Research Gaps in AI-Supported Disassembly Sequence Planning in Remanufacturing

Abstract

Artificial intelligence has become increasingly important in research on Disassembly Sequence Planning (DSP). AI-based methods can now solve more complex problems and consider several economic, environmental and operational objectives. However, most studies still focus on optimizing the disassembly process after the recovery strategy and target components have already been selected. This study reviews AI-supported disassembly planning in remanufacturing and identifies the main research gaps and future research directions. The review combines bibliometric mapping of 241 publications with a qualitative analysis of 35 selected studies. Of these studies, 23 directly address DSP, seven concern related disassembly problems, and five focus on remanufacturing planning or scheduling outside DSP. The results show progress in metaheuristic optimization, reinforcement learning, digital twins, knowledge graphs, computer vision, large language models and robotic planning. At the same time, the analysis identified several recurring limitations that can be grouped into three broader dimensions: product-state knowledge and adaptability, validation and transferability, and decision scope and strategic-operational integration. The reviewed studies show that uncertainty, adaptive planning, digital twins and multi-objective optimization have already received considerable attention, but their application remains limited by predefined product representations, case-specific validation and weak integration between product-level recovery assessment and operational disassembly planning. The findings suggest that future research should complement further algorithmic development with more integrated, adaptive and knowledge-supported decision-support approaches that connect recovery assessment with the planning of the required disassembly process.