Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Personalized and Context-aware Route Planning for Edge-assisted Vehicles

Dinesh Cyril Selvaraj, Falko Dressler, Carla Fabiana Chiasserini

TL;DR

The paper addresses the need for personalized, context-aware routing for edge-enabled vehicles by integrating graph neural networks with deep reinforcement learning. It models the road network as a graph and learns end-to-end routing policies that maximize driver satisfaction while considering travel time and congestion, via a two-stage training process: a generic model trained on traditional routing factors and driver-preference models trained to reflect individual preferences. Key contributions include the use of a GIN-E graph representation, a DQN-based DRL agent with a multi-objective reward, and a driver behavior classifier that links driving patterns to road attributes; inference selects the appropriate driver-specific model using cosine similarity. Experimental results on the LuST dataset show up to a 17% improvement in aligning routes with driver preferences and up to a 46% reduction in travel time compared to a shortest-distance baseline, demonstrating the method's practical potential for real-time, personalized navigation in autonomous and connected vehicles.

Abstract

Conventional route planning services typically offer the same routes to all drivers, focusing primarily on a few standardized factors such as travel distance or time, overlooking individual driver preferences. With the inception of autonomous vehicles expected in the coming years, where vehicles will rely on routes decided by such planners, there arises a need to incorporate the specific preferences of each driver, ensuring personalized navigation experiences. In this work, we propose a novel approach based on graph neural networks (GNNs) and deep reinforcement learning (DRL), aimed at customizing routes to suit individual preferences. By analyzing the historical trajectories of individual drivers, we classify their driving behavior and associate it with relevant road attributes as indicators of driver preferences. The GNN is capable of representing the road network as graph-structured data effectively, while DRL is capable of making decisions utilizing reward mechanisms to optimize route selection with factors such as travel costs, congestion level, and driver satisfaction. We evaluate our proposed GNN-based DRL framework using a real-world road network and demonstrate its ability to accommodate driver preferences, offering a range of route options tailored to individual drivers. The results indicate that our framework can select routes that accommodate driver's preferences with up to a 17% improvement compared to a generic route planner, and reduce the travel time by 33% (afternoon) and 46% (evening) relatively to the shortest distance-based approach.

Personalized and Context-aware Route Planning for Edge-assisted Vehicles

TL;DR

The paper addresses the need for personalized, context-aware routing for edge-enabled vehicles by integrating graph neural networks with deep reinforcement learning. It models the road network as a graph and learns end-to-end routing policies that maximize driver satisfaction while considering travel time and congestion, via a two-stage training process: a generic model trained on traditional routing factors and driver-preference models trained to reflect individual preferences. Key contributions include the use of a GIN-E graph representation, a DQN-based DRL agent with a multi-objective reward, and a driver behavior classifier that links driving patterns to road attributes; inference selects the appropriate driver-specific model using cosine similarity. Experimental results on the LuST dataset show up to a 17% improvement in aligning routes with driver preferences and up to a 46% reduction in travel time compared to a shortest-distance baseline, demonstrating the method's practical potential for real-time, personalized navigation in autonomous and connected vehicles.

Abstract

Conventional route planning services typically offer the same routes to all drivers, focusing primarily on a few standardized factors such as travel distance or time, overlooking individual driver preferences. With the inception of autonomous vehicles expected in the coming years, where vehicles will rely on routes decided by such planners, there arises a need to incorporate the specific preferences of each driver, ensuring personalized navigation experiences. In this work, we propose a novel approach based on graph neural networks (GNNs) and deep reinforcement learning (DRL), aimed at customizing routes to suit individual preferences. By analyzing the historical trajectories of individual drivers, we classify their driving behavior and associate it with relevant road attributes as indicators of driver preferences. The GNN is capable of representing the road network as graph-structured data effectively, while DRL is capable of making decisions utilizing reward mechanisms to optimize route selection with factors such as travel costs, congestion level, and driver satisfaction. We evaluate our proposed GNN-based DRL framework using a real-world road network and demonstrate its ability to accommodate driver preferences, offering a range of route options tailored to individual drivers. The results indicate that our framework can select routes that accommodate driver's preferences with up to a 17% improvement compared to a generic route planner, and reduce the travel time by 33% (afternoon) and 46% (evening) relatively to the shortest distance-based approach.
Paper Structure (13 sections, 8 equations, 6 figures, 2 tables)

This paper contains 13 sections, 8 equations, 6 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Preliminaries
  4. Graph neural networks
  5. Reinforcement learning
  6. Personalized and Context-aware Route Planning
  7. System model
  8. Characterizing the driver's behavior and preferences
  9. GNN-based DRL agent
  10. Performance Evaluation
  11. Model pre-training
  12. Training driver's preferences
  13. Conclusions

Figures (6)

  • Figure 1: An overview of the driver's preference identifier.
  • Figure 2: An overview of the proposed GNN-based DRL framework.
  • Figure 3: GNN-DRL learning progress: Mean reward trend.
  • Figure 4: Route selection considering traffic light phases.
  • Figure 5: GNN-DRL learning progress: Mean reward trend with driver preferences. $DPM-V1$ (Left) considers [straight, two, simple, low] as the preference vector, $DPM-V2$ (Left) considers [straight, one, simple, low] as the preference vector.
  • ...and 1 more figures