Project R.A.P.T.R

[TechRoo]

Welcome to this page explaning project raptr.

R -	Realtime
A -	Autonomous
P -	Pathfinding
T -	Test
R -	Robot

Well, i have been active in the robotics world for a couple of years now. I kinda got hooked on robot building during various university projects with autonomous path finding robots. Now i want to continue one of these projects on my own, or possibly, with some outside support  .

What is RAPTR?
Well, raptr stands for "Realtime autonomous pathfinding test robot". In plain english (or plain american for you guys oversees) it means a robot that uses sensors to plot a save route from point A to point B. This route is decided by pathfinding software without any operator intervention.

What kind of robot?
The robot itself will be a medium sized biped robot. It will stand about 140cm tall (55") with 7 servo's per leg. The feet will contain 4 load/pressure sensors to detect foot placement and absolute floor height. Detection of the surroundings will be done by means of stereovision with two USB camera's.
A GHI G400 embedded .NET module will mount on the robot and send image data and servo/gyro data back to a master PC through Ethernet. Commands are also executed from this module. The  Processing will be done on a Intel NUC running OpenCV and some C code to run the inverse kinimatics. All development will be done in Eclipse IDE.

Hardware

Servo's	Robotis Dynamixel MX106T & MX64T
Camera's	Microsoft lifecam Cinema HD
Processing	GHI G400-S on robot, Intel NUC as remote processing unit

4-9-2013 - Initial Hardware and Design.

Well, i am still busy designing the legs in autodesk inventor, but should have a draft ready somewhere next weekend. Torque calculations make my head explode. I already have about half of the required servo's from previous projects, at least enough to test the first leg. Since these servo's are really expensive i will buy the remaining servo's when i need them.



These dynamixel servo's are great. They are very durable model servo's but with metal gears and dual ball bearings on the output shaft. To make mounting easier they have a free running flange on the backside of the output shaft. The servo's are connected using a 1 megabaud daisy chained TTL Uart link, making them easy to drive. Speed, Acceleration, deceleration, max torque and target position are all programmable. The servo reports back its current position, speed, torque, temperature etc.

Also, where normal servo's have a simple potmeter as an encoder, these have real contactless absolute encoders.

For anyone interested in these servo,s, check out robotis at www.robotis.com

[anoni]

Good luck!

You should look into this thing called "RoboBall", essentially your robot is pitted against other robots in a game of high stakes soccer/football! Your path finding software could be useful for that

[TechRoo]

Thanks Anoni,

The pathfinding software i worked with/on before is really designed for visual plane finding and uses a sequential (reverse kinematic) motion planner to seek available steps.

Basicly, i create a 3d height map in which i search planes of equal height which are within a certain allowed degree of tilt. These planes are locations that the robot can stand on without falling. The motion planner has a set of rules (reverse kinematic) which allow calculations to be made on the current plane (normal walking on the same height), and traversing to a different plane (stepping up or down stairs for example).

The planner basicly starts planning step after step, moving one step "back" if it runs into a dead end.


UPDATE
On a designing point of view, i am making progress but running into some issues around the ankle because of the size of the servo's and the amount of joints required down there. I want most of the frame to be made of aluminium sheet because this can be lasercut very cheap, while milling aluminium mounting blacks is expensive.

The upper en lower leg segment both use two 10NM servo,s , basicly giving me 40NM holding torque on a 400mm "arm".Taking into account that this is holding torque i should keep max torque around 30NM.

Torque = pi = rF sin O comes to 75Newton or 7,5Kg when the upper or lower leg is perpendicular to the floor. Right now the leg segments are 34,034mm³  > 34cm³ @ 2,7gr/cm³ = 91gr per segment.

91gr*8 = 728gr of aluminium per leg
153gr*6 = 918gr of MX106T servo's per leg
126gr*2 = 252gr of MX64T servo's per leg

Total = 728gr + 918gr + 252gr = 1898gr or 1.9Kg.

Obviously i am missing some brackets, the aluminium for the feet and hip but still.
1.9 * 2 = 3,8Kg, so that leaves 3,2Kg for that.

So it should be able to stand up straight, from a sitting down position
And it still stand 1 meter tall from feet to hip. Ill do some further torque calculations in Inventor when i finish the design. I have some doubts about how high it can "step" before the servo's give way.

[legolyle]

Man I love what you've been doing! where did you get the experience with CAD software and robotics?

[TechRoo]

Well, i kinda learned to work with Autodesk Autocad & Inventor by just fiddeling around with it... and bothering the guys at the design department where i work.
I am currently employed as a Software development and robotics engineer so i do a lot of work in this field, though mostly with industrial robots and vision systems.