Based on the complexity of global economic dynamics, people often cannot accurately identify the evolution of risk factors. Inevitably, financial modeling is inherently subject to the ambiguity of the model. The Ayersberg paradox believes that people repulse the ambiguity, that is, people are more inclined to choose certain things, and to reject things which are uncertain. A closedloop portfolio optimization method with a constant absolute risk aversion utility function is proposed to solve the problem of risk ambiguity and aversion of investors' doubts about the average yield and volatility. Given a volatility that achieves compact values, using a special uncertainties set to represent drift, using the Karush-Kuhn-Tucher condition of the optimization problem, based on the classical Dyson problem and the max-minimum Hamilton-Jacobi-Berman-Isaacs partial differential equation, the optimal portfolio for ambiguity aversive investors can be solved.