Road friction estimation algorithms had been studied for many years because it is very important factor for safety control and fuel efficiency of vehicle. The traditional solutions are hard to adapt in automotive industry because thier performance is not sufficient enough and expensive to implement. Therefore, this paper proposes a road friction estimation algorithm based a trained artificial neural-network which is low cost and robust. The suggested method doesn't need expensive additional sensors such as optical and luminar sensor, also it shows better performance in real car environment compared to other algorithms based vehicle dynamic. In this paper, we would describe this algorithm in detail and analyze the test results evaluated in the real road conditions.