TRATSS: Transformer-Based Task Scheduling System for Autonomous Vehicles

2025-04-07

AI Summary

This work addresses the NP-hard problem of single-agent task scheduling in autonomous driving scenarios. Methodologically, it proposes a dynamic graph-based scheduling framework that integrates Transformer architectures with reinforcement learning—specifically, it is the first to introduce self-attention mechanisms into vehicular task scheduling to model long-range task dependencies and dynamic environmental perception over graph-structured representations. The framework combines Proximal Policy Optimization (PPO), graph-state encoding, and online policy fine-tuning to enable real-time resource adaptation and joint optimization of multi-dimensional scheduling actions. Its core contribution lies in overcoming the generalization limitations of conventional heuristic methods under complex dependency constraints. Evaluated on standard benchmarks, the approach achieves an 18.7% improvement in task completion rate and a 22.3% increase in resource utilization, significantly outperforming existing state-of-the-art methods in overall scheduling quality.

Technology Category

• Planning, Routing, and Scheduling: Learning for Planning and Scheduling
• Search and Optimization: Learning to Search
• Humans and AI: Human-Aware Planning and Behavior Prediction

Application Category

• Search and Retrieval-Augmented AI: Agentic search
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Responsible Web: Machine-in-the-loop, human agency and autonomy

Abstract

Efficient scheduling remains a critical challenge in various domains, requiring solutions to complex NP-hard optimization problems to achieve optimal resource allocation and maximize productivity. In this paper, we introduce a framework called Transformer-Based Task Scheduling System (TRATSS), designed to address the intricacies of single agent scheduling in graph-based environments. By integrating the latest advancements in reinforcement learning and transformer architecture, TRATSS provides a novel system that outputs optimized task scheduling decisions while dynamically adapting to evolving task requirements and resource availability. Leveraging the self-attention mechanism in transformers, TRATSS effectively captures complex task dependencies, thereby providing solutions with enhanced resource utilization and task completion efficiency. Experimental evaluations on benchmark datasets demonstrate TRATSS's effectiveness in providing high-quality solutions to scheduling problems that involve multiple action profiles.

Problem

• Solve NP-hard scheduling for autonomous vehicles
• Optimize resource allocation using transformer architecture
• Enhance task completion efficiency with reinforcement learning

Innovation

• Transformer-based scheduling for autonomous vehicles
• Reinforcement learning with dynamic task adaptation
• Self-attention captures complex task dependencies