Thursday, 27 May 2010

MULTIPLE SENSORS FOR A MOBILE ROBOT

SUBSUMPTION ARCHITECTURE

In a seminal paper in AI which introduced a the paradigm of 'Subsumption Architecture', 'Robust Layered Control System for a Mobile Robot' by Rodney Brooks; Brooks argues that a mobile robot should be equipped with multiple sensors as that adds to the robustness of the control of the robot. The robot can respond efficiently to a dynamic environment thus the behaviours are apparently more 'smoother'.

Fig.1 Control schematics for multiple sensors, from Brooks
To come up with examples to illustrate Brook's theory led me to this simple experiment in Player-Stage.

I. A robot with sonars, STUCK !



II. Introducing lasers, VOILA it is unstuck and on its way





WHAT IS HAPPENING ? 

Here the simulation starts off with a robot equipped with long-ranged sonars, the robot is stuck.

 
Introducing a short-ranged laser (SICK LASER) unstuck the robot.

Obviously, other solutions can be by modifying the code (sonarobstacleavoid.cc) or changing the geometry of the placement of the sonars. However, the idea of this demonstration is to illustrate the robustness gained by the use of two sensors.

REFERENCES
(2) 6 Architectures

2 comments:

Anonymous said...

"Other solutions can be by modifying the code (sonarobstacleavoid)"

You say it yourself...its all about coding. Normally you design your robot according to the environment it is going to act at, but for such occasions, where the robot gets into a stuck, algorithms such as a simple wall follower or a coastal navigation algorithm with more tolerance in the robot-wall distance can lead to unstucking. And don't forget to add randomness somewhere in your code, because closed loops tend to lead to fixed robot paths...

check your mail ;)

Arkapravo said...

Dude, that entry on my blog has to be modified. Have been lazy of late ! ..... by the way, which e-mail are you talking of ? Did you send it to my gmail account ?