How to understand pushdown automata intuitively?

Question

What is an intuitive way of understanding what a push down automaton is capable of computing?

Answer 1

Intuitively, a pushdown automaton uses a stack and using a stack we can do a depth first traversal of a parse tree. It means that we can accept strings which are in context free languages by using stack of a pushdown automaton (left-most derivation). This is not a rigorous proof that languages of pushdown automata are context free. For proofs you must see various textbooks on the subject.

Answer 2

Let's start with a finite automaton. Such a device can be viewed as a machine that has a fixed, finite number of boolean variables, like $s = $ "we've seen an even number of $0$'s in the input so far", or $s_i = $ "the input so far, when interpreted as a binary number, is congruent to $i\pmod 3$". Using such a machine, we can count, but only up to a fixed, finite value.

A PDA can be viewed as a FA with one other variable, a stack. With such a machine, we can count up to an arbitrary number. This means that while the language $$ L=\{a^nb^n\mid n\ge 0\} $$ can't be accepted by any FA, it can be recognized by a PDA: for each $a$ we see in the input, we push a marker on the stack and for each $b$ we pop a marker off the stack. When we've seen all the input we stop and accept if the stack is empty and reject if it's not.

In a similar way, we can use the stack to store patterns in our input, which is why the palindrome language $$ P=\{wcw^R\mid w\in \{a,b\}^*\} $$ can similarly be recognized by a PDA: we push the characters seen on the stack until we see a $c$ and then match the rest of the inputs against the contents of the stack. Since the stack will contain the characters in $w$, we can match the reverse, $w^R$ part of the input against the contents of the stack.