auction-based task allocation

This work addresses the challenge of real-time multi-agent task allocation, which is often hindered by the high computational cost of high-fidelity path planning. To overcome this limitation, the authors propose a distributed two-stage, multi-fidelity bundle generation framework. In the first stage, low-fidelity heuristics rapidly explore the space of possible task bundles; in the second stage, only high-potential candidates undergo computationally expensive, high-fidelity path planning. A central coordinator then solves a set-packing problem to ensure global feasibility and utility maximization. The approach supports dynamic bundle sizes and enables reactive, real-time assignment while preserving agents’ internal states and proprietary cost models. Experimental results across diverse simulation scenarios demonstrate that the framework significantly enhances the performance of reactive auction-based task allocation, achieving both efficiency and scalability in multi-robot coordination.

This work addresses the challenge of balancing social welfare and mechanism simplicity in divisible resource allocation. The authors propose a unified framework that interpolates between the Kelly mechanism and first-price auctions, employing proportional allocation with uniform pricing. This approach retains structural simplicity while strictly improving upon the efficiency guarantees of the Kelly mechanism—achieving full efficiency in certain regimes—and offering revenue guarantees relative to the VCG mechanism. Leveraging mechanism design theory, equilibrium analysis, and interpolation techniques, the proposed family of mechanisms significantly enhances allocative efficiency at equilibrium while maintaining strong revenue performance.

This paper studies online resource allocation with complementary valuations: buyers arrive sequentially, each demanding a bundle of items, and their valuations are drawn from a known distribution. Standard item pricing fails to capture complementarities, while existing bundle-pricing mechanisms suffer from poor generalizability and reliance on restrictive structural assumptions. To address this, we propose the first unified static anonymous bundle-pricing framework, establishing the first general theoretical foundation applicable to diverse complementary settings—including combinatorial auctions and graph routing. Our competitive ratio improves exponentially with resource capacity and reveals a deep connection to qualitative independent partitioning in extremal combinatorics. For $d$-single-minded buyers, we achieve an $O(d^{1/B})$ competitive ratio; for general single-minded and graph-routing settings, we obtain $O(m^{1/(B+1)})$, where $B$ is the minimum bundle size and $m$ the number of resources. We further provide tight information-theoretic lower bounds, resolving a long-standing theoretical gap in this domain.

In electricity combinatorial auctions, the restriction on XOR bundle bids limits bidders’ preference expression and incurs significant social welfare loss. Method: This paper proposes a computationally feasible dynamic envelope bid selection mechanism: (i) it quantifies welfare loss from “exclusive block group” bids in European electricity markets; (ii) it integrates budget constraints and price predictability priors to design a bid-space compression algorithm and a counterfactual equilibrium price simulation method; and (iii) it formulates bid selection as an integer linear program augmented with welfare sensitivity analysis. Contribution/Results: Empirical evaluation demonstrates that the approach reduces social welfare loss by 12–19% at equivalent computational cost, while delivering a deployable bid-assistance tool and policy alternatives for European day-ahead and intraday electricity markets.

This paper studies a combinatorial principal-agent problem: a principal must incentivize an agent to select a subset of unobservable actions—each incurring cost—to complete a high-cost task, where success probability exhibits diminishing returns. We introduce the first combinatorial contract model capturing structural relationships between action subsets and success probability. Our theoretical contributions are threefold: (i) gross substitutes is a tight condition for polynomial-time computation of optimal contracts; under submodularity, the problem is NP-hard; (ii) we establish a deep connection to combinatorial auctions; (iii) we design a polynomial-time algorithm based on value queries and prove that linear contracts are robustly optimal under first-moment constraints on the agent’s type distribution. Collectively, our results provide both computational foundations and fundamental complexity boundaries for high-dimensional, structured incentive design.

本文研究带奖励的竞标游戏，通过引入新技巧消除次优玩法，解决无限配置空间问题，并提出一种保证可达性玩家获胜的新策略类型。

This study addresses the inefficiencies in pricing and resource allocation arising from complementarities under “No Assembly” constraints in combinatorial double auctions. The authors propose a constrained combinatorial buyer-bid double auction model that incorporates stability and price impact conditions, enabling bundled submarkets to inherit the competitive discipline of single-item large markets, thereby achieving efficient price discovery and eliminating strategic underbidding. Theoretically, in the two-good case, bundle bidding introduces no first-order strategic distortion, and complementarity mitigates frictions due to limited market size; clearing prices converge to competitive levels and track common values. Simulations reveal that welfare losses stem primarily from the “No Assembly” constraint rather than strategic behavior, are negligible in moderately thick markets, and further diminish as complementarity strengthens.

This study addresses the misalignment between individual bidding and global optimality in decentralized multi-robot task allocation by proposing a Grouped Auction Consensus Algorithm. The method reconstructs task-level bids into group-level structured actions via spatial clustering and employs a two-stage negotiation architecture compatible with partial acquisition mechanisms to effectively bridge individual-team objective gaps. Across 4,000 test scenarios, the algorithm achieves a median optimality rate of 97%, significantly outperforming CBBA’s 81–84% while demonstrating faster convergence and superior scalability. Validated against mixed-integer linear programming benchmarks as near-globally optimal, this approach establishes a novel paradigm for efficient coordination in large-scale robotic swarms.

研究探讨了在不同机制下，通过乘法调速方法解决数字平台预算管理问题，并分析其均衡存在性、唯一性和计算效率。

This study investigates multi-player discrete-bidding graph games, where token ownership is determined each turn via auction. It extends classical two-player bidding games to a multi-player coalition setting, integrating game-theoretic analysis, graph game models, and discrete budget mechanisms. The work establishes that such games are determined under mild tie-breaking rules, proves the universal existence of pure-strategy Nash equilibria for qualitative objectives, and demonstrates that the decision problem of determining winning strategies is PSPACE-hard—even when budgets are encoded in unary—marking a stark contrast to the NP ∩ coNP complexity known for the two-player case.