In this paper, we investigate a multivariable control of air path of a diesel-dual-fuel (DDF) engine. The engine is modified from a CI engine by injecting CNG in intake ports. The engine uses CNG as its primary fuel and diesel as its secondary fuel, mainly for initiation of combustion. The modification is economically attractive because CNG has lower price than diesel and the modification cost is minimal. However, for DDF engine, control of the air path becomes more difficult because the engine now has combined characteristics of the CI and the SI engines. The combined characteristics come from the fact that diesel is still directly injected into cylinders (CI engine) while CNG is injected at the intake ports (SI engine.) In pure CI engine, throttle is normally fully opened for maximum air intake, while EGR valve is actively actuated to obtain low emissions. In pure SI engine, however, throttle is an active actuator, driven by pedal. The air path control of the DDF engine, therefore, is a multivariable problem, where both throttle and EGR valve are actively actuated to obtain desired outputs. Two outputs that are of particular interest are mass air flow (MAF) and manifold air pressure (MAP). They are directly available from existing vehicle sensors and are good indicators of the air path characteristic. Their desired values are normally obtained as fixed maps during engine calibration. We formulate a control problem, having throttle and EGR valve openings as inputs and MAF and MAP as outputs. Because of high level of input-to-output interactions, a fully multivariable controller is preferred over the decentralized control. A multivariable control based on the quantitative feedback theory (QFT) is designed and implemented. The QFT-based controller is attractive because its robustness amount can be quantified. Control design plant model is allowed to have uncertainty within a boundary called plant template. For all uncertainty within a plant template, a controller and a prefilter are designed from loop shaping to enforce several frequency-domain specifications such as tracking, stability, disturbance and noise rejections, and control effort limitation. A 2KD-FTV Toyota CI engine is modified as a DDF engine and is installed in a test cell with an engine dynamometer. A system identification of the air path is performed to obtain a control design model. Both simulation and experimental results on the engine show the effectiveness of the proposed control system in tracking step changes of desired MAF and MAP.