Abstract: To handle the low consistency existed between the actual conditions and steady-state test of the power-train in tracked vehicles, and to reflect the operational status in real scenarios, a construction method was proposed for the duty cycle of tracked vehicles power-train. Firstly, establishing a power-train model and calculating resistance torque based on driving data, the method was arranged therewith to characterize the variations in road conditions. And then, referencing the road spectrum, a spectral analysis method was put forward based on spatial frequency, and the power spectrum characteristics of the resistance torque in typical regions were extracted. And the variation characteristics of the road resistance were reconstructed by K-means clustering and Markov chain to generate new duty cycles. Finally, the constructed load cycle was verified based on an integrated power-train test bench. Comparing the actual vehicle test data with duty cycle 1 test data, the errors of the joint probability distribution of engine speed-fuel and vehicle speed-gear are 1.5% and 1.2% respectively. Adjusting penalty factor, the generated duty cycles can carry out the bench test for vehicle speed, fuel consumption, engine speed, and so on, indicating the applicability of the proposed method.