Request for Comments: Next Iteration of Behavior Simulator

[knoellle]

New stuff

• Think about where to draw the line between simulator and scenario file/lua library.
  • Specifically regarding application of rules, game logic and physics. Most of this currently either isn't implemented or done by the simulator.
• Fix penalty handling
  • currently both the robots and the game controller state have a penalty state, both of which are used in different locations

Leftovers from #91:

Important

• Limit robot's field of view & view range Behavior simulator FOV #342
• Simulate node inputs that are not currently covered
  • obstacles (use actual node?)
  • penalty shot direction (use actual node?)
  • has ground contact
  • fall state
• HULK standard library in Lua
• Build & Run simulator in CI

Nice to have

• Rerun simulation when
  • scenario file changes
  • changing parameters via communication
• Ball "physics"
• Add configurable noise to
  • ball position
• message delay/jitter/dropping
• Apply SPL rules
• Automatically generate behavior cycler code. Probably after the wheels on the bus go round and round #143
• Improve lua interface, maybe have a look at lenscas/tealr

[h3ndrk]

Regarding architecture:

• Each robot should provide one communication server
• Time controlling should be separate from communication connections

From that follows that Twix needs to be able to display multiple robots which would also help with real NAOs.

[knoellle]

Agreed, but those are probably candidates for a behavior simulator 2.0 and twix 2.0.

[h3ndrk]

As proposed by Patrick, we should have a look in the simulation leagues before implementing our own simulator.