Question

Which search algorithm produces useful “anytime” algorithms: good for

response-time critical applications, e.g., games?

Answer 1

LRTA* (full-form- Learning Real-Time A*), is a search algorithm which produces useful "anytime" algorithms. An "anytime" algorithm is an algorithm which is able to find legit solutions even if the execution/process is stopped before end-point.

The longer the algorithm is allowed to work, the better the solution it would find.

A* algorithms are generally notated with the following method-

f(n) = g(n) + h(n)

n- Leaf/ last node.

g(n)- cost

h(n)- It is a heuristic which allows us to find the least expensive path till the end.

In LRTA*, we keep on executing the problem until we reach a certain stop state, or until we receive the globally best result. It can also be termed as real-time DP (dynamic programming), and is often used in response-time critical applications.

Main characteristic of this algorithm is that it keeps on updating the values of heuristics by visiting them many times in a certain period, altering the values to some extent for reaching the globally best result.

Other algorithms which produce "anytime" algorithms are- Dynamic A* algorithm- (DA*) , LPA*, also known as  Lifelong Planning A* algorithm, etc..

LRTA* algorithm is used in games frequently. However, it has some drawbacks-

1) Complex period where DB (database) is established for intermediates states.

2) It requires very high memory due to the above mentioned period.