This paper develops an adaptive idle speed control strategy for a V2, 1000 cc four-stroke, water-cooled, port injection SI engine. In order to verify the proposed strategy, the non-dimensional engine model including charging and torque dynamics is established in Matlab/Simulink software based on previously experimental verification. The integration of dynamics above will be a multi-input-single-output (MISO) system, which inputs are throttle angle and spark advance angle, and the output is engine speed. The proposed adaptive controller is developed on the model-based structure. The system parameters are updated by recursive least square (RLS) method so the system is able to represent the actual operation. The updated system parameters adjust control gain by derivation of closed-loop gain and pole placement. This research also applies RLS method in torque load identification, and uses feedforward control module to calculate the desired throttle opening for the corresponding engine torque. The feedforward throttle control makes controller less angle of spark advance, therefore, the tracking time could be reduced. In order to verify the proposed adaptive idle speed controller performance, the conventional proportional integral derivative (PID) idle speed controller is established. The results show the proposed feedforward adaptive controller can obviously improve about 40% of tracking time with ramps load. In addition, when the water pump load is added, the proposed strategy can also improve about 20%. Finally, the results of the proposed strategy are also improved comparing to the conventional PID control as deposits happen.