The process is Markov, so we can condition on the next car passing. The time to cross the road is 20 seconds if no cars pass for 20 seconds. Otherwise, the expected time to cross is the time the car passes plus the original expectation. The time a car passes is distributed as Exp(lambda = 1/10).
(E[T]=\int_0^{20}{(t+E[T])dP(\tau=t)}+\int_{20}^{\infty}{20dP(\tau=t)})
(E[T]=\int_0^{20}{t\lambda e^{-\lambda t}dt}+E[T]\int_0^{20}{dP(\tau=t)}+20\int_{20}^{\infty}{dP(\tau=t)})
(E[T]=10-20e^{-2}-10e^{-2}+E[T](1-e^{-2})+20e^{-2})
(E[T]=10(e^2-1)=63.89)