Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

SmartFlow Reinforcement Learning and Agentic AI for Bike-Sharing Optimisation

Aditya Sreevatsa K, Arun Kumar Raveendran, Jesrael K Mani, Prakash G Shigli, Rajkumar Rangadore, Narayana Darapaneni, Anwesh Reddy Paduri

TL;DR

SmartFlow presents a hybrid framework that merges deep reinforcement learning with an agentic AI layer to address dynamic bike rebalancing in urban networks. A DQN-based strategic agent learns high-level redistribution policies in a high-fidelity Citi Bike simulator, while a deterministic tactical planner assembles efficient multi-leg routes and just-in-time dispatches. An LLM-powered agentic AI translates the plan into clear, actionable dispatch instructions for operators, improving interpretability and execution readiness. Across three seeded runs on Citi Bike data, SmartFlow achieves an average imbalance reduction of about 95% with low travel distance and high truck utilisation, offering a scalable and transparent blueprint for AI-assisted urban mobility logistics.

Abstract

SmartFlow is a multi-layered framework that integrates Reinforcement Learning and Agentic AI to address the dynamic rebalancing problem in urban bike-sharing services. Its architecture separates strategic, tactical, and communication functions for clarity and scalability. At the strategic level, a Deep Q-Network (DQN) agent, trained in a high-fidelity simulation of New Yorks Citi Bike network, learns robust rebalancing policies by modelling the challenge as a Markov Decision Process. These high-level strategies feed into a deterministic tactical module that optimises multi-leg journeys and schedules just-in-time dispatches to minimise fleet travel. Evaluation across multiple seeded runs demonstrates SmartFlows high efficacy, reducing network imbalance by over 95% while requiring minimal travel distance and achieving strong truck utilisation. A communication layer, powered by a grounded Agentic AI with a Large Language Model (LLM), translates logistical plans into clear, actionable instructions for operational staff, ensuring interpretability and execution readiness. This integration bridges machine intelligence with human operations, offering a scalable solution that reduces idle time, improves bike availability, and lowers operational costs. SmartFlow provides a blueprint for interpretable, AI-driven logistics in complex urban mobility networks.

SmartFlow Reinforcement Learning and Agentic AI for Bike-Sharing Optimisation

TL;DR

SmartFlow presents a hybrid framework that merges deep reinforcement learning with an agentic AI layer to address dynamic bike rebalancing in urban networks. A DQN-based strategic agent learns high-level redistribution policies in a high-fidelity Citi Bike simulator, while a deterministic tactical planner assembles efficient multi-leg routes and just-in-time dispatches. An LLM-powered agentic AI translates the plan into clear, actionable dispatch instructions for operators, improving interpretability and execution readiness. Across three seeded runs on Citi Bike data, SmartFlow achieves an average imbalance reduction of about 95% with low travel distance and high truck utilisation, offering a scalable and transparent blueprint for AI-assisted urban mobility logistics.

Abstract

SmartFlow is a multi-layered framework that integrates Reinforcement Learning and Agentic AI to address the dynamic rebalancing problem in urban bike-sharing services. Its architecture separates strategic, tactical, and communication functions for clarity and scalability. At the strategic level, a Deep Q-Network (DQN) agent, trained in a high-fidelity simulation of New Yorks Citi Bike network, learns robust rebalancing policies by modelling the challenge as a Markov Decision Process. These high-level strategies feed into a deterministic tactical module that optimises multi-leg journeys and schedules just-in-time dispatches to minimise fleet travel. Evaluation across multiple seeded runs demonstrates SmartFlows high efficacy, reducing network imbalance by over 95% while requiring minimal travel distance and achieving strong truck utilisation. A communication layer, powered by a grounded Agentic AI with a Large Language Model (LLM), translates logistical plans into clear, actionable instructions for operational staff, ensuring interpretability and execution readiness. This integration bridges machine intelligence with human operations, offering a scalable solution that reduces idle time, improves bike availability, and lowers operational costs. SmartFlow provides a blueprint for interpretable, AI-driven logistics in complex urban mobility networks.
Paper Structure (43 sections, 2 equations, 8 figures, 1 table)

This paper contains 43 sections, 2 equations, 8 figures, 1 table.

Table of Contents

  1. Introduction
  2. Related Work
  3. Hybrid Frameworks in Bike Rebalancing
  4. Materials and Methods
  5. Theoretical Framework
  6. Value-Based Reinforcement Learning.
  7. Agentic AI and Grounded Reasoning.
  8. System Architecture
  9. Strategic Core: The Reinforcement Learning Agent.
  10. Tactical Execution: The Operational Planning Module.
  11. Communication Interface: The Agentic AI Layer.
  12. Data Acquisition and The SmartFlow-Prep Pipeline
  13. Reinforcement Learning Environment
  14. State and Action Spaces.
  15. Environment Dynamics and State Transitions.
  16. ...and 28 more sections

Figures (8)

  • Figure 1: System architecture of the SmartFlow framework. A strategic RL agent guides a tactical planning module, whose output is translated into human-readable instructions by an agentic AI layer.
  • Figure 2: Individual learning curves for each of the three seeded runs, showing the moving average reward over time.
  • Figure 3: The aggregated learning curve across three runs, showing the mean moving average reward and standard deviation.
  • Figure 4: Individual task prioritisation plots for each of the three seeded runs.
  • Figure 5: Aggregated density plot of the hours at which the agent identified the need for a transfer, showing a consistent pattern of proactive planning.
  • ...and 3 more figures