Sumrit D.Apiwatvaravong S.Mahidol University2026-06-012026-06-012026-06-01Transportation Engineering Vol.24 (2026)https://repository.li.mahidol.ac.th/handle/123456789/117029Autonomous delivery robots (ADRs) are widely promoted as a solution to improve efficiency in urban last-mile logistics, yet their large-scale adoption remains limited. This is mainly due to social, regulatory, and infrastructure-related challenges rather than technological constraints. This study develops a structured decision-making framework to analyze ADR implementation barriers and to identify effective multi-stakeholder solutions while accounting for uncertainty and interdependencies in expert judgments. The results show that ADR barriers are highly interconnected. Institutional arrangements and infrastructure conditions act as underlying drivers that influence many other challenges, whereas public acceptance, cybersecurity concerns, and human–robot interaction issues largely emerge as outcomes of weak governance and limited system readiness. Safety compliance and risk liability are identified as the most influential barriers, followed by cybersecurity vulnerabilities and cultural or behavioral constraints. These findings indicate that progress in ADR adoption depends more on clear responsibilities, trust, and social readiness than on further technological development alone. The solution analysis shows that governance- and infrastructure-oriented actions are more effective than isolated technical measures. Ethical guidelines, clear legal frameworks, and stronger industry collaboration are identified as the most effective strategies, highlighting the need for coordinated action among regulators, technology developers, and logistics service providers. Theoretically, the study reinforces the view that ADRs should be understood as part of a broader urban system rather than as standalone technologies. Practically, the findings provide clear guidance for stakeholders seeking to enable responsible and scalable ADR deployment in urban last-mile logistics.EngineeringArticleSCOPUS10.1016/j.treng.2026.1004462-s2.0-1050398526182666691X