Cooperative movements through hierarchical database search

The paper tackles the problem of synthesizing robot movements for human robot collaboration. The proposed approach employs a dual hierarchical database, which encodes multiple demonstrated human-robot collaborative movements. The primary database encodes demonstrated human movements and is enhanced with a directed weighted graph. It is used for human movement recognition. After recognition, the secondary database, encoding corresponding robot demonstrations, is used to synthesize appropriate collaborative movement. The proposed approach is evaluated through comparison to Interactive Primitives, a popular approach for synthesizing human robot collaborative tasks. Different sets from a database of two-dimensional human movements are used as example sets for evaluation.