Automated market makers (AMMs) are central to decentralized exchanges, but their reserve curves make mechanism design different from classical double auctions. We study a batch double auction with buyers, sellers, and a liquidity-nondecreasing AMM: post-trade reserves must have weakly larger product than the initial reserves. Each user has two-dimensional private information: a unit value and a demand or supply quantity. User trades clear at one uniform price subject to AMM feasibility. The objective is user-side trading volume, benchmarked against the optimal feasible individually rational (IR) uniform-price allocation. We give a randomized uniform-price mechanism based on binary quantity buckets. It randomly chooses one of three clearing modes: peer-to-peer matching, AMM sales to buyers, and seller sales to the AMM. Every deterministic realization of the mechanism is ex-post feasible and satisfies IR and weak quantity incentive compatibility: conditional on any quantity report, truthful unit-value reporting is dominant whenever it gives a non-negative utility, no unit-value deviation gives a positive utility if the truthful one doesn't, and users cannot benefit by under-reporting the quantity together with their truthful unit value. Let $V^\ast$ be the optimal feasible IR uniform-price volume without incentive constraints, and let $γ=\bar q/\underline q$, where $\bar q$ and $\underline q$ are the quantity-domain bounds. For $γ\ge1$, the mechanism achieves an expected trading volume of $ Ω(1/(1+\logγ)^2)V^\ast - O(\bar q). $ We also show that full incentive compatibility over the two-dimensional type space forces a worst-case approximation ratio of $O(1/γ)$ at every prescribed initial reserve ratio, even when both AMM reserves are scaled by an arbitrarily large common factor and the optimal volume $V^\ast$ is arbitrarily large relative to $\bar q$.

翻译：暂无翻译