Friday, 26 March 2010

On centralized control

Now, at the beginning of the project, I have to decide if I use a "emulated" local control or real local control.
By emulated I mean: Use a computer to send a low-level command (like "move foward" and so on) and emulate the local capabilities and knowledge of each agent on software or...
Real local control: Each robot have the complete control of it's action based on real local knowledge and processing.

I started creating a table of pros and cons of each approach to help me:


  • Much simpler robots.
  • Much Cheaper robots (an important consideration, since I'm paying for them).
  • It's easier to convert the project to an academic competition (like RoboCup) since the architecture is closer.
  • more... (help me to think about this)
  • Expend lot of time in Computer Vision, concurrence management, communication protocols, routing and other unrelated complex software.
  • Not real a swarm, since if the "brain" computer dies all agents stop...
  • Have to simulate sensors response.
  • more... (help me to think about this)
Real local:

  • True local processing.
  • Real swarm, robust in case o failure of agents.
  • Focus on behavioral algorithms instead of a simulation platform.
  • Real challenge for my MCU programming skills (and eLua's power =) ).
  • more... (help me to think about this).
  • More complex Hardware (electronics mainly).
  • Expensive agents.
  • More sensible to project errors (mostly a consequence of the complexity).
  • Turning away from other academic activities involving the built robots (again, something like RoboCup).
  • more... (help me to think about this)

No comments:

Post a Comment