Close
0%
0%

Ultimate running companion

Better than human

Similar projects worth following
Human runners telling you to get lost because you're not rich, bald, & fat? Need a way to carry your junk, get video of your running form, pace your workouts, light your path, play music?

When it was introduced in 2013, it was a modified G-buggy no-one cared about. 3 years later, it was still no more than an RC car, but self driving car startups saw their valuations explode & anyone in Silicon Valley who saw this machine suddenly wanted to write a $65 million check. Should have really driven it down Sand Hill Rd & fielded buyout offers.

The key development was the remote control.


The object of the robot is manetaining a constant speed & heading on its own.  This is just automated enough for a lion to drive it with 1 paw while doing something else.  The journey began in Jan 2014, with a $40 G-Buggy the lion kingdom got for free.


This was the seed of the idea.  The lion kingdom would probably have never developed the complete robot if it wasn't for the original free G buggy.

It was a very poor at setting a pace, couldn't carry cargo, but could carry a camera.  The potential for documenting runs was there.  There was no easy way to measure speed & the steering was binary.  With enough tuning however, the steering could be made proportional enough to hold a heading.  

The steering mechanism was an interesting reduction of a servo into a spring that centered the wheels & a clutch plate that applied a steering force proportional to the speed of a motor.   The controller it came with could only do binary steering, but with a fast rate sensor & feedback loop, the steering could be made somewhat proportional. 

The inner clutch plate & outer ring it pressed against while spinning, but not while stationary.

With a constant voltage applied to the traction motor, it achieved poor speed regulation but infinitely better than lions trying to pace themselves.  

The final form of the 1st model.

Several models came afterwards, as budgets increased, culminating in the Tamiya Lunchbox.   The steering was finally proportional & the speed could finally be regulated.

Attempts to make it self driving all failed, for reasons startups are too familiar with.  

Key to making it all work was a paw controller.  Proportional steering & throttle would be nice, but weight & durability dictated binary tact buttons.  China simply never made a durable, compact, proportional sensor.

The  final design had 2 buttons for slow steering, 2 buttons for fast steering, 1 button for throttle.  This was proportional enough for all needs, with the computer automating speed & heading.  1 switch controlled reverse.

The stick was further refined with 900Mhz frequency hopping radios, 4" packing tape to insulate the switches from sweat.

The current system is a modified lunchbox with a very large container.

Suspension was lowered.  Wheels were reinforced.  Tires were wrapped in nylon. 

Current mail order parts which have yielded the best results, after thousands of miles:



bearings    https://www.amazon.com/gp/slredirect/picassoRedirect.html/ref=pa_sp_atf_aps_sr_pg1_1?ie=UTF8&adId=A07934293L95IQDPPTPD4&url=https%3A%2F%2Fwww.amazon.com%2FTamiya-Lunchbox-Precision-Bearing-Rubber%2Fdp%2FB00V1QZ86S%2Fref%3Dsr_1_1_sspa%3Fie%3DUTF8%26qid%3D1549922090%26sr%3D8-1-spons%26keywords%3Dtamiya%2Blunchbox%2Bparts%26psc%3D1&qualifier=1549922089&id=4383766425134616&widgetName=sp_atf    10.48

Lunchbox kit    https://www.amazon.com/Tamiya-TMYTAM58347-1-12-Lunchbox/dp/B000BMVIGW/ref=sr_1_3?ie=UTF8&qid=1549922090&sr=8-3&keywords=tamiya+lunchbox+parts    147.2

motor    https://www.hobbypartz.com/96m276-2848-2400kv-16t.html discontinued

1:10 2400kv size 540 35mm diameter 3.17mm shaft.  The 2848 descriptor is not reliable.

bec    https://hobbyking.com/en_us/turnigy-5a-8-40v-sbec-for-lipo.html    6.86

Brushless servo    https://hobbyking.com/en_us/trackstar-ts-940hg-brushless-digital-helical-gear-high-torque-servo-25kg-0-1sec-72g.html    50

battery  ...

Read more »

  • sweat + switches don't mix

    lion mclionhead08/12/2019 at 02:55 0 comments

    The same throttle switch had been used for many thousands of miles, since it was built in 2015.  It gradually started failing on, when the lion got real sweaty.  It finally failed hard & drove into an intersection on its own.  When that happens, you can either kill the transmitter & wait 1 second for the watchdog timer to stop the motor, flip it into reverse, or steer it into a curb.  The fastest move is to steer it into a curb.  For the 2nd time in uncontrolled drives into roads, there happened to be no traffic.  It's a good example of the kind of durability required in an electric scooter.

    After failing for a few more miles & stopping by running into curbs, it started failing off & wouldn't go.  The lion kingdom long believed the problem was salt water getting into the switch, but didn't know of a foolproof solution or why it started failing off.  There was also the chance of salt water shorting pins.

    The outside was covered in salt & hot glue from a re gluing 2 years ago.  It actually resumed service in 2016.  Its previous life was a boundary sensor on the DVD robot from 2002-2009.

    The teardown revealed a clean spring with good tension, despite its mileage.  It makes sense since the switch is rarely toggled.  A 10 mile run with no traffic can be done in only a single button press.  The steering buttons take the most beatings.

    The problem was the wear on the conductor.  This clearly would have trouble conducting when it was dry.  It was more like arcing wear it could have experienced in a former life.  To fail on, a water drop must have formed in the tiny gap.  The pullup resistor was only 50k.  Unfortunately, there was no evidence of salt water anywhere inside the switch.  This leads to the failures coming from shorted pins.

    Another unlikely option is the spring getting weaker over time, until it would engage the conductor but  not disengage it, in hot enough weather.  The only force available to move the conductor is the spring tension & not the operator.  There were still definite click sounds of the spring disengaging.

    The switch was acquired around 2001 & it was already over 20 years old by then.  It was now over 40 years old.  Lions only used it in low current applications, as a boundary sensor.

    Technology connections funnily enough made a video about switches on the day of the failure.

    The mane idea is the force on the conductor comes entirely from a spring.  No amount of pressing on the switch makes the conductor move more forcefully.  They use a spring to make the switch bounce & arc as little as possible.

    A layer of E6000 was added to the new switch, to try to seal it from water.  The circuit board got another layer of polyurethane.

  • More gopro 7

    lion mclionhead07/04/2019 at 21:45 0 comments

    Based on a more complete picture of the features from Rainmaker Guy & the limited teardown information, the gopro7 is the same as the 6 except for the DDR being doubled. Doubling the DDR was the key to improving the stabilization.

     The stabilization sauce is implemented in Japan by socionext.com. The 2 key innovations in the last 5 years are read ahead stabilization & rolling shutter correction. Lions lived in a time when the 1st single chip encoder for H.264 came out, just 13 years ago. It was a stretch just to encode 720p on a single chip. It took 2 chips to encode 1080p. 


     Today, the Socionext SC2000 encodes 4k 60fps in H.265 while also doing read ahead stabilization, rolling shutter correction, & camera stitching. It was only a year ago that lions still had a computer with 2GB of DDR taking up a stick on a motherboard.


     All the image stabilization is driven by an IMU with a samplerate of 6400Hz. The fastest lions used in quad copters years ago was 1khz. At 6400Hz, they get 106 readings in each frame. It can easily detect changes in motion at different points in the scanning of the sensor. The memories of jello are gone. If only DSLR's implemented rolling shutter correction with gyros.


     The key to stabilization was pioneered at Google for stabilizing gootube videos. They realized buffering several seconds of video into the future allowed the algorithm to anticipate future motion out of the sensor area. The algorithm could begin panning the video projection sooner rather than after motion was already out of bounds, giving much smoother results. The same thing is done for limiting audio loudness, but here it's done with video. The effect of read ahead stabilization is quite dramatic, with no significant increase in sensor size.




    Cautionary tale for stock pickers.  Lions never bought any stock & weren't liked by their hiring manager, but despite all that going against them, the results from these cameras were always a cut above.  

  • More gopro 7

    lion mclionhead07/04/2019 at 21:40 0 comments

    Based on a more complete picture of the features from Rainmaker Guy & the limited teardown information, the gopro7 is the same as the 6 except for the DDR being doubled. Doubling the DDR was the key to improving the stabilization.

     The stabilization sauce is implemented in Japan by socionext.com. The 2 key innovations in the last 5 years are read ahead stabilization & rolling shutter correction. Lions lived in a time when the 1st single chip encoder for H.264 came out, just 13 years ago. It was a stretch just to encode 720p on a single chip. It took 2 chips to encode 1080p. 


     Today, the Socionext SC2000 encodes 4k 60fps in H.265 while also doing read ahead stabilization, rolling shutter correction, & camera stitching. It was only a year ago that lions still had a computer with 2GB of DDR taking up a stick on a motherboard.


     All the image stabilization is driven by an IMU with a samplerate of 6400Hz. The fastest lions used in quad copters years ago was 1khz. At 6400Hz, they get 106 readings in each frame. It can easily detect changes in motion at different points in the scanning of the sensor. The memories of jello are gone. If only DSLR's implemented rolling shutter correction with gyros.


     The key to stabilization was pioneered at Google for stabilizing gootube videos. They realized buffering several seconds of video into the future allowed the algorithm to anticipate future motion out of the sensor area. The algorithm could begin panning the video projection sooner rather than after motion was already out of bounds, giving much smoother results. The same thing is done for limiting audio loudness, but here it's done with video. The effect of read ahead stabilization is quite dramatic, with no significant increase in sensor size.




    Cautionary tale for stock pickers.  Lions never bought any stock & weren't liked by their hiring manager, but the results from these cameras were always a cut above.  

  • Gopro 7 stabilization

    lion mclionhead06/27/2019 at 22:41 0 comments

    Gopros had advertised stabilization for years, but only the marginal amount found on camcorders 20 years ago.  The mighty Yi 4k was a change in form factor with little improvement.  The gopro 7 was the 1st to advertise enough stabilization to approach gimbals.  The DJI action cam just looked like junk, cropped more than the gopro & had lousier color.

    Suddenly, the goo tubes lept ahead of the lion kingdom in terms of stabilization & the quest for improved stabilization got more intense.  The lion videos were truly shaky relics in a world of rock solid videos.

    The lion kingdom got by with software stabilization for 20 years & never really matched today's standards.  The lion kingdom was so desperate to capture decent footage of its mane, the decision was made.

    In reality, the gopro stabilization doesn't work in 4k wide mode.  It works in 4k linear & all modes below.  The most aggressive gimal-like stabilization only works in the "linear" modes.  The fisheye/wide modes offer minimal camcorder stabilization from 20 years ago.  There's no difference between 4:3 & 16:9 modes, so pick 4:3.  

    It doesn't change the cropping to enable stabilization & there's no way to manually crop the modes to vary the stabilization.  In unstabilized modes, the sensor area is thrown away.  It's disappointing that the amount of cropping can't be adjusted to adjust the amount of stabilization.

    The integrated video screen is a bulky, fragile, heavy, slow, battery draining pain.  There is no longer a way to use it without the screen.  Many methods have been tried to keep it from shattering.  The enclosure provides some protection, but lions often use it out of the enclosure.

    Like the other gopros, it uses the battery to record 4K, despite USB being plugged in.



    There's no way to turn off the screen manually, only set the timeout as short as possible & reduce the brightness.  The GUI is so slow, it's like 2007 again before Steve Jobless showed everyone how to do it.

    The front lens cover is replaceable & the lion kingdom wishes it knew all the lenses were replaceable, 4 years ago.

    25" has been the maximum height for a camera pole with significant reduction in maneuverability & very shaky results.  Compare the stabilized gopro 7 with the unstabilized 4.

    Stabilized:

    Unstabilized:

    The 1st run video was amazing.  The stabilization on the robot wasn't steadicam grade & slightly behind gimbal grade, but far ahead of the footage on the other goo tubes.  Would call it slightly behind the steadycam & maybe equivalent to the 1980's Star Wars footage.  It was worth $330 for that level of improvement.  The cropping cut off the lion's head for most of it.  You either have to stand farther back or aim it carefully.


    They dramatically increased the saturation.  Audio is heavily processed to remove wind noise.  It's so processed & stabilized, it's like more of the information is coming from the software than the sensor.

  • Tire wrapping

    lion mclionhead05/31/2019 at 04:39 0 comments

    The new front tires were last year's old rear tires.  After a few rotations, they were done after no more than 200 miles.  The nylon wrapping was the same as before, gluing 1/3 at a time & holding it until it stiffened slightly, then gluing around a 2nd time.  This time, a bead was laid on the 2nd day to try to make the irregularities slap around less.  

    The handle strap proved to have another use besides making the handle removable.  It allowed easily strapping stuff on top of the container.  It was top heavy but a game changer.

  • The removable handle

    lion mclionhead04/02/2019 at 05:05 0 comments

    The transmission was extremely noisy for its 1st 20 miles, despite lubrication & the sound damper.  Despite the noise, the efficiency was a record high of 240mAh/mile empty.  2200kV is a better speed than 2400kV.  The hall effect sensor worked without any firmware changes.  It actually manetaned the same speed going downhill.  There's no brake modulation, so it would have to always deliver the minimum power rather than brake.

    A 14 mile drive with 4.5 miles at 10mph & the heavy speaker pulled it up to 300mAh/mile.  The speaker at 6mph only pulled 270mAh.  Drove at high speed to try to break in the transmission & make it quieter, but it made no difference.  The latest theory is the motor is making the noise.

    The slightly enlarged cargo compartment was a huge improvement.  There's definitely incentive to upgrade the old vehicle.  Another idea gaining traction is a return to trying to make the handle removable.  Removing the tape & relying on the stiffness of the coroplastic to keep it on was a failure.  

    A hose clamp could go on the handle.

    https://www.walmart.com/ip/10-Constant-Tension-Band-Hose-Clamps-5-8-23-32/123944532

    They're expensive, too tight, & would scratch the carbon fiber.

    There's wrapping rubber bands on the handle.  They would be too lose.

    There's cutting a notch in the plastic & adding a key to the handle.  The coroplastic is too soft to retain the notch shape.  It would just become a larger hole for the handle to slide through.

    There's augmenting the handle with a big rubber band.  It would be soft enough to not crush the plastic, yet have enough tension to keep the sides in.


    The leading idea was simply augmenting the handle with string to pull the coroplastic in.  It would be a pain to remove, but it would work.  The string could use a fastener or a simple shoelace knot.

    The string idea worked quite well.  There are points of wear, like the coroplastic bending, the tape giving way, & the ends of carbon fiber delaminating.  This allowed the largest single items to be transported.

  • New motor 2

    lion mclionhead03/24/2019 at 20:09 0 comments

    The next motor was a

    SODIAL(R) High Efficiency 540 17.5T 2200KV Sensored Brushless Motor for 1/10 RC Car Truck

    https://www.amazon.com/gp/product/B075LFGCK9/ref=ppx_yo_dt_b_asin_title_o00_s00?ie=UTF8&psc=1

    After arriving 4 weeks later, this one finally fit.

    Most intriguing was the hall effect sensor connector.  It could allow speed control while coasting.  Sensorless motors just go out of control while coasting.  It would require depotting the ESC, routing 3 more wires from the ESC to the motor.

    Takes a lot more wires to access the hall effect sensor.

    A 1.7k was put in series with the hall effect output to lower it to 3.3V.  The 5V was extracted from the ESC.  Sadly, the hall effect sensors proved very fragile.  Just connecting 5V or GND to the signal line fried one.  The motor was down to 2.

    Some past telemetry was restored for viewing radio packet rates, throttle PWM, & RPM.  For the 1st time, the app would record PWM to a file.

    This revealed the slight throttle changes from changing the wheel alignment.  The lower the PWM, the higher the power.  The changes were also caused by the battery draining, but there was hope the battery drain changed it less than the alignment.  The current alignment was copied with caliper measurements.  More precise wheel alignment would still require many long drives.

    High quality, weatherproof, color coded LED connectors.  The frame is 7V, but the lion kingdom figured it wouldn't matter.

    Oil cover installed & transmission lubricated.  Struts are still not threadlocked, but it only needs 1 to finish a drive.

    Old rear tires were installed as front tires to wear them down before installing nylon wrappers.

  • New motor

    lion mclionhead02/28/2019 at 17:43 0 comments

    In a common theme with limited ordering time between long commutes & worthless currency, another wrong part arrived from China.  The 2848 designator on the Tacon 2848-2400kv does not symbolize diameter & length while it does on the B28-47-17S Brushless Inrunner 2300kv, so the required diameter is not 28mm but 35mm.   Another motor won't arrive from China until April.


    A large collection of useless servos & motors has accumulated.

    After much soldering, the electronics without the motor were finished & the cargo bolted on.

    The lights got much higher quality connectors.  Most of the other connectors were replaced by point to point soldering.  All the soldering was a bit better.

    Being a rare new build of a finished design that evolved to handle extreme abuse, it felt like a very high quality, commercial product from Japan.  The paint smelled fresh.  Everything fit together tightly.  The battery loading door was still stiff.  All the corners were still right angles & flat surfaces were still flat.  Lions never saw this design made out of virgin plastic.

    The size increase was significant.  The lion kingdom desperately wanted more room.  It made it much harder to plug in the battery.  It'll wear out the front tires faster.  Dual speaker straps were added, to try to wear the tires more evenly.  The previous camera tower won't work, because it's taller.

    There's still installing the motor in April, doing a final transmission lube, threadlocking the nuts, doing the final rear wheel installation, installing the oil cover, modifying the firmware for the gyro being right side up, fine tuning the wheel alignment, reinforcing the front tires, waterproofing the radio.  Wearing down the old rear tires as much as possible before putting them on the new front wheels.  The new front tires will go on the old rear wheels.

  • Vehicle #6 begins

    lion mclionhead02/13/2019 at 21:49 0 comments

    The total for the next vehicle ended up being around $320.  Should note the rising price of the lunchbox kit.  Lions have now built 5 vehicles.  Lions don't eat up robots.  Years of experience have shown 2 lunchbox derived vehicles to be the required fleet.  

    Hobbyking required using a chrome browser to enable shipping.  After 20 years, Firefox still can't render javascript, but doesn't lock up.  Chrome renders javascript but constantly locks up.  Nothing will ever be as fast as page loads in 1997.  That year, the backbone 1st lept ahead of demand while clockspeeds lept ahead of the static pages of the time.

    The journey began with 2 batteries.  Note China has shrunk the size, despite a negligible decrease in current.  Innovation ahead of refinancation.

    The new, higher priced kit arrived.  Remember how marvelous these boxes looked in hobby shops 20 years ago & how they're manely discarded now.

    Parts to be discarded & parts to be used are separated.
    Front wheels are strengthened with shapelock.



    Paint is applied to all exposed surfaces.  Very important to install plastic bushings as temporary masks.

    Servo is pre threadlocked.

    Battery tray is modified.


    The chassis is ground down, the hardest part of customizing this $150 kit.

    Chassis modification.

    A key modification was allowing the transmission to be removed without disassembling the entire chassis.

    Battery tray receives 1/4" spacers.

    Transmission receives wood blocks to attach an oil cover.

    Wood struts are installed.







  • Tire wear & Plans for another vehicle

    lion mclionhead02/07/2019 at 03:10 0 comments

    The 1st 2 nylon wrapped tires suddenly disintegrated after an extended period of good robustness, maybe from driving more on sidewalks instead of the bike path, so the 2nd set came together with gloves.  Lions hate wearing gloves, but  this operation is the 1st mess of carcinogenic adhesive big enough to leave no choice.  

    Though the problem is friction rather than tensile strength, there definitely should be an experiment with kevlar.  Kevlar, gloves, & some dogs would make it TSA compliant.

    With long commutes forecasted, plans for a new vehicle began.  The new vehicle will be exactly the same as the existing lunchbox.  The mane problem is the number of custom parts which must now be fabricated is quite high.  Custom tires are just the beginning.  Reinforced front wheels, suspension struts, chassis modification, cargo bay, ESC reprogramming & potting, headlights, handle, battery spacers, battery cage, transmission lubrication, radio stands, battery cables, all make you realize what it's like to be Chinese & create the modern world from nothing.

View all 70 project logs

Enjoy this project?

Share

Discussions

Similar Projects

Does this project spark your interest?

Become a member to follow this project and never miss any updates