Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Bi-LSTM based Multi-Agent DRL with Computation-aware Pruning for Agent Twins Migration in Vehicular Embodied AI Networks

Yuxiang Wei, Zhuoqi Zeng, Yue Zhong, Jiawen Kang, Ryan Wen Liu, M. Shamim Hossain

TL;DR

The paper addresses dynamic task migration in Vehicular Embodied AI Networks by modeling AVs and RSUs as actors in a Multi Leader Multi Follower Stackelberg game to optimize bandwidth with QoS considerations. It introduces a Tiny MABLPPO framework, a Bi-LSTM based MADRL method, to approximate Stackelberg equilibrium under partial observability, and enhances it with computation-aware PX pruning to accommodate heterogeneous AV hardware. Theoretical equilibrium analysis establishes conditions for a unique SE and practical algorithms for follower and leader strategies. Numerical results on an embedded platform show improved load balancing and reduced latency, demonstrating the method’s potential for scalable, edge-powered VEAN deployments.

Abstract

With the advancement of large language models and embodied Artificial Intelligence (AI) in the intelligent transportation scenarios, the combination of them in intelligent transportation spawns the Vehicular Embodied AI Network (VEANs). In VEANs, Autonomous Vehicles (AVs) are typical agents whose local advanced AI applications are defined as vehicular embodied AI agents, enabling capabilities such as environment perception and multi-agent collaboration. Due to computation latency and resource constraints, the local AI applications and services running on vehicular embodied AI agents need to be migrated, and subsequently referred to as vehicular embodied AI agent twins, which drive the advancement of vehicular embodied AI networks to offload intensive tasks to Roadside Units (RSUs), mitigating latency problems while maintaining service quality. Recognizing workload imbalance among RSUs in traditional approaches, we model AV-RSU interactions as a Stackelberg game to optimize bandwidth resource allocation for efficient migration. A Tiny Multi-Agent Bidirectional LSTM Proximal Policy Optimization (TMABLPPO) algorithm is designed to approximate the Stackelberg equilibrium through decentralized coordination. Furthermore, a personalized neural network pruning algorithm based on Path eXclusion (PX) dynamically adapts to heterogeneous AV computation capabilities by identifying task-critical parameters in trained models, reducing model complexity with less performance degradation. Experimental validation confirms the algorithm's effectiveness in balancing system load and minimizing delays, demonstrating significant improvements in vehicular embodied AI agent deployment.

Bi-LSTM based Multi-Agent DRL with Computation-aware Pruning for Agent Twins Migration in Vehicular Embodied AI Networks

TL;DR

The paper addresses dynamic task migration in Vehicular Embodied AI Networks by modeling AVs and RSUs as actors in a Multi Leader Multi Follower Stackelberg game to optimize bandwidth with QoS considerations. It introduces a Tiny MABLPPO framework, a Bi-LSTM based MADRL method, to approximate Stackelberg equilibrium under partial observability, and enhances it with computation-aware PX pruning to accommodate heterogeneous AV hardware. Theoretical equilibrium analysis establishes conditions for a unique SE and practical algorithms for follower and leader strategies. Numerical results on an embedded platform show improved load balancing and reduced latency, demonstrating the method’s potential for scalable, edge-powered VEAN deployments.

Abstract

With the advancement of large language models and embodied Artificial Intelligence (AI) in the intelligent transportation scenarios, the combination of them in intelligent transportation spawns the Vehicular Embodied AI Network (VEANs). In VEANs, Autonomous Vehicles (AVs) are typical agents whose local advanced AI applications are defined as vehicular embodied AI agents, enabling capabilities such as environment perception and multi-agent collaboration. Due to computation latency and resource constraints, the local AI applications and services running on vehicular embodied AI agents need to be migrated, and subsequently referred to as vehicular embodied AI agent twins, which drive the advancement of vehicular embodied AI networks to offload intensive tasks to Roadside Units (RSUs), mitigating latency problems while maintaining service quality. Recognizing workload imbalance among RSUs in traditional approaches, we model AV-RSU interactions as a Stackelberg game to optimize bandwidth resource allocation for efficient migration. A Tiny Multi-Agent Bidirectional LSTM Proximal Policy Optimization (TMABLPPO) algorithm is designed to approximate the Stackelberg equilibrium through decentralized coordination. Furthermore, a personalized neural network pruning algorithm based on Path eXclusion (PX) dynamically adapts to heterogeneous AV computation capabilities by identifying task-critical parameters in trained models, reducing model complexity with less performance degradation. Experimental validation confirms the algorithm's effectiveness in balancing system load and minimizing delays, demonstrating significant improvements in vehicular embodied AI agent deployment.
Paper Structure (25 sections, 2 theorems, 34 equations, 10 figures, 1 table, 1 algorithm)

This paper contains 25 sections, 2 theorems, 34 equations, 10 figures, 1 table, 1 algorithm.

Table of Contents

  1. Introduction
  2. Related Works
  3. Embodied AI in Vehicular Networks
  4. Service Migration with Stackelberg Game
  5. Model Compression for DRL
  6. System Model
  7. Migration Model in Embodied AI
  8. Latency Model
  9. MLMF Stackelberg Game between AVs and RSUs
  10. Utility models of AVs and RSUs
  11. Stackelberg game formulation
  12. Equilibrium Analysis for MLMF Stackelberg Game
  13. Follower-level Equilibrium Analysis
  14. Leader-level Equilibrium Analysis
  15. Positiveness
  16. ...and 10 more sections

Key Result

Lemma 1

A function $\mathcal{H}_r(\boldsymbol{B})$ is a standard function if and only if it satisfies the following three conditions: where $\mathcal{H}_r(\boldsymbol{B})$ denotes the optimal strategy of RSU $r$.

Figures (10)

  • Figure 1: The system model for VEAAT migration.
  • Figure 2: TMABLPPO algorithm's Framework for the VEAAT migration.
  • Figure 3: Comparison of the total reward curves of MABLPPO and baselines for the Stackelberg Game.
  • Figure 4: Comparison of the total reward curves of TMABLPPO and baselines for the Stackelberg Game with 90% density.
  • Figure 5: Comparison of the total reward curves of TMABLPPO and other pruning algorithms for the Stackelberg Game with 90% density.
  • ...and 5 more figures

Theorems & Definitions (4)

  • Definition 1
  • Lemma 1
  • Theorem 1
  • proof