Page 107 - Read Online

Page 14 of 19 Mai et al. Intell Robot 2023;3(4):466-84 I http://dx.doi.org/10.20517/ir.2023.37

Figure 7. The influences of on path length, fitness values, and mean convergence generation.

Table 5. Simulation results in complex terrain environment
AS MMAS Improved ACO by Chen et al. DSACO
Optimal path length 79.9710 79.2324 77.4436 74.1204
Optimal fitness value 90.7962 89.5247 84.5741 83.0249
Number of iterations 421 407 398 383
Running time/s 3.3321 3.7479 4.2152 3.9541

to optimizing the pheromone update mechanism and limiting the pheromone value, the approximate global
optimal path is found.

Figures 12-14 show the best individual fitness trends of the four different algorithms under simple, medium,
complex, and complex environments, respectively. By comparing the simulation results, we found that the tra-
ditional ACO algorithm has a good convergence speed in a simple environment. By comparing the simulation
results, we found that AS and MMAS have an excellent convergence speed in a simple environment. However,
they are prone to get stuck in local optima in a complex environment. Compared with the improved ACO
algorithm by Chen et al., the fitness of DSACO can be reduced to a lower level, indicating that the algorithm
has a better path search capability. At the same time, DSACO can reach a smaller fitness with fewer iterations,
which indicates that the algorithm has a faster convergence speed.

In order to minimize errors, we conducted 30 experiments and calculated the average values of the optimal
path length, the optimal fitness value, the number of iterations required to reach the optimal fitness value, and
the running time for the four different algorithms in complex mountain environments, as shown in Table 5.