Existing Machine Translation approaches frequently neglect the necessity for translations that are cognitively accessible to children and individuals with limited cognitive or language abilities, such as those with autism. Traditional methods that combine standard translation with post-editing text simplification are prone to amplifying semantic biases and incur additional computational costs. In this work, we propose a unified framework for accessible machine translation that directly generates simplified translations by regulating lexical complexity during decoding through quantitative complexity metrics. Experimental results demonstrate that our approach not only improves translation quality but also enhances comprehensibility.