Can PI-Bot calculate motor rotations to get travel distance

[claudep]

Greetings

Question - Can the PI-Bot record the number of rotations that the motor makes?

Discussion - 

Recently got my PI-Bot and as a project would like to record a room layout including distance traveled using motor rotation count, gear ratio and tire circumference to get distance traveled.  I think this would be better than setting robot speed and measuring time to travel eg 10 ft and using this conversion factor to calculate distance traveled.  

The robot would use the sonar sensor on a 45 deg angle (two would be better, one facing side and one ahead.) The robot would start from a corner, and travel recording distance until it reached an inside or outside corner. Eg wall change, furniture, etc.  It would then stop, record distance (length of wall) , then make appropriate turn, calculate angle and record or transmit a stream, then proceed along next wall leg and repeat.  Inside corners might be a little tricky with only one sensor.  The robot would continue until a net angle of 360 deg is reached.  (initially doors would be closed) The recorded data could be streamed or uploaded from Stem board.  Could also try to send data to Raspberry PI via serial or other port connection.  Here is a video I made if you are interested   youtu.be/kAMaUuBVK9I

I currently have one of my Raspbery PI's ( B+) setup with opencv using the instructions from this UK group.  blog.dawnrobotics.co.uk/2014/06/programming-raspberry-pi-robot-using-python-opencv/
This provides a web interface and their robot includes a pan/tilt camera mount as an option. This is a learning experience but eventually I would like to build my own self balancing, opencv vision capable somewhat autonomous robot.  I am currently looking at power packs with built in photo cells.  It would be nice to have the robot monitor batter and self charge by finding a sunny spot in the house to recharge itself.

Anyway one step at a time.  I expect it will be a while before my goal is reached but everyone has to have a vision.

After building the PI-Bot it struck me that this was like the early days of computers. Like when I got my original Apple ][. Each computer company produced it's own unique computer hardware, OS, software with no interchangeable parts, component standards, cable connectors or fasteners.  There were computer clubs to provide support. Assembling the PI-Bot was more like a mechno set with nuts and bolts, tiny wires, no built in cable management or cable standards, no standard mounting or snap in fasteners.  Even cars and truck use snap in fasteners.  It is a leaning tool but I have build many computers from scratch and at least power supplies, cables, cable connectors, card slots, accessories, Etc. are standardized to a great extent.  

The price of computer components, reliability and computing power has steadily improved over the years so that we take things for granted.  I believe this is where personal robotics should be heading.  All kit building aside. After putting the PI-Bot together and redoing wiring once to get it to work OK, It appeared to me that if personal robotics is to achieve what the PC industry has done, it needs to have standards with USB style plug-in components and sensors, standard high level operating systems, robot app store, standard robot connections, communications protocols, robot training protocols Etc.

At any rate I believe personal robotics has a long way to go.

Thanks for listening
and If someone can answer my distance question, that would be great.

Update - After doing some searching again it looks like motors are pretty dumb and don't provide feedback as far as I can tell. Might be able to put a white tape line on one wheel and have light sensor read each wheel rotation via the white mark on the tire.  Cludgy but will give it a try.

Found these www.pololu.com/product/2459 light sensors.  Will try them out with one on each wheel that has a stripe pattern on the inside of the  tire or hub alternating black and white.  This would give a more accurate distance.  Problem will be that I may need light source (white LED) since it relies on reflected light and won't work on when it is dark or low light conditions.  Hoping beam will be narrow enough to not get false readings.  Using one on each wheel will allow calculating radius of turn and average distance to center line of robot. (average distance between each wheel.  This might give a more accurate reading.  Need to see how hub speed affects reading as well.  If the wheel turns too fast then possibility that light pulse will be missed.

Still looking at the possibility of using motor PWM signal to motors to get shaft revolutions.  Not sure why rotation counter is not part of motor design.  Cost I guess.


Claude Pageau



 

[jasea]

Hi Claude, It is good to see projects like this, especially one that is intended progress onto an autonomous robot. My target is to produce an autonomous model yacht that will sail across the Atlantic!

You are correct in that the motors of the Pi-Bot are dumb, they are fed pulses of electricity (PWM) and the length of the pulse determines the effective voltage applied to the motor. I like your idea of using a white line on the tyre, it is better than my first thought of changing the drive to stepper motors, I would use paint between the treed of the tire, may keep cleaner.

On another robot similar to the Pi-Bot I have used a simple compass module to keep track of the direction being taken, and also to get out of "hitting the wall" situations, where I turn 90 degrees one way and see if it is clear if not then turn 180 degrees to see if it is clear in that direction, if I can't go in either direction then about turn and try somewhere else. I started this as part of the boat project using a GPS but the small distances involved inside the house meant that the GPS was ineffective.

Regarding the use of the sonar sensor, Google "hc-sr04 manual " and you will see various documents about this sensor, the important bit is the measurement angle of the sensor, it is reported as being only 15 - 30 degrees, which explains why my Pi-Bot keeps hitting walls at a slight angle, and that you may need more that two.

I think that the need for sensor connection standards will be eliminated by the "Internet of Things " where all devices will use wireless connections and the need for hardwiring will be reduced to power connections.

[claudep]

Hi Jasea

Personal Robotics appears to be somewhat in the stone age as far as hardware standards and connectivity is concerned. I have added to my posting on installing special light sensors on each wheel (see original post for details). Allows calculating radius of turns and gives some way of double checking distance accuracy based on PWM number sent to each wheel. If accurate both wheels should show same distance with same PWM number signal to each wheel.  Reading wheel tracks will still be a problem in a dark room unless I install LED light source so the wheel light sensors will work.

I am having fun even if this is a basic toy learning robot. Just looks kind of Frankenstein with everything hanging out in the open even with good wire routing.
BTW just received my gyro, compass, accelerometer from Pololu www.pololu.com/product/2468  Will do some testing on the PI-Bot but this will be used for the self balancing robot which won't be for a while yet.  Still have to build Carbon Fiber chassis and drive train etc plus software.  Also looking at battery packs with photo cell array built-in so the robot can recharge batteries itself by sitting in the sun in front of a window. Not sure how to accommodate free standing if it needs to sit but probably a solenoid operated landing gear of some kind might work.  I have a long way to go.  Getting a few ideas form dog behavior as well.

Thanks for your comment
Claude.

[jasea]

Hi Claude,
I have had a look at your updated post. I am not too sure that these Pololu 2459 sensors are what you need. It looks to me as they determine how much reflected light is available by measuring the decay of the charge on a capacitor. The rate of decay is varied by a light detector increasing the current follow as the reflected light increases.
For what it worth, I think the you need an encoder that will count pulses from a disc attached to a wheel. Have a look at www.societyofrobots.com/sensors_encoder.shtml for a general description, it might be useful for your robot as well. At the bottom of the page you will also see a description on how to make an encoder wheel. Also have a look at www.societyofrobots.com/schematics_infraredemitdet.shtml which shows a basic circuit for the emitter and detector. What you will then need is code that attaches an interrupt for the pin ( either D2 or D3) and will be run on a rising signal. You can then keep count of the number of white bars that have passed under your sensor.
Photocells for battery charging have been a problem with my yacht project, the size of the yacht is being determined by the size of the photocells that I need to keep the batteries charged. So every good idea that I get that increases the total current consumption can result in an increase of the hull size to accommodate additional photocells.

Regards

John

[claudep]

Hi John

Thanks for your reply. I did try the light sensors that came with the PI-Bot. The test white stripe on the wheel was 1/8 " wide. I did get the PI-Bot line following reflected light sensor to reliably and accurately count tire revolutions but needed a short delay to avoid repeat signals (I guess this is what you meant). Did not try multiple stripes but will test this as well. Thought a printed round disk on the inside of the hub might work better than tape on the wheel tread. Read the articles and have a better understanding of the wheel rotation encoder limitations. I am thinking that a magnetic sensor might be better since it will work in the dark. This might affect compass in gyro/accelerometer chip (more testing).  Magnet would be similar approach to bicycle type system only magnets would be glued/fastened to inside of wheel hubs. The plan is to have accurate waypoint to correct robot navigation errors that might accumulate. Something like doing course corrections based on sextant readings and clock. I plan to use opencv to read room markers/signs so robot can triangulate exact location when needed/scheduled.

Currently I am working on my self balancing robot chassis prototype. Have a drawing and bought some black Elmer's foam board 20"x 30" x 1/16" (Walmart $4.98 Canadian). Looks quite strong, durable and very, very light. Cheaper than carbon fiber sheets and should be more than strong enough. Currently testing various joints but looks like the best will be a v cut groove with 90 deg bend where foam backing is not cut on the outside of joint.  Inside of Joint is glued with Elmers Xtreme (still need a foam board X-Acto cutting tool that is designed to cut accurate 45 deg angles at precise depth).  Might use straight pins on joints as needed to add joint strength as well.  I have used a few pins to hold joints while glue dries (takes 24 hrs) 

The PI-Bot gearbox chassis will be removed from PI-Bot plastic mount and will be mounted vertically on my new chassis. Still working out details of best gearbox chassis mount point and design. Will use double layer of foam board for stress points like gearbox mount point. Will allow for mounting shims (spacers) to easily adjust gearbox balance point under robot. I have designed a removable shelf system and will use velcro to lock in place. Have a very rough prototype partially built but will make a better one with new v-groove joints. So have not tested hardware and new chassis yet so PI-Bot (I-Frank for Frankenstein) still has his original home plastic base for a while.

Will initially mount the PI-Bot caster on back to allow the robot to drive while standing without self-balancing. Final design will use a linear motor driven support system that will be activated to prevent robot falling when it shuts down self balancing. This would also save battery when robot is stationary and does not need to self balance.
Expecting this design will do a good job and look very nice as well.

Received my openelectron pan/tilt PI camera module mount recently as well. Works nice but took a while to put together. Working on better camera and servo cable management (ordered longer RPI ribbon cable 30 cm.  Might need to design a cable caddy of some sort to keep pan/tilt operation from touching cables.  Will try to capture video stream to RPI and have camera track an object with pan and tilt to keep in frame. Code is written in python.

Working on a YouTube video update.






Sorry for rambling on.  Bye for now
Claude ...

[jakkals]

A quick and cheap sensor is to paint the top half of the axle with an isolator and have 2 wires pressing against it. The rotation of the axle will complete the circuit every turn.

[claudep]

Hi Jakkals

Wow Great idea. Thanks. No magnets to disrupt compass or light sensors that need lighting in the dark. This would only need a low voltage to the contacts. It almost sounds too obvious. I wonder if there is a shaft hub or wheel hub with contacts and a contact pickup that can be purchase, rather than make one.

Love your idea.

[jasea]

Hi Claude,
Regarding the self balancing robot, have you seen this ozzmaker.com/2013/04/18/success-with-a-balancing-robot-using-a-raspberry-pi/ ?

John