GenMRP: A Generative Multi-Route Planning Framework for Efficient and Personalized Real-Time Industrial Navigation

TL;DR

GenMRP presents a generative multi-route framework for efficient, personalized real-time industrial navigation by integrating a skeleton-to-capillary sub-network construction with an iterative route-generation process guided by a correctional boosting objective. The framework leverages a Link Cost Model that fuses user context, historical sequences, and rich link features via a DIN-GAT-Multi-Scenario network, and it uses a Bidirectional Dijkstra routing module to produce routes within a dynamically updated sub-graph. Key innovations include the Link Memory mechanism and a correctional boosting loss that balance route quality and diversity across iterations, plus computational efficiency strategies like STC and incremental online inference. The authors release PRN, a large request-level road-network dataset, and demonstrate superior offline and online performance, including deployment in a real-world navigation app, underscoring GenMRP’s practical impact on personalized, diverse, and scalable routing.

Abstract

Existing industrial-scale navigation applications contend with massive road networks, typically employing two main categories of approaches for route planning. The first relies on precomputed road costs for optimal routing and heuristic algorithms for generating alternatives, while the second, generative methods, has recently gained significant attention. However, the former struggles with personalization and route diversity, while the latter fails to meet the efficiency requirements of large-scale real-time scenarios. To address these limitations, we propose GenMRP, a generative framework for multi-route planning. To ensure generation efficiency, GenMRP first introduces a skeleton-to-capillary approach that dynamically constructs a relevant sub-network significantly smaller than the full road network. Within this sub-network, routes are generated iteratively. The first iteration identifies the optimal route, while the subsequent ones generate alternatives that balance quality and diversity using the newly proposed correctional boosting approach. Each iteration incorporates road features, user historical sequences, and previously generated routes into a Link Cost Model to update road costs, followed by route generation using the Dijkstra algorithm. Extensive experiments show that GenMRP achieves state-of-the-art performance with high efficiency in both offline and online environments. To facilitate further research, we have publicly released the training and evaluation dataset. GenMRP has been fully deployed in a real-world navigation app, demonstrating its effectiveness and benefits.

Figures (6)

• Figure 1: Personalization and diversity are essential for route planning.
• Figure 2: A real-world road network and its corresponding abstracted graph and dual-graph.
• Figure 3: The framework of GenMRP and the Link Cost Model (LCM). The link set are generated based on user requests and the skeleton-to-capillary method. Training involves route sampling, where the link set and previously generated routes are input in each iteration. For inference, the trained link cost model calculates link costs in each iteration, and bidirectional Dijkstra is applied for routing. LCM is designed to compute link costs through its user preference capture module, link representation module, and multi-scenario network. A detailed description of the input features is provided in \ref{['tab: key features']}.
• Figure 4: Skeleton-to-capillary (STC).
• Figure 5: Diversity analysis.