Abstract
The Internet of Vehicles (IoV) is undergoing a transformative evolution, enabled by advancements in future 6G network technologies, to support intelligent, highly reliable, and low-latency vehicular services. However, the enhanced capabilities of loV have heightened the demands for efficient network resource allocation while simultaneously giving rise to diverse vehicular service requirements. For network service providers (NSPs), meeting the customized resource-slicing requirements of vehicle service providers (VSPs) while maximizing social welfare has become a significant challenge. This paper proposes an innovative solution by integrating a mean-field multi-agent reinforcement learning (MFMARL) framework with an enhanced Vickrey-Clarke-Groves (VCG) auction mechanism to address the problem of social welfare maximization under the condition of unknown VSP utility functions. The core of this solution is introducing the "value of information" as a novel monetary metric to estimate the expected benefits of VSPs, thereby ensuring the effective execution of the VCG auction mechanism. MFMARL is employed to optimize resource allocation for social welfare maximization while adapting to the intelligent and dynamic requirements of IoV. The proposed enhanced VCG auction mechanism not only protects the privacy of VSPs but also reduces the likelihood of collusion among VSPs, and it is theoretically proven to be dominant-strategy incentive compatible (DSIC). The simulation results demonstrate that, compared to the VCG mechanism implemented using quantization methods, the proposed mechanism exhibits significant advantages in convergence speed, social welfare maximization, and resistance to collusion, providing new insights into resource allocation in intelligent 6G networks.
1. Introduction
In the era of 6G and the Internet of Vehicles (IoV), Network Service Providers (NSPs) utilize Network Function Virtualization (NFV) to "slice" resources—such as bandwidth, computation, and cache—for diverse Vehicle Service Providers (VSPs). The primary objective is to maximize social welfare, defined as the aggregate benefit all VSPs receive from these resources.
However, two major hurdles impede this goal. First, the "Black Box" Utility Problem arises because NSPs typically lack knowledge of VSPs' internal utility functions, which are often private or mathematically complex. Second, VSPs are profit-driven and may engage in collusion or dishonest bidding to manipulate prices. While traditional Vickrey-Clarke-Groves (VCG) auctions encourage truth-telling, they generally require explicit knowledge of these utility functions to calculate fair payments.
2. The Dual-Currency VCG Auction Mechanism
To overcome these challenges, the authors propose a three-phase auction mechanism that introduces "Value of Information" (VoI) as a novel monetary metric, referred to as VoI Currency (VC). This currency acts as a proxy for the expected benefit a VSP derives from a resource slice. A neutral third-party entity, the "Bank," is introduced to manage the VC lifecycle, separating financial incentives from the NSP's resource allocation to protect VSP privacy.
As illustrated in Figure 3, the mechanism operates in three distinct phases:
- Phase I (Issuance): The Bank estimates total VoI based on historical billing messages to create a VC pool.
- Phase II (Bidding): VSPs submit "intention" bids in VC to the NSP, who conducts a sealed auction to determine resource distribution.
- Phase III (Disclosure & Taxation): After allocation, the Bank compares VSPs' bids against their actual realized performance (gross profits). A "VC interest" tax is applied to ensure that truthful bidding remains the most profitable strategy for every participant.
3. MFMARL-Based Optimization Framework
Standard multi-agent reinforcement learning struggles with the "curse of dimensionality" as the number of vehicles grows. To ensure scalability, the authors employ Mean-Field Multi-Agent Reinforcement Learning (MFMARL). In this framework, each agent reacts to the average (mean field) behavior of its neighbors rather than tracking every individual agent, allowing the system to efficiently manage 50+ VSPs.
Figure 6 details how the MFMARL framework is embedded within the auction phases. The Bank and NSP utilize Online and Target Critic/Actor networks to iteratively learn optimal allocation strategies ($a$) and pricing based on observations ($o$) from the environment. This integration allows the NSP to learn the relationship between bids and resources without requiring direct access to sensitive VSP utility data.
4. Key Results and Insights
- Truthfulness and Collusion Resistance: The mechanism is mathematically proven to be Dominant-Strategy Incentive Compatible (DSIC). Simulations showed that even when VSPs attempted to collude, their net profits were consistently highest when returning to truthful bidding. The exploration-based randomness of MFMARL further prevents stable collusive strategies.
- Efficiency: The proposed method drastically reduces information exchange. For a network of 50 VSPs, it required 2000x fewer floating-point value exchanges compared to high-precision quantized VCG methods.
- Performance: DVCG-MFMARL reached a stable, high social welfare state significantly faster than standard MADDPG-based models. The achieved social welfare closely approached the theoretical optimal allocation across various scenarios.
5. Conclusion and Future Work
This research provides a robust framework for resource allocation in intelligent 6G networks by combining an enhanced VCG auction with MFMARL. By utilizing VoI as a currency and introducing a neutral Bank, the system maximizes social welfare while preserving privacy and deterring collusion. Future work will focus on expanding this model to complex MIMO systems and addressing the high mobility of vehicles in real-world dynamic network topologies.
Summary updated