Efficient parallel algorithm for robot inverse dynamics computation

Dynamic Analysis of Robot Manipulators

“…However, this approach leads to algorithms which are applicable to particular classes of manipulators with two and three degrees of freedom only. Finally, to facilitate real-time implementation of advanced robot control strategies, parallel processing techniques have been used [56][57][58][59][60] to implement many of the existing algorithms which compute inverse dynamics.…”

A multi-microprocessor-based decentralized adaptive controller for a robotic manipulator

“…This design approach has been utilized by several investigators [6][7][8][9]. The other way to enhance the computational efficiency of dynamic control is to exploit the parallelism of dynamic control algorithms [8][9][10][11][12][13][14]. There exists an obvious parallel in Lagrangian dynamic equations as the computations of all the terms in the equations are independent of one another [9].…”

“…There exists an obvious parallel in Lagrangian dynamic equations as the computations of all the terms in the equations are independent of one another [9]. Many researchers have investigated the recursive Newton-Euler dynamics and have proposed parallel computational algorithms [11][12][13]. However, the parallel efficiency of those algorithms is not very high.…”

Simplification of manipulator dynamic model for nonlinear decoupled control

“…This computation is equivalent to the robot inverse dynamics computation [6]. Since various parallel algorithms have been developed to compute the robot inverse dynamics based on the Newtoh-Euler equations of motion [10], the efficiency of the simplified dynamic model(s) can be gauged by comparing the required number of mathematical operations in terms of multiplication and addition with those stated in [10]. Table 5 compares the time complexity of calculating (75) on a uniprocessor computer using simplified models with the parallel computation of Newton-Euler equations of motion on a multiprocessor system [10].…”

“…Since various parallel algorithms have been developed to compute the robot inverse dynamics based on the Newtoh-Euler equations of motion [10], the efficiency of the simplified dynamic model(s) can be gauged by comparing the required number of mathematical operations in terms of multiplication and addition with those stated in [10]. Table 5 compares the time complexity of calculating (75) on a uniprocessor computer using simplified models with the parallel computation of Newton-Euler equations of motion on a multiprocessor system [10]. Table 5 shows that the computation of simplified dynamic models on a uniprocessor has about the same amount of computation as the parallel algorithms on a multiprocessor system with six microprocessors [10].…”