The demand for massages is rising daily in the modern day due to population aging. The creation of an intelligent physicaltherapy massage robot is currently a pressing task due to the severe scarcity of qualified masseurs in the field of physical therapy massage. In this paper, we use the back therapeutic massage robot as our research subject, build the system’s framework, and conduct a thorough analysis of the therapeutic massage operation’s task planning technique using customized features. While there are many different types of physical therapy robots on the market, the majority of them are limited to basic massage tasks and do not take the patient’s psychological state into account. They also have issues with basic trajectories. This research proposes a new task planning method for robotic physical therapy massage that uses a depth camera to identify the patient’s back acupoints, which greatly increases the safety and comfort of back massage. To achieve effective massage task planning, the method combines linear interpolation in Cartesian space with fifth-degree polynomial interpolation. The results of simulations and real-world experiments demonstrate that the suggested approach has great generality and operability, fully accounting for patient individual differences and psychological comfort in addition to having smooth and stable qualities.