Revolutionizing Robot Building

[xaockhaos]

Collaborative robot building

My suggestion for revolutionizing robot building is to tap on the knowledge of the web to build a robot that is comparative to the commercial robots;
 http://www.galaxyzoo.org/ // the idea of collaborative robot building is taken from this website, it currently produces multiple research papers from observation of the public
Arduino , Linux , Human Genome Project are prefect examples to prove that collaboration is better than going alone;

I believe that generalizing (multi-purpose) robotics platform will be the direction to the future;
Key properties of the Multipurpose Robotics platform (MRP) = self assembly + self sufficient (Taps into renewable energy / voltage source) + self navigation;

The combination of the 2 ideas that i had just proposed may have huge implications;
I have the general idea in mind and i love to share my ideas, if anybody have any suggestion/ comments about the ideas proposed please share and let us collaborate together;
May it be the first Networked Robot to be design and created by all;

[robertc]

Few days ago I was searching around the web for information because I was trying to understand the implications of building a small humanoid robot capable of executing small tasks like self balancing, walking, executing commands or avoiding objects while walking. I found out it's not easy but indeed very complex because you really need a good engineer mind and something like a humanoid robot would involve a big investment ($$$).

I like your idea because I'm thinking the same. I think that the key to building advanced robots like commercial ones is to use the existing resources from the web but in a collaborative way.

Wasn't linux created in the same fashion ? Programmers world wide joined togheter to develop programs, distros and so on.

A networked robot that would be great !
I have so many ideas

[xaockhaos]

There are cheap alternative to humanoid robot

Hardware

I had purchased an humanoid robot (with 996R servos) which cost less than USD 300, shipping fees included.
It worked just fine, i use Arduino Mega to control the 17 servos and up till now i only manage to make it stand.
Next step i am going to add in the 2 mechanical grippers (USD 15) to achieve the some self assembly capability. To achieve full self assembly capability the humanoid arms must be longer and there must be a rotary tool attached to the robot.
Main purpose of rotary tool with changeable tips is to cut grind, screw and many more.
To recognize each tips will need a very strong computer vision program.
The camera module chosen is ov7670. (USD 10)

Software
As for software i had decided to use matlab to control the robot, matlab have very powerful software codes that can tackle any problems but the learning curve is quite steep.
Computer vision can be achieved relatively easy with matlab.


Links
http://item.taobao.com/item.htm?spm=a1z10.3.w475680914.47.lGemxZ&id=16790573580&      //use google translate to read (Humanoid robot)
http://item.taobao.com/item.htm?id=14782419292&spm=2014.12011554.13.68                         // Mechanical grippers
http://www.dremel.com/en-us/tools/pages/ToolDetail.aspx?pid=3000                                          // Rotary tools


http://www.seeedstudio.com/depot/robovie-x-p-1211.html?cPath=170 (USD 2203)                     // Commercial humanoid robot with no self assembly capability

[xaockhaos]

To initiate collaborative robot build i would need guidance/ questions/ problems from anybody so please try to reply and share your ideas no matter how ridiculous it is. I would write it down in the word format and summarize what i had gathered through the forums and share with all of you.

I am going to kick start with my ideas.

Key properties of the Multipurpose Robotics platform (MRP) = self assembly + self sufficient + self navigation

Try to visualize each section as an 3D modular electronics block.
- Display (TFT)
- Power supply ( Solar + Lipo battry + charging circuit + battery level indicator)
- Communication ( Bluetooth + Xbee or NRF24L01 )
- Navigation ( Camera + ultrasonic sensors + lasers + human detectors + IMU + barometric +  photo resistor +IR + sound sensor)
- Interfacing ( Servos connectors, Motor connectors , External sensor connectors, magnets, current sensor)
- CPU (Arduino or STM32F4)
- Motor (L293D motor driver + motor + encoders)

Need not implement all of the sensor, but think of it as interchangeable modules that allows the support for all sensors.
All of the 3D modular block can be removed except for the CPU & power supply , else it will still work.
The problem now is the Interfacing circuits to the various platform the MRP will work with.
There are 3 platform this MRP is planned to work with, Quadcopter, Humaniod and a Truck.
Feel free to suggest more platform , but more platform means more problems to the interfacing circuit design.

[robertc]

Why not Raspberry PI(25$). It's a 700Mhz ARM CPU but I don't think it's powerful enough for computer vision. It handles well blob detection with SimpleCV and python using a cheap(10$) webcam.

Matlab is more powerful but you have to run on a dedicated host computer with wireless(either wi-fi or bluetooth) connection between MRP and the host. Is it possible for the software to run on board the MRP ?

[robertc]

Can you provide more feedback about the humanoid robot you just bought ?
I'm wondering how are the servos working because I don't really trust these analog servos. They brake often and due to low resolution and slow response time they don't fit well into a humanoid robot. Digital ones are great but more expensive.
Do you program the movement step by step or are you using some kind of algorithm like inverse kinematics? Which one would be better ?

[xaockhaos]

It is possible to run Simulink models as standalone applications in BeagleBoard and Panda Board, but not Rasberry PI.
http://www.mathworks.com/academia/pandaboard/

There is premade models in simulink for Arduino and STM32F4 but like what you had said an bluetooth / wifi connection to computer will be needed.

However the price will increase by a lot

- BeagleBoard ($149)
http://sg.mouser.com/ProductDetail/BeagleBoard-by-CircuitCo/BeagleBoard-xM/?qs=sGAEpiMZZMsi5STK09QP%252bKiOKGrBbCoZ

- PandaBoard ($181)
http://sg.mouser.com/ProductDetail/PandaBoard/UEVM4460G-02-01-00/?PandaBoard/UEVM4460G-02-01-00/&qs=sGAEpiMZZMvFPGEOwQcrY4/7rxV7jhbyK5P6Sy2KQB8=

- Arduino ($38)
- STM32F4 Discovery ( $14)


I had just tried to control the servos (946R) individually and to make it stand. It worked just fine and it is able to withstand the weight of the humanoid.
Digital servos are great but with so many servos on hand it is over my budget so i had to make do with these servo during the experimentation phase.

As for the humanoid motion i am currently implementing it on simmechanics with reference to the pdf below.
I am also going to place the IMU on the CG of the humanoid so that the robot could balance.
The simMechanics modeling is going to take some time thou.
http://www.internetjournals.net/journals/tar/2011/January/Paper%2005.pdf

Cheap IMU + Barometer  http://item.taobao.com/item.htm?spm=a1z10.3.17-5928574584.37.TLlWjQ&id=17523656949& ($20)
Work very well with arduino mega.

[xaockhaos]

After researching on the above mention processor, an STM32F4 will be enough for an humanoid, quadcopter and truck.

2 main things stm32F4 is that it have digital signal processing and floating point unit (allow us to control video/voice signal and implement mathematical function better than arduino)
ARM cortex A8/A9 which is the raspberry pi, panda and beagleboard IC chip will be overkill and over complicated for this project. Although it can run standalone with matlab but there STM32F4 can be implemented without the use of matlab so an bluetooth/ wifi connection will be great but non essential.

I am currently studying electrical engineering so my strength will be designing PCB.

The light at the end of the tunnel will be a 3D modular electronics block / quadcopter/ truck / humanoid ,which the humanoid robot can assemble and disassemble easily. (Self assembly)

An a power supply module which can monitor the amount of energy consumed and navigate itself to a USB charging port. Or using photo resistors to move to place that have access to solar energy. (self sufficient)

Self navigation which is the easiest to implement just need an GPS, IMU,a few ultrasonic sensors to have coverage on each angle an last but not least a pair or a camera module.

The problems that will be obstructing the progress is
- Control of movement of humanoid robot using STM32F4
- The types of connectors between the modular block that the humanoid robot can self assemble easily.
- Assembly of quadcopter and truck.
- Assembly of humanoid robot itself (very hard problem indeed)
- Types of connectors in the interfacing circuit ( 17 PWM , 4 high power motor (for quadcopter and truck)
- How modular is 3D modular electronics board, it is quite hard for self navigating layer to control all or the sensor mention. So which are more impt and which are not.
IMU (must have)
Camera (must have)
GPS (can live without but good to have)
Ultrasonic sensors (can live without but good to have)

So that is to say all the tools i had used to create anything, the humanoid robot must be capable of controlling it.