Single-inductor multiple-output (SIMO) dc-dc converter is a preferred candidate to generate multiple regulated outputs because of reduced circuit volume and avoided coupling of the inductors. Because one common inductor is shared by all output branches, cross regulation (CR) is an important issue, which will result in degraded output performance and undesired voltage variations. As shown in Fig. 1, the load variation of output V1 leads to serious voltage fluctuation of output V2 because the duty cycle of switches changes slowly under conventional Proportional-Integral (PI) control. To improve the transient response speed, the inner loop control [1, 2] is applied in ripple control technique. However, the implementation complexity is increased. Moreover, conventional PI control is unable to deal with functional interdependency among its output branches. From the perspective of decoupling, a small-signal modeling approach based on averaging the inductor current ripple is developed to propose a cross derivative state feedback controller, which decouples outputs in frequency domain [3]. But the self-loop and cross-loop compensation of this controller is designed around a specific operating point. Therefore, the performance may be influenced by the variations in the input voltage and load. To consider a number of operating points, a digital multivariable controller design methodology [4] is proposed. But multivariable control has the limitation of linear regulation based on PI control method. To overcome the defect of linear regulation, comparator control [5] technique exploits high-speed response of comparator to suppress CR. But cross-regulation problem will focus on the last branch. Considering voltage fluctuation of all outputs, model predictive control [6–8] takes the output with the largest deviation from the reference voltage as the priority. It uses algorithm to optimize control duty cycle for next several periods. However, the duty cycle of output switch is either 0 or 1. The limited duty cycle results in large ripple and affects the effect of restraining CR. Due to the above limitations on duty cycle, the CR limit is still unclear. To explore the CR limit, this paper utilizes differential evolution (DE) algorithm to solve the corresponding duty cycle that varies freely in transient process.