In this paper, we discuss the sensor allocation problem in detecting obstacles in robot manipulators. The detection of obstacles in a work area is important for safety purposes and for the efficiency of robot control. Therefore, it is necessary to allocate the sensors properly on the links in a robot manipulator. Here, we propose two types of effective sensor allocation methods. One is based on the joint coordinates of the robot, and the other is based on the orthogonal work space. In addition, we show the allocation of additional sensors based on the quantitative conditions of the robot and its obstacles. The optical proximity sensor, which was developed by the authors, is used, and the proposed allocation methods are applied using a SCARA-type robot. It is proved, by experiments on obstacle avoidance control, that effective sensor allocations can be found.