TO ROBOTIS A PLEA FROM A BIOLOID DEVELOPER
TO ROBOTIS A PLEA FROM A BIOLOID DEVELOPER
by Dewey
Tue Jun 26, 2007 12:02 pm
This post is my plea to Robotis for what I need next from them. I am sure fellow forum members will let me know if I have it right or wrong. I make no claim that anything is really new here, just a pulling together of things, in the hope that it may reach Robotis.
Perhaps Limor you are in a position to carry the candle to them? - If you agree of course.
I will start by stating that I believe that the if you wish to experiment in biologically inspired robotics the Bioloid System is a great start. The basic actuating servo is designed for the job not a fix-up from an RC Servo. The CM5 is a powerful simple to operate controller.
If biology ‘tells us’ anything, and its worth ‘listening’ to its been practising for over 2 billion years, its that success comes from ‘sensor rich’ environment monitoring backed by ‘distributed computation’ of modest power. When a fly flies it does not solve the equations of motion like we do in a Eurofighter using large amounts of computing power, its segmented eye gives an overlap view of the area around it, this overlap with modest post processing provided by the back layer of the eye generates the motion vectors directly.
Ok so we must have more sensors, cameras, pressure switches, accelerometer etc., etc. It would be unreasonable and too restrictive to expect Robotis to make all these different sensors what we need is a Digital/Analogue Input/Output unit, perhaps in an AX12 case with standardised connectors. I know such things have been built by forum members but we need a Robotis Unit for standardisation and availablility. Standardisation would mean it would be supported directly by Robotis Software and availability means people like me who want to get on and solve the interesting things like dynamic walking can buy them and do so (well at least try). Evolution developed dynamic walking with small feet, it uses some 25 times less energy per unit mass than the static walking most of our Bioloid Bipeds are using!
The development of anything but the most simple robot requires a Software Development Environment that supports Projects with multiple Code Units.
The Behaviour Control Program Editor/Compiler is fine to get a beginner going but is inadequate/unmanageable for ‘developing in the large’. Maybe Robotis should embrace Microsoft’s Open Robotic Development Platform.
So Robotis:
Phase 1 Bioloid System (as is today) - Great Start.
Phase 2 Analogue/Digital Input/Output Unit - As soon as possible please.
Plase 3 Enhanced Software Development - When you can, it really is needed!
Dewey
Perhaps Limor you are in a position to carry the candle to them? - If you agree of course.
I will start by stating that I believe that the if you wish to experiment in biologically inspired robotics the Bioloid System is a great start. The basic actuating servo is designed for the job not a fix-up from an RC Servo. The CM5 is a powerful simple to operate controller.
If biology ‘tells us’ anything, and its worth ‘listening’ to its been practising for over 2 billion years, its that success comes from ‘sensor rich’ environment monitoring backed by ‘distributed computation’ of modest power. When a fly flies it does not solve the equations of motion like we do in a Eurofighter using large amounts of computing power, its segmented eye gives an overlap view of the area around it, this overlap with modest post processing provided by the back layer of the eye generates the motion vectors directly.
Ok so we must have more sensors, cameras, pressure switches, accelerometer etc., etc. It would be unreasonable and too restrictive to expect Robotis to make all these different sensors what we need is a Digital/Analogue Input/Output unit, perhaps in an AX12 case with standardised connectors. I know such things have been built by forum members but we need a Robotis Unit for standardisation and availablility. Standardisation would mean it would be supported directly by Robotis Software and availability means people like me who want to get on and solve the interesting things like dynamic walking can buy them and do so (well at least try). Evolution developed dynamic walking with small feet, it uses some 25 times less energy per unit mass than the static walking most of our Bioloid Bipeds are using!
The development of anything but the most simple robot requires a Software Development Environment that supports Projects with multiple Code Units.
The Behaviour Control Program Editor/Compiler is fine to get a beginner going but is inadequate/unmanageable for ‘developing in the large’. Maybe Robotis should embrace Microsoft’s Open Robotic Development Platform.
So Robotis:
Phase 1 Bioloid System (as is today) - Great Start.
Phase 2 Analogue/Digital Input/Output Unit - As soon as possible please.
Plase 3 Enhanced Software Development - When you can, it really is needed!
Dewey
by Dewey
Tue Jun 26, 2007 7:26 pm
Jon
Its not the sensors that I wish Robotis to address. I want a flexible I/O portal into the Bioloid System for connecting both analogue and digital sensors. I appreciate you and others have expended a lot of effort in achieving this. The fact that you and others all need such a capability is what I want Robotis to run with. Perhaps they have something in the offering perhaps not. Maybe they can ‘commercialise’ the work you and others have done.
I have little interest in using Motion Editors to carefully craft walking postures so my biped struggles along hopefully without toppling. I want it to walk like you or I (Thats the Goal at least). That requires sensors, at least I believe it does, (not withstanding Passive Dynamic Walking). I don’t know what the required minimal set is or the software required. On the software side I favour the neural net approach and have been looking at the work done on the Spike Stream Neural Simulator. Even though I am not the worlds greatest fan of Microsoft I am also looking into using their Robotics Studio again the high level stuff would run on the PC. The CM5 would just run low level stuff, but writing the runtime would be some undertaking. - You have a brother don’t you Jon!.
To answer your question. In handcrafted robots I have built in the past I have used:
Accelerometers based on the ADXL202E
Flex Sensors
Polaroid Ultrasonic Rangefinders
Sharp GP2D12 Analogue Infra-Red Rangers.
1655 Analogue Compasses
Simple Micro Switches
Photo Transistors
Photo Diodes
Of course I am really looking for that new sensor that I didn’t know of or hadn’t thought of by watching sites like this.
To date I have not got involved in complex vision systems involving cameras for two reasons:
1. If my robot cannot walk, it don’t matter what it sees.
2. Suspect I’m not smart enough and I would spend all my time on the vision subsystem. I am happy to pickup on the work of others who are clever enough though!
Dewey
Its not the sensors that I wish Robotis to address. I want a flexible I/O portal into the Bioloid System for connecting both analogue and digital sensors. I appreciate you and others have expended a lot of effort in achieving this. The fact that you and others all need such a capability is what I want Robotis to run with. Perhaps they have something in the offering perhaps not. Maybe they can ‘commercialise’ the work you and others have done.
I have little interest in using Motion Editors to carefully craft walking postures so my biped struggles along hopefully without toppling. I want it to walk like you or I (Thats the Goal at least). That requires sensors, at least I believe it does, (not withstanding Passive Dynamic Walking). I don’t know what the required minimal set is or the software required. On the software side I favour the neural net approach and have been looking at the work done on the Spike Stream Neural Simulator. Even though I am not the worlds greatest fan of Microsoft I am also looking into using their Robotics Studio again the high level stuff would run on the PC. The CM5 would just run low level stuff, but writing the runtime would be some undertaking. - You have a brother don’t you Jon!.
To answer your question. In handcrafted robots I have built in the past I have used:
Accelerometers based on the ADXL202E
Flex Sensors
Polaroid Ultrasonic Rangefinders
Sharp GP2D12 Analogue Infra-Red Rangers.
1655 Analogue Compasses
Simple Micro Switches
Photo Transistors
Photo Diodes
Of course I am really looking for that new sensor that I didn’t know of or hadn’t thought of by watching sites like this.
To date I have not got involved in complex vision systems involving cameras for two reasons:
1. If my robot cannot walk, it don’t matter what it sees.
2. Suspect I’m not smart enough and I would spend all my time on the vision subsystem. I am happy to pickup on the work of others who are clever enough though!
Dewey
by robbybot
Tue Jun 26, 2007 8:57 pm
hi dewey, just read this thread and checked out your bioloid progress on the forum also. i am not yet a robot owner but have spent many hours researching the hobby as it is a passion of mine, i share your frustrations with the lack of standardisation available not only with robotis but also accross all the major players. very difficult to choose which kit, i personnally have opted to wait and monitor things as im sure the industry is about to take quantum leaps forward as robotics isnt going away!
my personal choice at present is robotis as they seem to be heading for a modular , interchangeable future but they are random with what they release to the public software wise! eg. the cycloid 2 software wich allows you to operate greater DOF than the bioloid is not available for dowload yet the bioloid is? i wanted to produce a 25 dof humanoid using dx 117 and rx 64 actuators but my lack of spftware and technical prowess and general availabillity of ready to run software is sadly lacking. i feel despite this the average enthusiast if supplied with better and more singign and dancing packages can offer the hobby every bit as much in other fields than the hardened pros who have been hacking for decades (much respect to them obviously). my field of expertise is martial arts and i would love to have a package to allow me to build my robot get it mobile then spend time refineing its robo1 combat skills instead of spending hours getting it to walk three steps without falling over!
so come on robotis (and everyone else) put your money were your mouth is and lets take the next step. sorry for that terrible moan but i got the doe just loathed to spend it at present!
check out virginia techs darwin 2 humanoid interesting stuff using embedded labview running real time ...ta rob and good luck.
my personal choice at present is robotis as they seem to be heading for a modular , interchangeable future but they are random with what they release to the public software wise! eg. the cycloid 2 software wich allows you to operate greater DOF than the bioloid is not available for dowload yet the bioloid is? i wanted to produce a 25 dof humanoid using dx 117 and rx 64 actuators but my lack of spftware and technical prowess and general availabillity of ready to run software is sadly lacking. i feel despite this the average enthusiast if supplied with better and more singign and dancing packages can offer the hobby every bit as much in other fields than the hardened pros who have been hacking for decades (much respect to them obviously). my field of expertise is martial arts and i would love to have a package to allow me to build my robot get it mobile then spend time refineing its robo1 combat skills instead of spending hours getting it to walk three steps without falling over!
so come on robotis (and everyone else) put your money were your mouth is and lets take the next step. sorry for that terrible moan but i got the doe just loathed to spend it at present!
check out virginia techs darwin 2 humanoid interesting stuff using embedded labview running real time ...ta rob and good luck.
You and your friend are in an open field and are confronted by a hungry lion.The lion starts to charge and it is clear you must run.What should your immediate goal be?Some say "Outrun the lion."Others say "Outrun your friend."
by limor
Wed Jun 27, 2007 3:09 pm
There are a few Bioloid-compliant sensor boards developed on by forum members.
You can find on this forum info about JonHylands sensor board Mark Pepper's sensor board which I have used successfully with a KRG-3 gyro (see blogs link on the left).
I am involved with some close-loop control research where the PC updates the AX12 motors 100 times/sec with input from AX12 positions and a KRG-3 gyro (using the pepper board). Matlab/Simulink was used as a prototype "brain" that controls the Bioloid in real-time.
I am not aware of Robotis having plans to make such a generic little board and i have asked them in the past about it. I think that there's a better chance that community members will make a batch of a few hundreds and sell them at cost+.
A very very simple experiment you can try at home will take you a few hours of fun with basic soldering, involves Lego'ing the following components:
1) Adruino mini (see robosavvy.com/store) - simple board incorporating an Atmega168 running at 16mhz as do the AX12 servos and CM5. - well, you will need an Atmel programmer to program this guy these are available for 30$ on ebay...
2) 5V voltage regulator - to convert 10V from Bioloid batteries
3) some wires and basic soldering gear
4) SN74LVC2G241DCTR - (google) this is an IC that helps to buffer the Rx and Tx into a the Dynamixel 1-wire scheme (as suggested by Dave Hylands). appearently shorting the Rx and Tx, which would seem like a logical thing to do, is not a good idea.
5) Veroboard or similar prototyping 1mm spaced board (which can be sawed to shape)
6) Gyro, accelerometer, odometer, flexometer or any other analog or serial sensor - (see robosavvy.com/store) + it may require a 3.3 voltage regulator as most of these dont work with 5V or 10V
I posted some working software for Atmega8 pepper board on the forum previously for communicating on the Bioloid bus.
If you have never done any electronics or embedded programming, this may seem like a big deal but it really isnt and it is a great opportunity to learn while having a clear objective. The total out-of-pocket cost of this experiment is about 100$ (40-50GBP)
You can find on this forum info about JonHylands sensor board Mark Pepper's sensor board which I have used successfully with a KRG-3 gyro (see blogs link on the left).
I am involved with some close-loop control research where the PC updates the AX12 motors 100 times/sec with input from AX12 positions and a KRG-3 gyro (using the pepper board). Matlab/Simulink was used as a prototype "brain" that controls the Bioloid in real-time.
I am not aware of Robotis having plans to make such a generic little board and i have asked them in the past about it. I think that there's a better chance that community members will make a batch of a few hundreds and sell them at cost+.
A very very simple experiment you can try at home will take you a few hours of fun with basic soldering, involves Lego'ing the following components:
1) Adruino mini (see robosavvy.com/store) - simple board incorporating an Atmega168 running at 16mhz as do the AX12 servos and CM5. - well, you will need an Atmel programmer to program this guy these are available for 30$ on ebay...
2) 5V voltage regulator - to convert 10V from Bioloid batteries
3) some wires and basic soldering gear
4) SN74LVC2G241DCTR - (google) this is an IC that helps to buffer the Rx and Tx into a the Dynamixel 1-wire scheme (as suggested by Dave Hylands). appearently shorting the Rx and Tx, which would seem like a logical thing to do, is not a good idea.
5) Veroboard or similar prototyping 1mm spaced board (which can be sawed to shape)
6) Gyro, accelerometer, odometer, flexometer or any other analog or serial sensor - (see robosavvy.com/store) + it may require a 3.3 voltage regulator as most of these dont work with 5V or 10V
I posted some working software for Atmega8 pepper board on the forum previously for communicating on the Bioloid bus.
If you have never done any electronics or embedded programming, this may seem like a big deal but it really isnt and it is a great opportunity to learn while having a clear objective. The total out-of-pocket cost of this experiment is about 100$ (40-50GBP)
by Dewey
Wed Jun 27, 2007 8:21 pm
Limor
Thanks a lot for taking the time to reply in such depth.
I guess your right, Robotis are not going to produce an I/O module, certainly in the near future. Seems such a pity though as it would be a clear winner.
Building my own interface doesn’t technically daunt me, you, Jon, Mark and others have done all the hard work, its just the aside time. I have been building robots of varying complexity for 20+ years. I always seem to end up spending lots of time building the bricks of the house instead of the house often with much frustration.
Would RoboSavvy consider acting as a focus and offer a kit? I for one would be prepared to pre-pay for a couple prior to manufacture.
A kit is a good second because it brings together all the ‘awkward’ bits. For example I have not been able to find a source of the molex 3pin plugs and sockets here in the UK, I know Jon has a US source, a kit solves those type of difficulties.
Anyway I will have a delve through old posts for the technical details, I have read most of it but like an idiot didn’t note the thread numbers.
Again thanks Limor, you have convinced me to get on and make it happen not wait on Robotis.
Thanks a lot for taking the time to reply in such depth.
I guess your right, Robotis are not going to produce an I/O module, certainly in the near future. Seems such a pity though as it would be a clear winner.
Building my own interface doesn’t technically daunt me, you, Jon, Mark and others have done all the hard work, its just the aside time. I have been building robots of varying complexity for 20+ years. I always seem to end up spending lots of time building the bricks of the house instead of the house often with much frustration.
Would RoboSavvy consider acting as a focus and offer a kit? I for one would be prepared to pre-pay for a couple prior to manufacture.
A kit is a good second because it brings together all the ‘awkward’ bits. For example I have not been able to find a source of the molex 3pin plugs and sockets here in the UK, I know Jon has a US source, a kit solves those type of difficulties.
Anyway I will have a delve through old posts for the technical details, I have read most of it but like an idiot didn’t note the thread numbers.
Again thanks Limor, you have convinced me to get on and make it happen not wait on Robotis.
To Robotis a plea from a bioloid developer
by splod
Thu Jun 28, 2007 12:43 pm
Hi,
You can get Molex plugs, pins and sockets from Farnell In One in the UK.
I offered on a thread way back to make up spare cables with plugs attached to any members that required them to any length for a small nominal fee.
Cheers
Dave
You can get Molex plugs, pins and sockets from Farnell In One in the UK.
I offered on a thread way back to make up spare cables with plugs attached to any members that required them to any length for a small nominal fee.
Cheers
Dave
by Dewey
Thu Jun 28, 2007 6:13 pm
Thanks for that Dave.
Do you know a source of suitable flexible cable. Farnell do a range of silicone cable but the specs talk about temperature tolerance and nothing about flexibility. Its quite pricy and I don't want to buy it if it doen't fit the bill.
In case the UK Farnell part numbers haven't been posted here they are:
3 way 2.5mm Shroud (i.e. the cable end housing) Farnell No: 977-9557
Crimps for above 100 off Farnel No: 977-9530
3 way 2.5mm Vertical Connector (i.e. the pcb connector) Farnell No: 977-9620
Do you know a source of suitable flexible cable. Farnell do a range of silicone cable but the specs talk about temperature tolerance and nothing about flexibility. Its quite pricy and I don't want to buy it if it doen't fit the bill.
In case the UK Farnell part numbers haven't been posted here they are:
3 way 2.5mm Shroud (i.e. the cable end housing) Farnell No: 977-9557
Crimps for above 100 off Farnel No: 977-9530
3 way 2.5mm Vertical Connector (i.e. the pcb connector) Farnell No: 977-9620
by splod
Thu Jun 28, 2007 8:23 pm
Dewey wrote:Thanks for that Dave.
Do you know a source of suitable flexible cable. Farnell do a range of silicone cable but the specs talk about temperature tolerance and nothing about flexibility. Its quite pricy and I don't want to buy it if it doen't fit the bill.
In case the UK Farnell part numbers haven't been posted here they are:
3 way 2.5mm Shroud (i.e. the cable end housing) Farnell No: 977-9557
Crimps for above 100 off Farnel No: 977-9530
3 way 2.5mm Vertical Connector (i.e. the pcb connector) Farnell No: 977-9620
Hi Dewey,
I had to get my cable from the US.
Code WB-22A44 22AWG High Strand Count Silicone Wire,
Black, 10ft mini coil.
It worked out fairly expensive but I got 40ft which will last me forever for
$36 including international postage to UK.
I got the 22G because whilst being a fraction thicker than the standard it is easier to crimp when you haven't got a dedicated crimper.
Address www.banebots.com
Dave