This paper aims to promote the development of agricultural intelligence, further enrich the perceptual image of agricultural machinery users, and solve the problem of homogeneity of agricultural robot modeling design. Combining the eye tracking experiment and VR simulation, a modeling design method of agricultural robot is proposed. Firstly, this paper determines the sample and preliminary perceptual image, and conducts the weight analysis and screening of perceptual image. Secondly, it uses the eye tracking experiment to analyze the modeling feature points, and mainly determines the key modeling elements based on the hot spots. Thirdly, according to the image weight analysis and eye movement experiment results, it carries on the targeted modeling image innovation, forms the scheme set and completes the computer-aided design. Finally, it uses entropy weight theory and VR simulation technology to complete the comprehensive design evaluation and obtain the optimal solution. Taking the shape design of apple picking robot as an example, the analysis shows that the application of eye tracking technology and VR simulation evaluation in the shape design of agricultural robot can more accurately innovate the shape design according to the user's perceptual image demands, and improve the design efficiency. In a nutshell, the method has certain applicability.