Abstract : 
In crutch-free walking with powered exoskeletons, pilots instinctively engage their upper body to adapt its motion and maintain balance, especially in the absence of lower-limb sensory feedback or external stabilizing aids. These self-balancing efforts, often involving significant head and trunk movement, not only increase physical and cognitive load but also reduce the overall usability of the exoskeleton. This study proposes a human-adaptation-in-the-loop optimization method that minimizes the need for voluntary upper-body adjustments, particularly head movement. This approach aims to enable crutch-free walking by minimizing the pilot's voluntary balancing, achieved through the iterative optimization of ankle joint trajectory based on the modeling of the pilot's head movements and the center of pressure (COP). As a result, the proposed human-adaptation-in-the-loop optimization minimized the instability caused by the pilot's adaptation motion that is not reflected within the human-robot integrated system, enabling continuous walking for people with spinal cord injury (SCI) at a speed of 0.24 m/s without the use of crutches. This demonstrates an effective solution for achieving natural, crutch-free walking in a powered exoskeleton.