Categories
Engineering

Gaitbot

The aim of this project was to design a simple autonomous walking robot that would traverse 3 obstacles as fast as possible: stairs, stones, and hurdles.

Duration: 30 hours
Team Members: 4
Skills Applied: Laser Cutting, Mechanical Systems Analysis, SolidWorks
Tools Used: Rabbit Laser Cutter

The aim of this project was to design a simple autonomous walking robot that would traverse 3 obstacles as fast as possible: stairs, stones, and hurdles. We were given coefficients for each obstacle, which represented the importance of that obstacle to our robots movement. Variable factors were linkage, feet, and gear ratios. The outer body, baseplate, and motor gearbox were fixed variables.

We based our coupler curve off the gait of a white-tailed deer, as they are nimble creatures that have to traverse many obstacles in their daily life. We noticed that their first movement from the ground is backward and upwards, which is useful for clearing obstacles in the stair and hurdle courses, our two highest coefficients. The curve then follows a horizontal ellipse, which makes sense for speed purposes since the foot touches down earlier and further forward than if the ellipse is angled upwards. To minimize this angle and keep the curve flat, we chose to place the foot centrally relative to the coupler. To summarize, our coupler curve is useful in avoiding obstacles without losing out on speed.

​​Link​ ​Lengths:
r​1​​ ​=​ ​2.00”​ ​(Frame)
r​2​​ ​=​ ​1.50”​ ​(Rocker)
r​3​​ ​=​ ​2.25”​ ​(Coupler)
r​4​​ ​=​ ​.65”​ ​(Crank)
Note​ ​that​ ​this​ ​satisfies​ ​Grashof’s​ ​Law​ ​(r​3​​ ​+​ ​r​4​​ ​=​ ​2.9”​ ​<​ ​r​1​​ ​+​ ​r​2​​ ​=​ ​3.5”).

​​r​p​​ ​is​ ​the​ ​distance​ ​from​ ​the​ ​coupler/rocker​ ​junction​ ​to​ ​the​ ​foot.​ ​Beta​ ​is​ ​the​ ​angle​ ​between​ ​the coupler​ ​and​ ​this​ ​length. r​p​​ ​=​ ​2.24” β​ ​=​ ​41.2​

Overall​ ​our​ ​final​ ​coupler​ ​curve​ ​is​ ​similar​ ​to​ ​our​ ​initial​ ​design,​ ​though​ ​there​ ​are​ ​some discrepancies.​ ​These​ ​arose​ ​from​ ​the​ ​holes​ ​we​ ​were​ ​attaching​ ​to​ ​having​ ​fixed​ ​lengths,​ ​limiting how​ ​much​ ​we​ ​could​ ​fine-tune​ ​our​ ​frame​ ​length.​ ​In​ ​addition,​ ​our​ ​crank​ ​length​ ​was​ ​also​ ​fixed​ ​as we​ ​chose​ ​to​ ​use​ ​the​ ​crank​ ​that​ ​came​ ​with​ ​the​ ​motor.​ ​By​ ​using​ ​these​ ​constrained​ ​lengths​ ​and​ ​using ratios​ ​from​ ​MotionGen​ ​to​ ​determine​ ​the​ ​remaining​ ​link​ ​lengths,​ ​we​ ​were​ ​able​ ​to​ ​produce​ ​a coupler​ ​curve​ ​that​ ​is​ ​fairly​ ​representative​ ​of​ ​our​ ​initial​ ​design.

Motor Characterization and Gear Ratio Selection

We​ ​measured​ ​the​ ​stall​ ​torque​ ​of​ ​the​ ​motor​ ​at​ ​3V​ ​to​ ​be​ ​​0.29​ ​oz-in​.​ ​We​ ​did​ ​not​ ​use​ ​a​ ​gear​ ​ratio for​ ​this​ ​-​ ​we​ ​simply​ ​fixed​ ​the​ ​motor​ ​and​ ​used​ ​a​ ​notched​ ​lever​ ​arm​ ​and​ ​known​ ​weights​ ​to​ ​measure the​ ​stall​ ​torque.

​​Our​ ​measured​ ​stall​ ​torque​ ​of​ ​0.29​ ​oz-in​ ​is​ ​significantly​ ​lower​ ​than​ ​the​ ​manufacturer-specified stall​ ​torque​ ​of​ ​0.5​ ​oz-in,​ ​found​ ​in​ ​an​ ​online​ ​description​ ​of​ ​the​ ​motor.​ ​We​ ​expected​ ​the​ ​measured torque​ ​of​ ​the​ ​motor​ ​to​ ​be​ ​less​ ​than​ ​the​ ​advertised​ ​one;​ ​this​ ​could​ ​be​ ​due​ ​to​ ​a​ ​variety​ ​of​ ​factors including​ ​testing​ ​error.

For​ ​the​ ​gearbox​ ​we​ ​are​ ​using​ ​the​ ​configuration​ ​with​ ​the​ ​highest​ ​gear​ ​ratio,​ ​which​ ​is​ ​​203.7:1​. We​ ​initially​ ​had​ ​the​ ​gearbox​ ​set​ ​to​ ​the​ ​mid-speed​ ​configuration​ ​(with​ ​a​ ​gear​ ​ratio​ ​of​ ​58.2:1)​ ​but this​ ​did​ ​not​ ​provide​ ​sufficient​ ​torque​ ​for​ ​our​ ​robot​ ​and​ ​we​ ​switched​ ​to​ ​the​ ​higher​ ​ratio.​ ​The​ ​gear ratio​ ​for​ ​the​ ​additional​ ​stage​ ​outside​ ​of​ ​the​ ​gearbox​ ​was​ ​chosen​ ​to​ ​be​ ​​2:3​.​ ​The​ ​total​ ​gear​ ​ratio​ ​is thus​ ​​305.55:1​.

The​ ​maximum​ ​mass​ ​that​ ​our​ ​robot​ ​can​ ​support​ ​without​ ​stalling,​ ​given​ ​our​ ​measured​ ​motor torque​ ​and​ ​gear​ ​ratio,​ ​was​ ​estimated​ ​below​ ​to​ ​be​ ​​0.278​ ​slugs​ ​​=​ ​​4.06​ ​kg​.​ ​This​ ​seems​ ​very​ ​high, but​ ​is​ ​most​ ​likely​ ​artificially​ ​inflated​ ​due​ ​to​ ​our​ ​assumption​ ​that​ ​the​ ​4-bar​ ​linkage​ ​perfectly transmits​ ​motor​ ​torque​ ​(in​ ​reality​ ​there​ ​is​ ​definitely​ ​a​ ​loss​ ​of​ ​torque​ ​here)​ ​and​ ​also​ ​our assumption​ ​that​ ​friction​ ​is​ ​negligible

Foot Design

​​For​ ​our​ ​feet,​ ​we​ ​initially​ ​planned​ ​on​ ​using​ ​the​ ​yellow​ ​foam​ ​balls​ ​provided.​ ​These​ ​provided great​ ​stability​ ​but​ ​not​ ​enough​ ​friction,​ ​so​ ​we​ ​experimented​ ​with​ ​adding​ ​different​ ​friction coverings​ ​to​ ​the​ ​balls​ ​-​ ​high-friction​ ​silicone​ ​rubber,​ ​electrical​ ​tape,​ ​hot​ ​glue​ ​strips,​ ​etc.​ ​However, these​ ​still​ ​didn’t​ ​provide​ ​enough​ ​friction​ ​to​ ​make​ ​it​ ​up​ ​the​ ​stairs​ ​or​ ​over​ ​the​ ​hurdles.​ ​We​ ​decided to​ ​go​ ​with​ ​the​ ​flat​ ​design​ ​(acrylic​ ​covered​ ​with​ ​electrical​ ​tape​ ​and​ ​press-fitted​ ​onto​ ​the​ ​legs)​ ​to provide​ ​a​ ​greater​ ​surface​ ​area​ ​in​ ​contact​ ​with​ ​the​ ​ground.​ ​We​ ​tested​ ​it​ ​at​ ​several​ ​different​ ​angles with​ ​the​ ​leg​ ​and​ ​with​ ​several​ ​different​ ​tape​ ​coatings,​ ​on​ ​each​ ​of​ ​the​ ​three​ ​obstacle​ ​courses.​ ​We found​ ​that​ ​these​ ​feet​ ​work​ ​best​ ​when​ ​press-fitted​ ​all​ ​the​ ​way​ ​onto​ ​the​ ​legs​ ​(so​ ​that​ ​the​ ​foot​ ​sticks out​ ​in​ ​both​ ​the​ ​front​ ​and​ ​the​ ​back),​ ​perpendicular​ ​to​ ​the​ ​plane​ ​of​ ​the​ ​leg.
Our​ ​process​ ​for​ ​optimizing​ ​foot​ ​design​ ​was​ ​highly​ ​iterative​ ​with​ ​a​ ​trial-and-error​ ​component. We​ ​first​ ​iterated​ ​through​ ​different​ ​variants​ ​of​ ​the​ ​provided​ ​yellow​ ​foam​ ​balls​ ​with​ ​different friction​ ​coatings​ ​and​ ​positions​ ​on​ ​each​ ​obstacle.​ ​Once​ ​we​ ​abandoned​ ​the​ ​foam​ ​balls,​ ​we​ ​iterated through​ ​different​ ​positions,​ ​angles,​ ​and​ ​coatings​ ​on​ ​the​ ​flat​ ​acrylic​ ​feet​ ​before​ ​choosing​ ​the option​ ​that​ ​performed​ ​the​ ​best​ ​overall​ ​on​ ​all​ ​three​ ​obstacles.

Manufacturing a Better Mount Plate

We were also asked to design and prepare for manufacture a new baseplate design.

Design Requirements ​for ​Mount ​Plate:
1. The​ ​mount​ ​plate​ ​can​ ​be​ ​manufactured​ ​quickly​ ​(50,000​ ​units​ ​in​ ​9​ ​months)​ ​and​ ​cheaply​ ​(so that​ ​the​ ​toy​ ​yields​ ​profit​ ​when​ ​sold​ ​at​ ​a​ ​reasonable​ ​market​ ​price,​ ​not​ ​yet​ ​determined).
2. Mating​ ​components​ ​can​ ​attach​ ​firmly​ ​and​ ​be​ ​positioned​ ​accurately.​ ​This​ ​requires​ ​slots and​ ​holes​ ​in​ ​the​ ​mount​ ​plate​ ​to​ ​have​ ​precise​ ​sizes​ ​and​ ​positions.
3. The​ ​mount​ ​plate​ ​is​ ​strong​ ​enough​ ​to​ ​be​ ​played​ ​with​ ​(e.g.​ ​dropped​ ​or​ ​kicked​ ​by​ ​a​ ​child) without​ ​yielding
4. The​ ​mount​ ​plate​ ​is​ ​stiff​ ​enough​ ​that​ ​deflections​ ​under​ ​typical​ ​loading​ ​don’t​ ​compromise the​ ​robot’s​ ​ability​ ​to​ ​walk
5. The​ ​mount​ ​plate​ ​is​ ​light​ ​enough​ ​to​ ​be​ ​played​ ​with​ ​by​ ​a​ ​child.
6. The​ ​mount​ ​plate’s​ ​geometry​ ​is​ ​suitable​ ​for​ ​the​ ​chosen​ ​manufacturing​ ​process.

The​ ​mounting​ ​plate​ ​is​ ​a​ ​rectangular​ ​plastic​ ​platform​ ​for​ ​the​ ​walking​ ​robot’s components​ ​to​ ​mount​ ​to.​ ​It​ ​is​ ​stiffened​ ​with​ ​ribs​ ​on​ ​the​ ​bottom​ ​side.​ ​Slots​ ​are​ ​cut through​ ​the​ ​top​ ​face​ ​so​ ​that​ ​vertical​ ​plates​ ​(which​ ​make​ ​up​ ​the​ ​sides,​ ​front,​ ​and​ ​back of​ ​the​ ​robot​ ​body)​ ​can​ ​be​ ​press-fit​ ​into​ ​the​ ​mounting​ ​plate.​ ​Holes​ ​are​ ​cut​ ​through​ ​the top​ ​face​ ​so​ ​that​ ​the​ ​robot’s​ ​motor​ ​and​ ​gearbox​ ​can​ ​mount​ ​to​ ​the​ ​plate​ ​with​ ​screws​ ​and nuts.​ ​Our​ ​part​ ​is​ ​designed​ ​to​ ​be​ ​injection-molded:​ ​vertical​ ​faces​ ​are​ ​drafted​ ​and thicknesses​ ​are​ ​kept​ ​uniform.​ ​See​ ​the​ ​‘Manufacturing​ ​Processes’​ ​section​ ​below​ ​for the​ ​reasoning​ ​behind​ ​the​ ​decision​ ​to​ ​injection​ ​mold.​ ​The​ ​mount​ ​plate​ ​is​ ​made​ ​from HDPE​ ​plastic:​ ​see​ ​the​ ​‘Material​ ​Choice’​ ​section​ ​below​ ​for​ ​the​ ​reasoning​ ​behind​ ​that decision.

Rationale: ● The​ ​choice​ ​to​ ​injection​ ​mold​ ​fulfills​ ​the​ ​first​ ​and​ ​second​ ​design​ ​requirements:​ ​the process​ ​can​ ​produce​ ​our​ ​required​ ​volume​ ​quickly​ ​and​ ​economically,​ ​and​ ​it​ ​allows the​ ​part​ ​to​ ​have​ ​precise​ ​dimensions.​ ​See​ ​the​ ​‘Manufacturing​ ​Processes’​ ​section below​ ​for​ ​more​ ​detail​ ​on​ ​this​ ​choice. ● The​ ​slots​ ​and​ ​holes​ ​through​ ​the​ ​top​ ​face​ ​of​ ​the​ ​plate​ ​help​ ​to​ ​fulfill​ ​the​ ​second design​ ​requirement​ ​(mates​ ​with​ ​robot’s​ ​other​ ​components). ● The​ ​ribs​ ​on​ ​the​ ​mounting​ ​plate’s​ ​underside​ ​fulfill​ ​the​ ​third​ ​and​ ​fourth​ ​design requirements​ ​(adequate​ ​strength​ ​and​ ​stiffness). ● The​ ​choice​ ​of​ ​HDPE​ ​plastic​ ​fulfills​ ​the​ ​fifth​ ​design​ ​requirement​ ​(low​ ​weight). ● The​ ​part’s​ ​uniform​ ​thickness​ ​and​ ​draft​ ​on​ ​vertical​ ​faces​ ​fulfill​ ​the​ ​sixth​ ​design requirement​ ​(suitable​ ​geometry​ ​for​ ​the​ ​chosen​ ​manufacturing​ ​process).
Improvements​ ​over​ ​Kit​ ​Mount​ ​Plate: This​ ​injection-molded​ ​plate​ ​can​ ​be​ ​manufactured​ ​much​ ​more​ ​quickly​ ​and​ ​cheaply than​ ​the​ ​laser-cut​ ​acrylic​ ​kit​ ​plate.​ ​The​ ​achievable​ ​dimensional​ ​precision​ ​and​ ​ability to​ ​mate​ ​with​ ​other​ ​components​ ​are​ ​similar​ ​for​ ​the​ ​kit​ ​plate​ ​and​ ​our​ ​new​ ​design.​ ​The high-density​ ​polypropylene​ ​material​ ​has​ ​a​ ​lower​ ​yield​ ​strength​ ​(4,600​ ​psi)​ ​than 1 acrylic​ ​(10,000​ ​psi) ,​ ​but​ ​our​ ​plate’s​ ​vertical​ ​ribs​ ​provide​ ​more​ ​strength​ ​and​ ​stiffness 2 than​ ​the​ ​flat​ ​design​ ​of​ ​the​ ​kit​ ​plate.

Manufacturing

Process: ​Injection​ ​Molding
We​ ​desired​ ​an​ ​inexpensive,​ ​rigid,​ ​relatively​ ​light,​ ​and​ ​relatively​ ​strong​ ​part,​ ​so​ ​we chose​ ​to​ ​make​ ​it​ ​from​ ​plastic.​ ​At​ ​a​ ​batch​ ​size​ ​of​ ​50,000​ ​units,​ ​compression molding​ ​and​ ​injection​ ​molding​ ​are​ ​the​ ​most​ ​economical​ ​plastic​ ​manufacturing processes.​ ​Our​ ​mount​ ​plate’s​ ​geometry​ ​is​ ​too​ ​complex​ ​to​ ​be​ ​compression-molded, so​ ​we​ ​chose​ ​to​ ​injection-mold​ ​it.
Material ​Choice: ​High​ ​Density​ ​Polyethylene​ ​(HDPE)​ ​Plastic
Our​ ​design​ ​requirements​ ​specify​ ​a​ ​material​ ​with​ ​light​ ​density,​ ​relatively​ ​high​ ​strength, and​ ​relatively​ ​high​ ​stiffness.​ ​These​ ​are​ ​properties​ ​of​ ​plastics.​ ​Of​ ​the​ ​plastics​ ​suitable​ ​for injection​ ​molding,​ ​HDPE​ ​is​ ​among​ ​the​ ​least​ ​expensive​ ​while​ ​having​ ​relatively​ ​high strength.​ ​This​ ​makes​ ​HDPE​ ​an​ ​excellent​ ​choice​ ​for​ ​our​ ​mount​ ​plate.

Cost Estimate

From​ ​the​ ​IPD​ ​Product​ ​Costing​ ​Guidelines,​ ​the​ ​fixed​ ​cost​ ​of​ ​injection​ ​molding​ ​(in​ ​U.S. dollars)​ ​can​ ​be​ ​estimated​ ​with:
Fixed​ ​Cost​ ​=​ ​(Base​ ​Cost)​ ​+​ ​(Mold​ ​Machining​ ​Cost)
Base​ ​Cost​ ​=​ ​1000​ ​+​ ​(10.58)​ ​(A)​ ​(d+6)^(.4)
Mold​ ​Machining​ ​Cost​ ​=​ ​​ ​(75)​ ​(A^.5)​ ​+​ ​(2700)​ ​[(.08)​ ​+​ ​(.04)​ ​(SP)]^(1.27)​ ​+​ ​300 +​ ​(120)​ ​[A^(1.2)​ ​]

A​ ​=​ ​area​ ​of​ ​mold​ ​base​ ​=​ ​(13​ ​in.)​ ​(6.5​ ​in)​ ​=​ ​84.5​ ​in^2
d​ ​=​ ​depth​ ​of​ ​mold​ ​=​ ​0.3​in
SP​ ​=​ ​number​ ​of​ ​surface​ ​patches​ ​=​ ​16

​Base​ ​Cost​ ​=​ ​$2,867.72
​Mold​ ​Machining​ ​Cost​ ​=​ ​$27,395.30
​​Estimated​ ​Fixed​ ​Cost​ ​=​ ​$30,263.02

The​ ​variable​ ​cost​ ​can​ ​be​ ​estimated​ ​with:
Variable​ ​Cost​ ​=​ ​(Material​ ​Cost)​ ​+​ ​(Cost​ ​of​ ​Cooling​ ​Time)
Material​ ​Cost​ ​=​ ​(1.03)​ ​(C/W)​ ​(W)
Cost​ ​of​ ​Cooling​ ​Time​ ​=​ ​(2.18)​ ​(hmax)^2​ ​/​ ​(alpha)

C/W​ ​=​ ​cost​ ​per​ ​weight​ ​of​ ​material​ ​=​ ​0.90​ ​$/kg
W​ ​=​ ​weight​ ​of​ ​part​ ​=​ ​0.0406​ ​kg​ ​​ ​(estimated​ ​in​ ​Solidworks)
hmax​ ​=​ ​maximum​ ​wall​ ​thickness​ ​=​ ​0.10​​in
Alpha​ ​=​ ​thermal​ ​diffusivity​ ​of​ ​material​ ​=​ ​0.11​ ​mm^2​ /​ ​s

Material​ ​Cost​ ​=​ ​$0.0376
Cost​ ​of​ ​Cooling​ ​Time​ ​=​ ​$0.1982
​​Estimated​ ​Variable​ ​Cost​ ​=​ ​$0.2358

Total​ ​Cost​ ​=​ ​(Fixed​ ​Cost)​ ​+​ ​(Num.​ ​Parts)​ ​(Variable​ ​Cost)
​​Estimated​ ​Total​ ​Cost​ ​=​ ​$42,053.02

Cost​ ​per​ ​part​ ​will​ ​be:
Cost​ ​Per​ ​Part​ ​=​ ​(Total​ ​Cost)​ ​/​ ​(Num.​ ​Parts)
Estimated​ ​Cost​ ​Per​ ​Part​ ​=​ ​$0.84

Documentation can be seen on the next page.