Multi-attribute Auction-based Resource Allocation for Twins Migration in Vehicular Metaverses: A GPT-based DRL Approach

TL;DR

This paper tackles VT migration in Vehicular Metaverses where RSU-limited edge resources and high vehicle mobility challenge real-time VT updates. It presents MADDA, a two-stage mechanism that first performs resource-attributes matching via Kuhn-Munkres on a weighted bipartite graph, then uses a GPT-based DRL auctioneer to adjust Dutch clocks and determine winners, optimizing social welfare under a multi-attribute market that includes location and reputation. The approach combines a reputation model with latency-aware valuation to align trust with efficiency, proving strong budget balance and incentive compatibility while enabling fast convergence and reduced information exchange costs; experiments show near-optimal welfare across varying market sizes and RSU capabilities. The work provides a scalable, dynamic framework for resource allocation in vehicular metaverses, with practical implications for edge-enabled VT migrations and immersive vehicle services.

Abstract

Vehicular Metaverses are developed to enhance the modern automotive industry with an immersive and safe experience among connected vehicles and roadside infrastructures, e.g., RoadSide Units (RSUs). For seamless synchronization with virtual spaces, Vehicle Twins (VTs) are constructed as digital representations of physical entities. However, resource-intensive VTs updating and high mobility of vehicles require intensive computation, communication, and storage resources, especially for their migration among RSUs with limited coverages. To address these issues, we propose an attribute-aware auction-based mechanism to optimize resource allocation during VTs migration by considering both price and non-monetary attributes, e.g., location and reputation. In this mechanism, we propose a two-stage matching for vehicular users and Metaverse service providers in multi-attribute resource markets. First, the resource attributes matching algorithm obtains the resource attributes perfect matching, namely, buyers and sellers can participate in a double Dutch auction (DDA). Then, we train a DDA auctioneer using a generative pre-trained transformer (GPT)-based deep reinforcement learning (DRL) algorithm to adjust the auction clocks efficiently during the auction process. We compare the performance of social welfare and auction information exchange costs with state-of-the-art baselines under different settings. Simulation results show that our proposed GPT-based DRL auction schemes have better performance than others.

Figures (6)

• Figure 1: The muti-attribute double Dutch auction-based resource allocation mechanism for reliable VTs migration in vehicular Metaverses.
• Figure 2: The proposed efficient GPT-based incentive mechanism to improve the efficiency of the DDA.
• Figure 3: Seller reputation under considering freshness compared with random weight scheme and no freshness scheme
• Figure 4: Convergence of the GPT-based DRL compared with DDA, SAC and random methods with or without considering reputation attributes.
• Figure 5: Performance evaluation under different system parameters. (a) and (b) shows the social welfare and auction information exchange cost versus market sizes. (c) and (d) show the winning pairs and trade success rates under different market sizes, respectively. (e)-(h) show the performance of social welfare, auction information exchange cost, winning pairs, and trade success rates under different computational powers owned by RSUs.

Theorems & Definitions (17)

• Definition 1: Individual Rationality
• Definition 2: Incentive Compatibility
• Definition 3: Strong Budget Balance
• Theorem 1
• proof
• Definition 4: Monotonicity
• Definition 5: Criticality
• Theorem 2
• Lemma 1
• proof