This paper introduces a new superquadric-based human body modeling technique. The model is used as part of an on-line path planning scheme. The path planning scheme utilizes a previously proposed danger evaluation metric in which danger is characterized based on human and nonhuman factors. A new factor that accounts for the human body orientation is introduced and used along with other factors for danger evaluation. A superquadric model of the human is used to determine the values of the factors used for danger evaluation including body orientation. The resulting danger value is then used to direct the search for an alternative robot path in a direction that minimizes the danger. The use of superquadric-based human model for danger evaluation and subsequently path planning provides an accurate and computationally efficient solution. At the same time, the resulting solution guarantees a safe and danger-free path, given the factors used to characterize the danger. The approach exhibits adequate speed of decision making, rendering it potentially suitable for real-time applications involving human-robot interaction. The proposed method is evaluated using a CRS-F3 industrial manipulator through various case studies.