Parallel implementation of Newton-Euler algorithm with one step ahead prediction

The authors propose a parallel Newton-Euler algorithm for calculating the inverse dynamics of manipulators whose computational time is independent of the number of links. The algorithm consists of two paths: a forward path for calculating a kinematic chain, i.e., linear velocities, linear accelerations, etc.; and a backward path for determining a dynamic chain, i.e., forces and moments. In order to determine each moment or force perfectly, the paths must be tracked sequentially. Complete parallel implementation of the Newton-Euler algorithm with data buffering is proposed for constructing a control scheme whose control period can be reduced drastically and is independent of the number of links. Delays due to the data buffering are suppressed by a one-step-ahead prediction using a linear model. The effectiveness of the proposed algorithm is demonstrated by extensive numerical simulations