Skip to content

Latest commit

 

History

History
33 lines (29 loc) · 1.55 KB

Strategy.md

File metadata and controls

33 lines (29 loc) · 1.55 KB

Slim strategy description

This file goal is to define the behavior of the robot during a combat.

Inputs:

  1. Motor measured speed left and right (RPM)
  2. Distance sensor left and right (active / not active)
  3. Euler angles defined orientation (deg)
  4. Orientation in quaternions (quaternions)
  5. Acceleration of the robot in 3D (m / s^2)
  6. Angular speed of the robot (deg / s^2)
  7. Tire traction available, left and right (available / not available)
  8. TODO

Outputs:

  1. Motor set point speed, left and right (RPM)
  2. TODO

States:

  1. Initial state - the moment after the start signal was applied.
  • General approach is to start going trough the terrain in a pattern, pattern is TODO
  1. Contact waiting state
  • We need to detect contact event (TODO: how to distinguish contact).
  • We need to control position changing using the pattern decided in the previous state (TODO: maybe PID control of path travelled by every tire).
  1. Contact detected state
  • Once contact is detected we need to decide action (TODO).
  • Different contacts must have different actions (TODO).
  1. Action tracking state
  • Once the opponent position is known and the counter action is chosen we need to make sure the action is done properly.
  • Here we need to watch out for disposition problems like lifting us up while trying to push the opponent on the front.
  1. OPTIONAL: Action failure state
  • The action has failed in execution (TODO: definition of action failure).
  • New action is to be chosen and passed to execution.