algorithm is employed with 30 particles and 200 iterations.
To verify the advantages of utilizing motion primitives in
the planning process, the performance of RRT-MP is compared
with the RRT-CPG [45] and RRT-K [47] methods. In RRT-CPG, the
actuators are also driven by sin CPGs. However, their parameters
are not predefined as in RRT-MP, but are generated randomly in
each expansion step. In the RRT-K approach, the actuators are
not controlled by a periodic signal. Instead, they are moved to
a fixed angle ai, which is generated randomly in each expansion
step within the range (−π/2, π/2). To enable a time comparison
of all the three planners, the complexity of their expansion steps
should be equivalent. In RRT-MP, the complexity is determined
by the number of utilized motion primitives. Therefore, RRT-CPG
generates four random CPG settings in each iteration and RRT-K
generates four random vectors of desired angles in each expansion
step. The maximum number of planning iterations is k = 5000 for
all algorithms.