­ma­ry struc­tural ele­ments of the chas­sis.

The­se ver­ti­cal plates are also con­nect­ed to anoth­er ver­ti­cal piece run­ning left-to-right and to the scoop. That is more or less the entire chas­sis, and it is excep­tion­al­ly stur­dy and rigid com­pared to a tra­di­tion (and easy to build) plat­form style design robot. In fact, even at the tour­na­ments, Harpy uses no glue to hold its plas­tic parts togeth­er.

Pho­to Gallery


Harpy has immense push­ing pow­er, due to its 550-size gear­mo­tors and cus­tom-mold­ed wheels. Thanks to cylin­dri­cal gear­heads over the gear­mo­tors that are flush and inline to the motor cans, Harpy has wheels that fit over the gear­head cas­ing, rather than pro­trud­ing out from the side of the gear­mo­tors. The wheel forms are made from poly­car­bon­ate tub­ing (razor-sawed by hand) with inserts laser-cut from PETG. They are then placed inside a mold of sim­i­lar con­struc­tion. Path­way Poly­mers Por-A-Mold S-111 (which has a cured hard­ness of Shore A 20) was then mixed and poured into the mold.

The wheel and tire assem­bly was then bro­ken out of the mold by cut­ting away the out­er mold tub­ing on two sides, break­ing the tube along one of the cuts, and fold­ing the tube along the oth­er cut. Think open­ing up a wrist shack­le.

The result­ing wheel and tire assem­bly were extreme­ly sticky. After the wheel was mount­ed to the gear­mo­tor, each of which is about a pound, it was found that the assem­bly could stick to ver­ti­cal paint­ed dry­wall indef­i­nite­ly, and even upside-down to the ceil­ing. The robot, with clean wheels, can push more than 2 to 3 times its weight, depend­ing on the con­di­tion of the wheels, align­ment, etc. In fact, if the robot were to dri­ve into a wall from a stand­still, it can actu­al­ly pop apart all of its plas­tic parts at the tab-and-slot joints, and essen­tial­ly dis­in­te­grate. That is why it does not use any glue or plas­tic weld cement—if the struc­ture can not relieve stress by pop­ping apart at Lego-like joints, then it will crack and cleave.

New York City Region­als
Match 1 – Stuyvesant Team C vs. Townsend Har­ris Team B

Match 2 – Stuyvesant Team C vs. Townsend Har­ris Team A

Match 3 – Stuyvesant Team C vs. Unknown Team

Match 4 – Stuyvesant Team C vs. Car­do­zo Team A

Match 5 – Stuyvesant Team C vs. Unknown Team

Match 6 – Stuyvesant Team C vs. Car­do­zo Team A

Match 6 – Stuyvesant Team C vs. Car­do­zo Team A (Redo)

Match 7 – Stuyvesant Team C vs. Car­do­zo Team A

All in all, Harpy’s mechan­i­cal design was extreme­ly sat­is­fac­to­ry, except that it has nowhere to mount the motor con­troller. Also, if the scoop is just a bit lift­ed up, then the rear rests on the bat­ter­ies rather than the wheels, and it begins to lose trac­tion, which can be seen in the video for Match 7. The motor con­troller is unac­cept­ably unre­spon­sive and under­pow­ered, and a deci­sion was made to replace them with Vic­tor 883s, from La Cucaracha. The­se and oth­er defi­cien­cies result­ed in Harpy II, which has not yet seen bat­tle, but is a clear improve­ment upon Harpy I.
Harpy II

Some­what irrel­e­vant, but pret­ty:
Victor 883