The report summarizes the approach and findings in a study on "Model Based Recognition and Learning of the Characteristics of Objects by Sensory Fusion in an Unknown Environment". As a first step ambiguities on the collected information due to the ultrasonic device uncertainities and the noise is reduced by preprocessing of the image. Information obtained from multiple sensors when fused reduces the inherited device uncertainly and noise in the final image so that information lacking in the image may be compensated by that supplied from another sensor. Following this phase algorithms related to object recognition are applied on the preprocessed image. The object recognition is based on the pattern of outer boundary of the object whose inner surface area taken out. The outer boundary, the path length of the boundary, center of gravity of the object, and the first and the second moments of the object are compared to match with those of the model in.recognition, of the object. Any object characteristics not matching with those of the model is identified as a new object which is to be added into the library. The system with this feature is considered to be a system having the ability of learning.In the hardware structure an array of 8 pairs of ultrasonic transducers are placed on the wrist of a robot arm for scanning the same involving the objects to be recognized. The robot arm recognizes the objects, determines their positions and. the orientations of the objects and then, plans its trajectory, preshapes and plans its fine motion for grasping in. accordance with this data.