Sep 2026· International Conference on Intelligent Transportation Systems and Automation Control· Vol 14368, pp. 143681X - 143681X-13· 0 citations
Engineering
TL;DR
A straightforward Deep Reinforcement Learning baseline framework that pairs a Double Deep Q-Network (Double DQN) with a deterministic pathfinding module to provide a simulation-based DRL baseline for future comparative studies in intelligent tourism mobility.
Abstract
Planning itineraries for urban tourist buses involves a critical trade-off between maximising attraction coverage and strictly adhering to operational constraints. While recent literature has introduced highly complex optimisation models, there remains a notable lack of a transparent, reproducible baseline tailored specifically for this domain. To address this gap, this paper presents a straightforward Deep Reinforcement Learning (DRL) baseline framework that pairs a Double Deep Q-Network (Double DQN) with a deterministic pathfinding module. Operating solely on fundamental static inputs, such as a bus-feasible road graph, points of interest (POIs), and fixed time budgets, the model generates structurally feasible routes within a simulated environment without requiring complex dynamic data. The framework is evaluated through a simulation-based case study in Ha Noi City, simulating both morning and afternoon shifts under a four-hour service constraint. Empirical results indicate that the Double DQN architecture exhibits improved training stability, yielding ~13% improvement in POI coverage compared to a Standard DQN baseline across different times of day. Also, the proposed model achieves ~90% time-budget utilisation rate, improving upon ~79% observed in standard methods, while converging roughly twice as fast. These findings provide a simulation-based DRL baseline for future comparative studies in intelligent tourism mobility.
The proposed hierarchical multi-agent proximal policy optimization framework can reduce total airlines' operational costs—including direct operating cost and capital cost and achieves a computation speedup in comparison with a conventional optimization baseline.
Li-Jing Liu, James M. Shihua, Qi-Yu Yan et al.· MATEC Web of Conferences· 0 citations
A deep architecture that is able to jointly optimize cost functions and route-ranking model towards any route preference is proposed, and a novel loss function is proposed that optimizes a single-objective variable, with other variables strictly under constraints.
Rui-Zhao-Min-,-Ning-Bo-Wen-,-Li-Si-Yao Zhao, Chao Chen, Long-Fei Xu et al.· 0 citations
It is demonstrated that integrating deep spatiotemporal prior learning with adaptive evolutionary optimization can effectively reduce the computational burden of mission replanning while maintaining responsiveness, stability, and scalability, providing a robust approach to resilient multi-robot management in unpredicta...
Xin-Yi-Gao-Yong Zhang, Xin-Qi Li, Wen-Bo Li· Frontiers in Robotics and AI· 0 citations
Urban transportation networks present complex optimization challenges spanning calibration of high-fidelity simulators and real-time operational control. This paper presents a shared latent-space framework that connects simulator calibration and reinforcement learning control through a common learned representation of...
Adewumi Augustine Adepitan, Christopher J. Haruna, Oluwasegun O. Adegoke et al.· 0 citations
Diffusion trajectory planners with domain-specific experts face a practical question in unseen regions: can they rely on selecting among existing experts, or must the planner itself adapt? We study this question using a planner with a shared adapter and per-domain expert adapters. Across three real-world held-out maps,...
Yun Li, Ehsan Javanmardi, Simon Thompson et al.· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.