Fig. 6: Performance Trade-offs with Varying C-network Channel Numbers. | Nature Communications

Fig. 6: Performance Trade-offs with Varying C-network Channel Numbers.

From: Optimization and control of actuator networks in variable geometry truss systems using genetic algorithms

Fig. 6

a C-network assignments for the quadruped robot with 2, 8, 16, 32, or 64 C-network channels at the 1000th iteration. b Performance metrics for a quadruped robot for four target objectives, including walking distance, turning alignment, tilting alignment, and crouching distance. c Box plots showing hypervolume results for designs with 2, 8, 16, 32, and 64 C-networks (n = 6 each). Each box shows the median (solid line) and mean (dahsed line), first and third quartiles (box boundaries), and data range (whiskers). Individual data points are shown as black crosses ( × ). Horizontal mean lines are displayed within each box. Statistical significance brackets show p-values from Tukey’s HSD post-hoc comparisons. ANOVA testing shows significant differences between groups (p < 0.001), with post-hoc Tukey’s HSD revealing significant improvement from 2 to 8 C-networks (p = 0.003) but no significant differences beyond 16 C-networks (p > 0.877). d Convergence of Pareto front hypervolume across iterations for designs with different numbers of C-networks.

Back to article page