title
How can we design a more intelligent path-planning?


intentional topic is
Is Artificial Intelligence (AI) really intelligent?

I doubt.
It might be more efficient and/or effective than human
but less intelligent


human intelligence is  flexible

behaviors differ  even in a identical two situations


google
what is intelligence => definition?

Goal is to make an agent behave differently
even when it encounters the same situation as before

Pardon?


Path Planning as a benchmark
from start to goal with limited fuel

to simply put it is usually minimization problem

let's consider by GA 

interpretation
evolution in an agent's brain (100 chromosome of up down right left) -> ga ->  convergence -> follow the chromosome


Fig => looks easy 
and everytime different route might be occur seems good
but ...  what we want
different behavior even if with exactly the same randome seed

let's dismiss GA

by NN for example

biannually we organize
ICNNAI neural network and AI

we've never seen intelligent performance of neural networks

fixed weight => identical behaviour in identical situation following same random seed


evolve weight  during behaviour

muchallo-pits hebb -> spiking nn STDN

architecture as an example

FIG -> no learning

might be not so difficult to find weight configuration that perform 
but


still not successful

motivation

for example
intelligent trafic signal


----------------------------------------
maximization from goal to goal infinite number of solutions
everytime maximum loop but different