With the trend of transportation electrification, driving cycles have been widely recognized as effective tools to tackle the challenges of the optimal design, management and evaluation of electric vehicles. In this work, real-world driving data recorded on 1 Hz of 40 electric taxis in Beijing area for 6 months are obtained and fused with road environment information to construct driving cycles tailored for electric vehicles. The conventional Micro-trip method is improved based on minimum comprehensive parameters deviation, which achieve better accuracy with less computational load. A novel improved Markov Monte Carlo method considering the driving features on different roads is proposed to reflect the features of road environment in the driving cycles. 53 parameters including characteristic and distribution parameters are extracted from driving data and used to comprehensively describe the features of driving process, in which the road environment and energy related parameters are also included. Based on Mean absolute percentage error and K-S test, the performances of the proposed methods have been investigated, and the constructed driving cycles as well as NEDC are verified and compared to real-world driving condition.