This paper introduces a polyhedral approximation algorithm for set-valued estimation of switching linear systems. The algorithm generates set-valued estimates for any possible sequence of switching parameters, under the assumption that the system has unknown but bounded disturbances and measurement noises. Our algorithm has practical implications; namely, set-valued estimates were generated for the position and electrical current of a real-time automotive electronic throttle valve, and the corresponding experimental data demonstrate the practical benefits of our approach.