Abstract:Real-time dynamic path planning in complex traffic environments presents challenges, such as varying traffic volumes and signal wait times. Traditional static routing algorithms like Dijkstra and A* compute shortest paths but often fail under dynamic conditions. Recent Reinforcement Learning (RL) approaches offer improvements but tend to focus on local optima, risking dead-ends or boundary issues. This paper proposes a novel approach based on causal inference for real-time dynamic path planning, balancing global and local optimality. We first use the static Dijkstra algorithm to compute a globally optimal baseline path. A distributed control strategy then guides vehicles along this path. At intersections, DynamicRouteGPT performs real-time decision-making for local path selection, considering real-time traffic, driving preferences, and unexpected events. DynamicRouteGPT integrates Markov chains, Bayesian inference, and large-scale pretrained language models like Llama3 8B to provide an efficient path planning solution. It dynamically adjusts to traffic scenarios and driver preferences and requires no pre-training, offering broad applicability across road networks. A key innovation is the construction of causal graphs for counterfactual reasoning, optimizing path decisions. Experimental results show that our method achieves state-of-the-art performance in real-time dynamic path planning for multiple vehicles while providing explainable path selections, offering a novel and efficient solution for complex traffic environments.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is to achieve real - time multi - vehicle dynamic path planning in complex traffic environments. Specifically, traditional static routing algorithms such as Dijkstra and A* perform well in calculating the shortest path, but have limitations when dealing with dynamic traffic conditions, for example, they are unable to effectively cope with traffic flow changes and traffic signal waiting times. In addition, although existing methods based on reinforcement learning (RL) have made some progress in dynamic navigation, they usually only focus on local optimal solutions and ignore global optimality, causing vehicles to potentially get stuck in dead ends or deviate from network boundaries. To this end, the paper proposes a new method based on causal reasoning - DynamicRouteGPT, aiming to achieve real - time dynamic path planning for multiple vehicles in actual road networks while balancing global and local optimality. This framework first uses the static Dijkstra algorithm to calculate the global optimal path as a baseline, and then guides vehicles to travel along the baseline path through a distributed control strategy. When a vehicle arrives at an intersection, DynamicRouteGPT will make real - time decisions based on real - time traffic information, driving preferences and unexpected events to optimize the dynamic path selection at each intersection. The DynamicRouteGPT framework integrates multiple technologies, including Markov chains, Bayesian inference and modern large - scale pre - trained language models (such as Llama3 8B), to provide an efficient and reliable path planning solution. This framework not only processes real - time traffic data, but can also dynamically adjust path planning according to various traffic scenarios and driving preferences. Experimental results show that DynamicRouteGPT has achieved state - of - the - art performance in multi - vehicle control and overall time optimization, demonstrating its effectiveness and potential in actual traffic scenarios. By combining global path optimization and local path dynamic adjustment, DynamicRouteGPT provides a novel and efficient dynamic navigation solution suitable for complex traffic environments.

DynamicRouteGPT: A Real-Time Multi-Vehicle Dynamic Navigation Framework Based on Large Language Models

A Machine Learning Method for Dynamic Traffic Control and Guidance on Freeway Networks

Dynamic Route Guidance Based on Model Predictive Control

Constrained Reinforcement Learning for Vehicle Motion Planning with Topological Reachability Analysis

Deep Reinforcement Learning Based Dynamic Route Planning for Minimizing Travel Time

G-Routing: Graph Neural Networks-Based Flexible Online Routing

Game Theory Based Approach for Massive Route Planning in Dynamic Road Networks

Urban Multiple Route Planning Model Using Dynamic Programming in Reinforcement Learning

Let Hybrid A* Path Planner Obey Traffic Rules: A Deep Reinforcement Learning-Based Planning Framework

Fast Multi-Class Vehicle Cooperative Path Optimization in Complex Urban V2X Transportation: A Novel Parallel Multi-Agent Reinforcement Learning Approach

XRouting: Explainable Vehicle Rerouting for Urban Road Congestion Avoidance Using Deep Reinforcement Learning

AlphaRoute: Large-Scale Coordinated Route Planning Via Monte Carlo Tree Search.

A decentralized path planning model based on deep reinforcement learning

Dynamic Route Guidance Using Neurodynamic Programming

PARP: A Parallel Traffic Condition Driven Route Planning Model on Dynamic Road Networks

GRL-PS: Graph Embedding-Based DRL Approach for Adaptive Path Selection

Learning How to Dynamically Route Autonomous Vehicles on Shared Roads

Graph Convolution-Based Deep Reinforcement Learning for Multi-Agent Decision-Making in Mixed Traffic Environments

Online Vehicle Routing With Neural Combinatorial Optimization and Deep Reinforcement Learning

Tactical Driving Decisions of Unmanned Ground Vehicles in Complex Highway Environments: A Deep Reinforcement Learning Approach

Real-time deep reinforcement learning based vehicle navigation