AI Recommendation Systems for Lane-Changing Using Adherence-Aware Reinforcement Learning

2025-04-28

AI Summary

In semi-autonomous driving, human drivers exhibit partial compliance with AI-generated lane-change recommendations, undermining system efficacy and safety. Method: This paper formulates single-vehicle lane-change recommendation as an Adherence-Aware Markov Decision Process (AA-MDP), explicitly modeling human–AI collaborative behavior through a novel “adherence-aware” mechanism. We propose the Adherence-aware Deep Q-Network (Adherence-aware DQN) algorithm, integrating reinforcement learning with human factors modeling in the CARLA simulation platform. Contribution/Results: Experimental evaluation under realistic traffic conditions demonstrates significant improvements in traffic throughput and decision reliability: lane-change failure rate decreases by 37% compared to baseline methods. The approach achieves a balanced trade-off among recommendation practicality, operational safety, and deployability—establishing a new paradigm for trustworthy human–AI shared driving decision-making.

Technology Category

• Humans and AI: Human-Aware Planning and Behavior Prediction
• Multiagent Systems: Adversarial Agents
• Search and Optimization: Learning to Search

Application Category

• Responsible Web: Machine-in-the-loop, human agency and autonomy
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Economics, Online Markets and Human Computation: Trust and reliance of crowd workers and data experts on GenAI

Abstract

In this paper, we present an adherence-aware reinforcement learning (RL) approach aimed at seeking optimal lane-changing recommendations within a semi-autonomous driving environment to enhance a single vehicle's travel efficiency. The problem is framed within a Markov decision process setting and is addressed through an adherence-aware deep Q network, which takes into account the partial compliance of human drivers with the recommended actions. This approach is evaluated within CARLA's driving environment under realistic scenarios.

Problem

• Optimizing lane-changing recommendations for semi-autonomous driving
• Addressing human driver partial compliance with AI suggestions
• Enhancing single vehicle travel efficiency via reinforcement learning

Innovation

• Adherence-aware reinforcement learning for lane-changing
• Deep Q network with partial human compliance
• Evaluation in CARLA's realistic driving scenarios