Need advice parts/sensor advice would like to rebuild my stingray robot.
rwgast_logicdesign
Posts: 1,464
in Robotics
So this is "frankBot" As it has sat for quite a long while. Basically two years since I had my daughter.
https://www.youtube.com/watch?v=tcex8l1Gwxc
Right now it is just a stingray chassis with the original motors and wheels. It uses two 8 cell NiMh battery packs along with custom buck conversion. It has a few sonars and BlueTooth which link to an Arduino and pass info on via SPI to a prop which control the motors and handles the home made encoders. There is a second Arduino which monitors battery's and can switch packs on and off, monitor charging etc, this Arduino handles the LCD too.
Originally I thought I would use a bunch of ATMega328s, with a UNO boot loader in order to easily allow for a lot of sensors to be connected and and send there data to a prop using SPI this way I was not wasting prop pins on things which don't require a ton of speed.
I think the way I originally made this robot was just a bad architecture and I was using the propeller in the wrong way. What I would like to do now is use two propeller chip to connect to most of the sensors on the bot. For motor and battery control I want to use a Cypress PSOC 5, I mean I can drag and drop hardware based quadrature decoders, and all kinds of analog components. Basically I see the PSOC 5 as a way to build something like the Polulu motion controllers with PID adjustment, it just has a lot more power and can fit a lot more motor control routines in it than the Polulu boards! I can also use the PSOC to help filter and merge data from the encoders/compass/accelerator for dead reckoning. And finally I want to use a parallella as the major brain to handle more complex sensors and crunch algorithms fast, and do things like text to speech and other misc things that require 50 dollar modules.
The first time I built this robot I wanted to build all the hardware by hand, this time I more interested in getting the hardware up and doing its thing as fast as I can, there is no reason to make any custom PCB's or veri-board anything unless I have to, or its a whole lot more cost effective to DIY.
So here is the run down Parallela main brain, PSOC5 for motor control and pid along with some/all dead wrecking math built in to motor control functions. Two propellers to connect all the sensors and controls together and then send the info to PSOC or Parallela depending on where it needs to go. What I want to do is primary autonomys navigation along with life detection (probably use a grid eye for that. or openCV). I plan to do 2D slam using, and save the data to SD so the bot can recognize where it is.
Anyways I need some help with part selection... First off im looking for the best sensors to help make point clouds for the 2d slam. I know I will be using sonar, but I also want to use laser. The problem is price, all the decent laser scanners I know of that can do 360 degree turret are out of my price range have not seen one under 500 bucks. I know that parallax is selling a 300 dollar laser module and and there is also the new lidar light. The problem is these are still KIND of pricey and they arent scanning. I am not sure how suitable they would be to slap on top of a motor with encoder and do scans, I would like to be able to do a at least 180 degrees per second. There is also the option of using a red laser with a WiiMote Pixart cam sensor, I will have to figure out how the cam was used to make a laser range finder for the Lassiter bot another member made. This is also not a great solution considering only official wee cams work and the wee is getting old. I know you can buy replacement laser modules for the Neato vacum bots, this would work really well but im not sure how easy they are to play with. Finally the Kinnect, this route requires the use of a laptop or full on mini itx PC on board though, although the Kinnect can easily do way MORE than a 2D point could I don't want to add a PC on board just to use a kinnect. So if anyone has any ideas for any other sensors I can use to average with sonar and IR sensors to make an accurate point could I am all ears!!
Secondly last time I made my own quad encoders out of IR and PCB with wedge designs glued to them attached to the stingray motors back shaft. I would like to find motors that can still go reasonably fast but have better gearing and maybe quad encoders built in, or at least a back shaft and a set of quad encoders I know will just be plug and play with the motors! I dont even know how to figure out what kind of motors are size compatible with the stingray kit.
Lastly Battery system!! Currently I am using two diy packs made from 8 1.2v 2500mah NiMh cells. One was for the motors only and the other was probably overkill for the logic, I never PWMed the motors of 7.2v and I used custom made buck converters for the electronics power rails the way I could get HOURS and HOURS of life out of the pack running the electronics. Problem is the logic pack time to Motor pack time is pretty ridiculous... Secondly I cant get the packs to charge all the way back up to there 11.2 volts, I am lucky to get 10 on either of them. I think Sealed lead acid may be a better way to go with a whole lot less fuss but these will be kind of heavy. I don't know if I really want to use lithium because I will have to invest in a new charger and I just don't trust lithium, I have them blow up on me a few times.
Sorry about long post but if anyone has some advice id appreciate it! The last two years I have basically been involved in analog land learning all I can mostly doing RF and in the middle of a Kick Starter team for Audio products. This stuff has all been educational and a bit fun but I am getting the itch to code again and intergrate all my new analog knowledge in to my digital projects!
https://www.youtube.com/watch?v=tcex8l1Gwxc
Right now it is just a stingray chassis with the original motors and wheels. It uses two 8 cell NiMh battery packs along with custom buck conversion. It has a few sonars and BlueTooth which link to an Arduino and pass info on via SPI to a prop which control the motors and handles the home made encoders. There is a second Arduino which monitors battery's and can switch packs on and off, monitor charging etc, this Arduino handles the LCD too.
Originally I thought I would use a bunch of ATMega328s, with a UNO boot loader in order to easily allow for a lot of sensors to be connected and and send there data to a prop using SPI this way I was not wasting prop pins on things which don't require a ton of speed.
I think the way I originally made this robot was just a bad architecture and I was using the propeller in the wrong way. What I would like to do now is use two propeller chip to connect to most of the sensors on the bot. For motor and battery control I want to use a Cypress PSOC 5, I mean I can drag and drop hardware based quadrature decoders, and all kinds of analog components. Basically I see the PSOC 5 as a way to build something like the Polulu motion controllers with PID adjustment, it just has a lot more power and can fit a lot more motor control routines in it than the Polulu boards! I can also use the PSOC to help filter and merge data from the encoders/compass/accelerator for dead reckoning. And finally I want to use a parallella as the major brain to handle more complex sensors and crunch algorithms fast, and do things like text to speech and other misc things that require 50 dollar modules.
The first time I built this robot I wanted to build all the hardware by hand, this time I more interested in getting the hardware up and doing its thing as fast as I can, there is no reason to make any custom PCB's or veri-board anything unless I have to, or its a whole lot more cost effective to DIY.
So here is the run down Parallela main brain, PSOC5 for motor control and pid along with some/all dead wrecking math built in to motor control functions. Two propellers to connect all the sensors and controls together and then send the info to PSOC or Parallela depending on where it needs to go. What I want to do is primary autonomys navigation along with life detection (probably use a grid eye for that. or openCV). I plan to do 2D slam using, and save the data to SD so the bot can recognize where it is.
Anyways I need some help with part selection... First off im looking for the best sensors to help make point clouds for the 2d slam. I know I will be using sonar, but I also want to use laser. The problem is price, all the decent laser scanners I know of that can do 360 degree turret are out of my price range have not seen one under 500 bucks. I know that parallax is selling a 300 dollar laser module and and there is also the new lidar light. The problem is these are still KIND of pricey and they arent scanning. I am not sure how suitable they would be to slap on top of a motor with encoder and do scans, I would like to be able to do a at least 180 degrees per second. There is also the option of using a red laser with a WiiMote Pixart cam sensor, I will have to figure out how the cam was used to make a laser range finder for the Lassiter bot another member made. This is also not a great solution considering only official wee cams work and the wee is getting old. I know you can buy replacement laser modules for the Neato vacum bots, this would work really well but im not sure how easy they are to play with. Finally the Kinnect, this route requires the use of a laptop or full on mini itx PC on board though, although the Kinnect can easily do way MORE than a 2D point could I don't want to add a PC on board just to use a kinnect. So if anyone has any ideas for any other sensors I can use to average with sonar and IR sensors to make an accurate point could I am all ears!!
Secondly last time I made my own quad encoders out of IR and PCB with wedge designs glued to them attached to the stingray motors back shaft. I would like to find motors that can still go reasonably fast but have better gearing and maybe quad encoders built in, or at least a back shaft and a set of quad encoders I know will just be plug and play with the motors! I dont even know how to figure out what kind of motors are size compatible with the stingray kit.
Lastly Battery system!! Currently I am using two diy packs made from 8 1.2v 2500mah NiMh cells. One was for the motors only and the other was probably overkill for the logic, I never PWMed the motors of 7.2v and I used custom made buck converters for the electronics power rails the way I could get HOURS and HOURS of life out of the pack running the electronics. Problem is the logic pack time to Motor pack time is pretty ridiculous... Secondly I cant get the packs to charge all the way back up to there 11.2 volts, I am lucky to get 10 on either of them. I think Sealed lead acid may be a better way to go with a whole lot less fuss but these will be kind of heavy. I don't know if I really want to use lithium because I will have to invest in a new charger and I just don't trust lithium, I have them blow up on me a few times.
Sorry about long post but if anyone has some advice id appreciate it! The last two years I have basically been involved in analog land learning all I can mostly doing RF and in the middle of a Kick Starter team for Audio products. This stuff has all been educational and a bit fun but I am getting the itch to code again and intergrate all my new analog knowledge in to my digital projects!
Comments
ARLO
Not so much for the robot itself, but maybe some sensor ideas and such since this uses SLAM and 3D for map and navigation.
It is what I have modeled my bot on.
JT
Anyways that robot link is very cool but not quite what im looking to do... I only want to do 2d mapping, I have nothing against 3D except the only way I know to do it is either slowly pan and tilting some distance sensor or using something like the kinnect/assus sensor. Those things are GREAT i would LOVE to play with them! The thing is I dont want a full blown PC on my bot. If I had a full blown PC I would definately set up an MIT cantenna aperture radar system too! Im sure the parallela board could handle doing the USB com with the kinnect/assus and even do a much better job using its data for DSP, for example if the kinnect for PC could be programmed using Nvidias CUDA for massive parallel streams im sure you would see some neat stuff going on.
Even is the parallela could connect to the kinnect/assus I dont know enough about these guys guys to know just how open they are, I would think the assus has no market except for developers so it may be a little more open. Is it possible to write or obtain source code for linux drivers? They would need to be cross compiled to the zynq fpga which as a hard arm core (A9 quad i think). Im all in for using these kinds of sensors if it is viable to ditch the full blown PC!
Now back to the some of the other issues, I remember Erco use to recommend some different gear motors with higher torque and less speed than the stock sting rays... Im not sure what they were and Im not sure about how to figure out what style motor I should look for that wont need any modification to bolt in to a stingray. Is there a numbering system on motors that dictates size or more importantly screw patterns?
Second can anyone recommend pre-made quad encoders?? If I have to use whole made again I will but mounting the IR sensors just right is mechanically a pita in the sting ray... Im looking for fully enclosed hi resolution quad encoders. I dont care if there optical or magnetic (well optical would be better as far as EMI) and I would like them to just mount on with no rigging extra parts or JB weld! I have seen a site selling encoders cant remember it right now, they looked nice the problem was I couldn't figure out how they mounted and the price was WAY to much like 1-200 per encoder. I would pay 50 per but not 100!! Im sure there awesome in a high accuracy servo driven CNC but probably overkill here! If I knew the style motors to look for maybe I could find some with encoders built in!!
Lastly what do you guys mainly use for batteries these days? I would like a 12 volt system the more maH the better, 5000 would be great. I know you can buy sealed lead acids that arent HUGE but I have no idea how heavy they are and if a stingray especially on stock motors will handle that load compared to the 16 AA Nimh i am using now. The stingray motors are nice they'll hit 7mph!!! but they have a tough time with torque I PWM them PASM and in order to get them to transition from hardwood to a carpet room I have to drop the PWM signal down to something like 500Hz and raid the duty cycle pretty high for a split second. With creative use they can do thing lower geared motors can but im not sure about lugging a heavy battery around all the time!
They are 12V brushed DC motors with a 102.083:1 metal gearbox and an integrated quadrature encoder that provides a resolution of 64 counts per revolution of the motor shaft, which corresponds to 6533 counts per revolution of the gearbox’s output shaft. These units have a 0.61"-long, 6 mm-diameter D-shaped output shaft. NOTE: the shafts are shorter than on the original Stingray motors, so the Stingray wheel hubs will secure onto the motor shafts with one set screw instead of two.
I have the obsolete older version of the motors without covers over the "encoders" on my Stingray forums.parallax.com/discussion/comment/1055275/#Comment_1055275
I have been running the motors off a 9.6 vDC NiMh RC car battery, but I am planing to run them off of a 20 vDC Li-Ion via DC-DC convertor (regulator) in the near future.
Below is a Picture of my stingray with Pololu "100:1 Metal Gearmotor 37Dx73L mm with 64 CPR Encoder" installed
My only issue with rc batteries is they cost a ton at least at the local shop, as they should since they have like 40c discharge rates but thats useless to me.