An Adaptive Vision-based Interactional Intention Inference System in Human / Robot Coexistence

I n the paper, we propose a vision-based system for adaptively inferring the interactional intention of a person coming close to a robot, which plays an important role i n the succeeding stage of human/robot cooperative operations typically i n production lines. It is naturally supposed that the interactional intention of a person induces her action of approaching to the robot and touchiny it. Therefore, i n our Adaptive Visionbased Interactional Intention Inference System (AVIIIS), the human interactional intention is inferred from the direction i n which the human body is moving and the tmjectory of the human movements. In the process, lrurr~an prlotion trajectory is generated by traces of the head moverrtent and interpolated wit11 a cubic Spline appro:rirnation teclrnique. Hidden Markov Model (HMM) is applied as an adaptive noise reduct ion filter at the stage of infe~.ring the Iturnan interactional intention. The HMM algovith7n with (I stochastic putter-11 prlatclling capability is extended to supply whether or. not a persort has an inte.rltion toward the isobot ut any tirne. Experimental results derrlonstrate the adapkiveness of the infe7~nce systern using tlre ezte*rlded WhIM ulgor.ithm titat filters out 7rtotio~t deviat ion over the trajectory.