Deictic State as Well

If we are willing to include a behavior with just one bit of variable state in it, then we can simplify the control state for the program. In the behavior deictic-avoid, the bit hit-left? serves as a deictic variable the-side-I-just-hit-on. Avoid-hit and compensate-avoid (primitives deictic-avoid supports) turn in the appropriate direction by accessing this variable. This allows a reduction in redundancy in the plan, including the elimination of one of the action patterns.

@1 &&&&&*+[F]deictic-avoid
hit-left? [lllll]_-avoid-hit, feeler-hit,^-compensate-avoid &&&[lll]^-feeler info

$$\mbox{{\bf walk} } \Rightarrow \left\uparrow \left\langle \b... ...x{ walk-straight}\nonumber \end{array} \right\rangle \right.$$

$$\mbox{{\bf avoid-obstacle} } \Rightarrow \langle \mbox{walk ... ... \mbox{avoid hit} \rightarrow \mbox{compensate avoid} \rangle$$ (4)