RAG Grounded Agentic AI for Green Vehicle Routing: A Systematic Literature Review and Case Study at CV RR Jaya Transindo

Abstract

The convergence of Agentic AI and Retrieval-Augmented Generation (RAG) presents
significant opportunities for logistics optimization, yet their integration into Green Vehicle
Routing Problem (GVRP) frameworks remains underexplored, particularly in SME
contexts within developing economies. This study conducts a Systematic Literature
Review (SLR) through a structured search, screening, and thematic synthesis process,
synthesizing 43 articles published between 2017 and 2026 across four thematic domains:
intelligent agent architectures, retrieval-augmented reasoning, green routing
optimization, and AI-driven logistics decision support. A prototype was subsequently
developed and validated at CV RR Jaya Transindo, a livestock feed distribution company
in East Java, Indonesia. The prototype implemented a multi-agent pipeline on Stack AI,
integrating Claude Opus 4 for RAG-based location extraction, OpenRouteService API for
GVRP optimization under HGV constraints, and GPT-5.1 for managerial report synthesis.
Black Box Testing confirmed 100% functional effectiveness. Quantitative evaluation on 35
paired Delivery Orders using Paired Sample T-Test revealed a statistically significant
travel time reduction of 24.06 minutes per delivery (t=4.467, p<0.001), with non
significant differences in distance, fuel consumption, CO₂ emissions, and cost consistent
with GVRP multi-objective trade-off theory.