Ankle exoskeletons have recently aroused increasing research interest owing to their potential in enhancing human locomotion. Nevertheless, inter-subject variability makes the control of human–exoskeleton interaction complicated. To handle this problem, we designed a human-in-the-loop (HIL) approach to optimization control for an ankle exoskeleton during walking based on an improved self-adaptive particle swarm optimization (ISAPSO) algorithm and the iterative learning control (ILC) algorithm. As part of the development, a self-adaptive updating strategy was first proposed to tune the three key parameters of each particle to obtain a better trade-off between the global and local search abilities of ISAPSO. Moreover, since the performance of the proposed ISAPSO heavily relies on its convergence property, we provided a convergence-guaranteed parameter setting rule for the proposed optimizer after analytically investigating its convergence. Finally, the developed HIL optimization approach was verified via experimental tests on eight subjects. The experimental results revealed that the proposed method reduced the soleus muscle activities of the eight subjects by 23.46 ± 10.21, 47.04 ± 13.54, 28.52 ± 8.14, and 8.58 ± 3.82% compared with those for the static assistance condition, zero-torque model, normal walking condition, and the case optimized by standard particle swarm optimization, respectively. Thus, the proposed method can be regarded as an alternative in the field of exoskeleton HIL optimization control.