Robotics and subsumption architecture
ImageCraft
Posts: 348
Subsumption architecture is a "new" (well, 20+ years old) way of thinking about designing intelligent systems. I am designing V2 of our subsumption architecture based OS, REXIS, and Propeller, with the multiple Cogs (and potentially stacking multiple Cogs together as suggested by Bob R.) is an ideal environment for REXIS. If you are interested, please take a look at our company's blog. I welcome any suggestions and comments here or in the blog.
http://imagecraft.wordpress.com
// richard
http://imagecraft.wordpress.com
// richard
Comments
I am fully awake now and took a look...I don't have the background to comment.
how would sensor fusion be handled?
Rich
You can handle it different ways, depending on whether the sensor effects are equally weighted or not. Lets say there are two sensors and they affect the same actuator.
Scenario One: Sensors have same priority:
A single behavior reads both sensor and acts accordingly.
Scenario Two: Sensor Two has higher priority
Behavior One handles sensor #1 and change the actuator
Behavior Two handles sensor #2 and suppresses the output from Behavior #1
etc.
in the 1991-1995 period. The company I worked for was NDC Automation, with my travel shared
between their US, Swedish and Australian offices.
These robots were laser-guided using a proprietary 360 degree scanning
head to perform triangulation with fixed reflector beacons.
As a result, I became acutely aware of Rodney Brooks' research, among others investigating subsumtive
architectures. I coded the micontroller and pc systems responsible for the vehicles guidance, and designed
the hardware for the communication links and RS422 interfaces.
This article definitely piques my interest. Do you have any more to share?
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
JMH
Post Edited (Quantum) : 8/27/2008 12:55:47 AM GMT
Not currently. All that is fit to print is on the blog entry. REXIS (V1) was demonstrated on a robot (named Ripley) converted from a RC car. I am not terribly good with hardware and my friend took out the steering mechanism and replaced it with a futuba drive. For sensor, we just have a polaroid sonar module mounted on a servo driven rotating platform. At my thesis presentation, Ripley performed perfectly by navigating out of the room, down the hallway and out to the parking lot.
What would be really neat is to get hexapod or a quadrapod, and re-design and implement something similar to what Genghis used. It was phenomenal to see it walk, using the subsumption program.
As for REXIS.MkII, I need to gauge demand for it first. As mentioned, Propeller is actually a great fit for the architecture. The language processor is relatively simple but not trivial and of course the API won't be easy to get right and tight. Lots of challenges, but they are good challenges.
i am verry interesting in this stuff from rodney brooks and his
<Subsumption architecture on ghenghis
i am work about 2 years on a quadrapod, with 4 ServoLegs with 3DOF
hardware (2 propellerProtoBoards, 3 PING-Distance-Sensors, 1 PIR and 1
memsic2125 3axis sensors.
for walking i hope can make this with walknet (stence and swing legs) and
the Subsumption architecture for the sensor fusioning and AI (avoid obstacles, look for
peoples (PIR) pathPlaning.
now i make a look at your blog.
ich somebody here hints and tips .......
excuse my bad english
regards
nomad
You said.
2 sensor´s. Why not 3 and fault tolerant system
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Nothing is impossible, there are only different degrees of difficulty.
Sapieha
Fault tolerance is one of the things at which this architecture excels.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
JMH
Yes, but!
2 it is to low to fault test.
3 and more have more precision. If 1 of 3 fault it is more precise why fault
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Nothing is impossible, there are only different degrees of difficulty.
Sapieha
I must be thick-headed this morning.
Could you clarify or explain using an example?
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
JMH
My point was.
In fault tolerant systems number on same sensor´s must be 3 or more to decide fault on one sensor.
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Nothing is impossible, there are only different degrees of difficulty.
Sapieha
▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Cheers,
Simon
www.norfolkhelicopterclub.co.uk
You'll always have as many take-offs as landings, the trick is to be sure you can take-off again ;-)
BTW: I type as I'm thinking, so please don't take any offense at my writing style