Microsoft Robotics Studio

[smile_nik]

Microsoft seems very confident that the Microsoft Robotics Studio is the solution to all problems in the robotics, what is your opinion?
Is there anyone that actually try this studio with his robot?

[Robo1]

I've read quite a bit about it in a year or two it could be a good product but at the moment it is only a beta so would require more effect to make it work then just programming you robot in something else.

have a look at robot-dreams they have a 30 min interview with one of the tests. It does promess to be quite good if and when it becomes available fully released. But from previews MS products the have good ideas lots of money nd they still don;t quite get it right. plus the fact do you think it will be open source no I don't think so will they give you lots of support unless your giving them millions no I don;t think so.

I'm not one of these MS is bad and there the devil I like office but how hard is it to write a word processor. But if you want you robot to crash then go ahead.

bren

[DerekZahn]

I played with it a few months ago and couldn't make any sense out of what they were even trying to accomplish, much less how one would use it to implement a real robot controller.

Beyond that I was attracted by the possibility of building a physics simulation for researching walking gaits but came to the conclusion that it would take a year to get any fruitful progress on a project like that since it's so complicated. I imagine somebody will get a RoboNova or other product into the simulation environment and we'll see how accurate the simulator ends up being. I imagine that the physics acceleration cards will help somewhat but it may be that small inaccuracies can add up to large behavioral differences (like whether it falls over or just wobbles).

I'm completely microsoft-neutral myself but didn't find their studio to be useful enough for my needs to invest the large amount of time required to adopt it as a development tool.

[smile_nik]

I played with it a few months ago and couldn't make any sense out of what they were even trying to accomplish....

The same picture here

[PaulP]

I have tried this with Lego Midstorms and it approaches programming from a different perspective. Rather than compile code and download to the robot it is intended to provide a common control language that is pc based and the commands are then send to the robot in realtime to be carried out. Essentially its moving the brains from the robot to the pc.

This requires a 'manifest' to be written for each robot type, that implements a service or protocol interface to the robot design in question. Lego did a similar thing a while back with mindstorms called spirit. This was in fact a software interface that allowed realtime VB code to send commands to the Lego Robot.

The advantage of this method is you have a common language that works across all P.C.s and all the computing power of your p.c. to drive it. The disadvantage is that you are tethered either physically or wirelessly to the controlling p.c.

In my mind this degrades the autonomous robot by making it a slave of your p.c.

The one really impressive thing I found was the simulations. Build a virtual version of your robot and play. No battery life to worry about, No problems if it falls top to bottom of the stairs, no disastrous burning smells of pinched wires...

[pepperm]

Microsoft Robotics Studio is out now. V1.0 was released in December (see here http://msdn.microsoft.com/robotics/getstarted/v1_0/default.aspx) and guess what....It's free, unless you plan to make money from it....which is good for us isn't it and only fair to Microsoft if you start profiting from their work?

I know nothing about Visual Studio but you are dam sure I will be trying to learn quick because I think the robotics scene is going mainstream very rapidly and I think it will be a good wave for us "early adoptors/geeks" to ride on. Short of maybe the Parallax Propeller the processing power in out robots these days is prety pathetic so connecting the power of the PC platform has got to be good. Couple that with the standard software interface that MSRS provides I believe things can only get better.

There are already some connectors (comms code on the PC and robot) available for cheap robots such as the Lego NXT, BoeBOT and KHR so lets get some more written for the Bioloid (just think of the flexibility there), the Robonova and the likes. Then Standard code on the PC will abstract all the higher stuff to the PC and leave the mundane and bespoke servo and sensor code to the robot's controller.

We have a fun few years ahead of us.

Mark

[slashsplat]

Short of maybe the Parallax Propeller the processing power in out robots these days is prety pathetic so connecting the power of the PC platform has got to be good.

I believe that Parallax is officailly hardware-certified with MSRS, but do not know if it is for Stamp or Prop.

Couple that with the standard software interface that MSRS provides I believe things can only get better.

AMEN!

[pepperm]

It's for the BASIC Stamp 2 on the Boe Bot. See here: http://msdn.microsoft.com/robotics/learn/tutorials/setuphdwr/default.aspx and here http://www.parallax.com/detail.asp?product_id=28118. You need the Bluetooth connection though.

To be honest you would just have to write some code that replied to serial commands sent over bluetooth with the appropriate answers and you could use any processor.

I'm trying to work it all out now

Mark

[PaulP]

I've delved a bit into how the lego rcx is controlled and its cleverly done, masking all the serious stuff and leaving you with some simple code objects to play with such as 'bumper' to represent touch sensors.

The involved bit requires the kit manufacturers to partner with MS and produce services and manifests for the robot type. This will implement the communications protocols to the pc and allow the virtual limbs or devices to be mapped for us to play with.

The one problem that springs to mind is that RN and KHR can be configured non-standard i.e. extra devices on ports of the constructors choosing and these may have to be standardised though im sure thats no great shakes.

As pepperm says, it increases the potential 'thinking' power of the kits and could be great fun to do.

I've pondered a bit about video sending and image recognition software, especially motion detection stuff that could allow robots to respond to particular things it sees..

Perhaps along the lines of mimicing movement. Learning by example etc..

Good stuff..

[PaulP]

Just had a wander through MS and found a pointer to the KHR-1 support in the \samples\platforms\kondo directory of the install.

Evidently it doesnt support services as such, just provides the ability to directly set the servos in code.

They are mapped:

Board Channel Assignment
0 1 Left Shoulder
0 2 Left Elbow
0 3 Left Wrist
0 4 unused
0 5 unused
0 6 Head
0 7 Right Shoulder
0 8 Right Elbow
0 9 Right Wrist
0 10 unused
0 11 unused
0 12 unused
1 1 Left Pelvis
1 2 Left Hip
1 3 Left Knee
1 4 Left Ankle
1 5 Left Foot
1 6 unused
1 7 Right Pelvis
1 8 Right Hip
1 9 Right Knee
1 10 Right Ankle
1 11 Right Foot
1 12 unused

This would be as easily achieved, with less of a PC processor overhead in VB.NET using the Kondo API (which has to be installed for MSRB to work anyway)

From this I am hoping that KHR-1HV will map to the same manifest and hopefully the new DOF will be accessible in the unused ones as I am about to order one.

I think that if RN1 has a similar manner of direct control then it would be possible to map it similarly although I dont believe there is an API for RN1 as yet.