peaq 백서

bitpro가 원문에서 글자만 모은 내용이에요. 표·그림·수식은 빠지거나 깨질 수 있어요. 전체 5쪽 중 3쪽

dapters Adapters are standardized interfaces that connect external service markets to the Orchestration layer. A compute network, an inference marketplace, a storage system or a payment rail each registers through an adapter that exposes its capabilities, pricing and availability in a format the system can reason about. Adapters are open and permissionless. Any provider can build one to make its offerings discoverable and bookable through the system. Their outputs are validated and reputation-scored like any other system participant, ensuring quality without gatekeeping growth. Discovery Discovery searches across all adapter-connected markets to find service providers that match a task’s requirements. When a task arrives, Discovery identifies candidates from all connected markets that can fulfil it. If the task has multiple requirements, discovery returns a structured candidate set per requirement. Discovery builds on the service listings published through the Coordination layer and aggregated by the peaq Validators across chains. The output is not a decision. It is a structured set of candidates that are technically capable of fulfilling each part of the task. Scoring Scoring evaluates each candidate against a policy that combines cost, latency, reputation and compliance constraints. Cost and latency come from the adapter. Reputation comes from the Validation layer. These are the same verified trust scores that the system computes for every economic machine. Two providers may offer the same service at the same price. If one has a significantly higher verified reputation score than the other, the Scoring layer reflects that difference. Machines and their owners can define custom policies that adjust the weight of each signal based on the task’s requirements. A routine low-value task may prioritize cost. A high-value settlement may prioritize trust. Planning Planning takes scored candidates and composes them into an execution plan. For compound tasks, the plan maps a primary provider to each sub-task, assigns fallback providers, sets spend limits and defines timeout thresholds. It determines which sub-tasks run in parallel, which depend on the output of others and what triggers a fallback. Once composed, the plan is submitted for approval. Approval can come from the machine’s owner policy or be granted autonomously if the task falls within pre-configured parameters. No economic commitment is made without authorization. Execution The approved plan is handed to the runtime layer for execution. The runtime handles authentication, skill invocation and external service calls. Orchestration monitors the lifecycle and triggers fallback providers if primaries fail or timeout. On completion, the Coordination layer anchors receipts and records settlement. The Validation layer evaluates the outcome and updates the reputation scores of all participants based on whether they delivered what they promised. Value is settled through the selected payment rails. Validation Validation is the economic accountability layer of the system. It turns raw, unverified data on permissionless registries into trustworthy, queryable trust signals, secured by staked capital and enforced with economic consequences. Without Validation, there is no way to know whether a machine is what it claims to be, whether its reputation is legitimate or whether it will deliver what it promises. Anyone can register a fake identity or post unaccountable claims. Validation closes this gap by making dishonesty economically irrational. The cost of being caught, slashed stake and damaged reputation, exceeds the potential gain. Validation does not verify absolute truth. It verifies consistency and commitment. For onchain actions such as payments or token transfers, claims are cross-referenced against settlement records and transaction data. For service delivery, both parties’ claims are evaluated against origin-signed telemetry data produced by the Verification layer. For subjective or unverifiable claims, trust is derived from the machine’s staked capital and accumulated reputation. Verification can be triggered in two ways. The system continuously samples claims across the network for proactive verification, ensuring that any claim has a meaningful probability of being checked regardless of whether it is challenged. Any machine can also trigger verification by challenging a specific claim with proportional stake. The result is that any machine or app on any chain can query whether a machine is trustworthy and get back an answer backed by staked capital, not just data from a permissionless registry. Reputation Reputation is the trust score of an economic machine. It is not self-reported. It is computed and verified by the peaq Validators based on a dynamic combination of three variables: Stake : the more economic risk a machine has at stake, the higher its trust score. History : the more honest claims the machine has made in the past, the higher its reputation. Inherence : the higher the trust score of the identity that owns, operates or manufactures the machine, the higher its trust score. This creates a Recursive Trust Lineage Graph where trust propagates along ownership relationships. A creator’s trust score is the average of the scores of its created machines, while each machine’s score is partly influenced by its creator’s score. This forms a decentralized, recursive structure of accountability and incentive alignment across owners and their machines. Reputation Formula Reputation Score = √( wₛ × Stake² + wₕ × History² + wᵢ × Inherence² ) Stake is the capital a machine puts at risk. The input gets normalized using a logarithmic curve to enforce diminishing returns, preventing wealthy machines from gaining disproportionate trust by simply outstaking others. The weight ranges between 40-60%. This makes stake the most important and game-resistant anchor of trust. History reflects the machine’s onchain history of honest claims and successful interactions. The weight ranges between 0-40%. It increases with consistent, verifiable behavior. It enables machines to earn high trust autonomously, reducing sole dependence on stake or inherence over time. Inherence reflects the trust score of the machine’s creator, such as its owner, operator or manufacturer. The weight ranges between 0-30% and is most influential when the machine first enters the system. It offers initial credibility when the machine lacks history and ensures lineage-based accountability. Machines without verified lineage data carry an Inherence score of zero, which limits their maximum achievable reputation. This creates a gradual incentive for manufacturers and owners to verify their identity and establish trust chains as the ecosystem matures. wₛ, wₕ, wᵢ are bounded dynamic weights that reflect the relative strength of each component within its range. They are always normalized to sum to 1, ensuring fair balance, compensatory potential and resilience to dominance by any single metric. Each component is squared to amplify strong inputs, penalize weak ones and sharpen trust distinctions. An approach widely used in traditional reputation and risk scoring systems. ERC-8004 and 8004-Solana defines the registries where identity, reputation and validation data is stored. peaq’s Reputation formula is the economically secured computation layer that reads those registries and produces a verified, queryable trust score. The registries provide the schema. peaq Validators provide the bureau. Reputation scores are recorded in the Reputation Registry and made queryable across all supported chains, ensuring that a machine’s trust is portable and discoverable regardless of where it was built. Stake In the Validation network, stake operates at two levels. Machines deposit stablecoin denominated capital to back the claims they make in the system. The more they stake, the higher their trust score and the more credible their commitments. Denominating machine stake in stablecoins ensures that accountability remains constant regardless of token price movements. Separately, Validator operators stake PEAQ to earn the right to perform verification work. Operators who verify honestly earn fees from the system. Operators who submit dishonest or failed verifications lose their stake. This dual staking model ensures that both the machines being verified and the operators doing the verification have skin in the game. Rules Rules define the accountability framework for machines coordinating and settling through the system. They are the codified logic that turns claims into economic commitments with clearly defined outcomes. machines who use external settlement or coordination mechanisms operate outside this framework at their own discretion. Every action must be expressed as a cryptographically signed claim linked to the machine’s ID and timestamped in Universal Machine Time. Each claim carries the machine’s stake as collateral. The stake is at risk for the duration of the claim’s lifecycle. If peaq operators determine a claim to be false through cross-referencing settlement records, signed telemetry data or counter-evidence from a challenging machine, the claimant faces penalties ranging from reputation downgrades to partial or full slashing of their bonded stake, depending on the severity of the violation. Any machine may challenge another machine’s claim by putting stake at risk proportional to the challenge. This prevents baseless or dishonest challenges while ensuring that legitimate disputes can be raised regardless of the challenger’s trust score. Machines who choose to interact with unscored machines do so at their own risk. The system does not provide accountability or Arbitration for those interactions. When scored machines have conflicting claims, the dispute is escalated to system Arbitration, executed by the peaq Validators. The relative trust scores of the machines, the size of the dispute and the available evidence all factor into the resolution process and its cost structure. Arbitration Arbitration is the system’s final decision mechanism, executed by the peaq Validators. When claims are conflicting or ambiguous, the dispute is escalated to the Validator network for resolution. The peaq Validators do not determine truth in an absolute sense. They render a verdict based on system rules, data consistency, the historical records of the machines involved and available evidence such as deposited funds, financial transactions or signed telemetry data. The cost of initiating Arbitration is influenced by the relative trust scores of the disputing machines and the size of the dispute, ensuring the process is accessible but not baseless. This ensures that even in complex or adversarial scenarios, outcomes are resolved within the system by staked operators with skin in the game, not by external arbitration. Arbitration closes the loop of accountability. Core Validation Services The peaq Validators provide four core services in their initial version: Identity Verification : Is this machine real, unique and operational? Staked operators verify the authenticity and uniqueness of machine identities registered across chains, detecting Sybil attacks, fake registrations and dormant identities. Reputation Scoring : Is this machine’s track record legitimate and un-gamed? Staked operators continuously evaluate a machine’s claim history, stake and inherence to compute verified reputation scores that cannot be Sybil-farmed or artificially inflated. Transaction Evaluation : Did this machine deliver what it promised? Staked operators assess whether service commitments were fulfilled by cross-referencing claims, signed telemetry data and settlement records. Cross-Chain Attestation : All of the above, queryable on any chain. Verification results, reputation scores and attestations generated by the peaq Validators are made available across all supported chains, ensuring that trust built on one chain is portable to every other. Examples A delivery robot operates on Base, earning USDC per completed drop-off. It stores proof of delivery through a decentralized storage network and sources real time navigation from a specialized AI provider. When a second delivery platform on Avalanche needs capacity, it can discover and hire the same robot without re-onboarding it. The robot’s identity, performance history and trust score are already portable. Escrow is released only once delivery conditions are verified through claims, telemetry and signed receipts. Trust Validators attest fulfillment. The robot’s reputation updates after every job, improving its future terms, fee rates and eligibility for financing. A GPU compute marketplace on Solana settles workloads locally. But its operators want access to lending, insurance and fleet financing on other chains. By activating trust tiers through bonded PEAQ, each GPU node builds a portable reputation queryable beyond Solana. A lending system on another chain can evaluate that reputation before extending credit. An insurance market on a third chain can price coverage based on verified uptime. The machine’s economic surface expands without migrating a single transaction. peaq does not internalize execution. It coordinates trust and settles guarantees across the chains where machines already operate. Validation The peaq Validators are a decentralized network of operator nodes that power the system. There are two types of validators: State Validators and Trust Validators. State Validators State Validators are a decentralized and permissionless onchain runtime that maintains the canonical shared state of the system. They host the identity registries, claim records, service registries, reputation scores and settlement data that every layer reads from and writes to. State is hosted in a chain-neutral environment accessible from all connected chains. This ensures that no single chain owns the system’s data and that state written from any chain is readable and verifiable on every other. Trust Validators Trust Validators are a decentralized and permissionless set of offchain Validators that perform the computation the system depends on. In Onboarding , Trust Validators aggregate identity registrations, wallet provisions, and service registrations from across chains, making them queryable in a unified layer. A machine onboarded to one chain is visible and verifiable on every other. In Coordination , Trust Validators route claims, synchronize service registries and execute settlement logic across chains. In Orchestration , Trust Validators search adapter-connected markets, score candidates against reputation data, compose execution plans and monitor their lifecycle. In Validation , Trust Validators verify identities, cross-reference claims against settlement records and signed telemetry data, evaluate transactions and compute reputation scores. Each of these services can be queried independently. A machine that only needs identity verification does not need to use orchestration. A system that only needs reputation scores does not need to use settlement. Trust Validators read from and write to State Validators. When a Trust Validator computes a reputation score, verifies a claim or anchors a settlement receipt, the result is submitted to the State Validator runtime for canonical recording. When a Trust Validator needs registry data for discovery or scoring, it reads from State Validators. State Validators are the shared source of truth. Trust Validators are the workforce that acts on it. The architecture described above is not theoretical. peaq has been live and operating at scale across its initial monochain phase. The following figures reflect network activity as of March 2026. Over 3,300,000 machines, robots, devices, vehicles and agents have been onboarded with verified machine identities. The network has processed more than 200 million transactions. Over 60 DePIN and Machine Economy applications have been onboarded across industries including energy, mobility, connectivity, spatial intelligence and AI. Among them, Silencio operates over 1.1 million devices mapping environmental noise data, Teneo runs more than 2 million nodes, and DualMint onboarded the world’s first revenue-generating robot valued at over $250,000. peaq’s identity and data-sharing standards were co-developed with Bosch, Denso, Continental, Airbus and DLR (German Aerospace Center) through Gaia-X moveID, a European initiative establishing decentralized identity and data-sharing standards for mobility and IoT, with peaq serving as the underlying blockchain infrastructure. peaq pioneered DePIN in the automotive sector, partnering with Audi within the Volkswagen Group on decentralized charging infrastructure. The peaq Validator network includes institutional operators spanning investment, telecommunications, aviation and academic research: Bertelsmann Investments, Deutsche Telekom, Lufthansa Innovation Hub and the Technical University of Munich (TU Munich). Institutional validator participation provides the network with operational credibility and economic depth from its earliest stage. peaq facilitated the tokenization of the world’s first revenue-generating robot, a vertical farming operating in Hong Kong that autonomously grows and sells produce, generating a ~20% annual yield profile for its token holders. This is the first instance of a physical machine operating as a live, investable onchain financial asset, proving the Machine Money Markets thesis described later in this paper with a working product, not a concept. peaq opened the Machine Economy Free Zone, officially recognized by Dubai’s Virtual Assets Regulatory Authority (VARA). The zone is developing legal and technical frameworks that allow autonomous machines to operate within the UAE, handling transactions and interacting with smart city infrastructure. This is, to our knowledge, the first regulatory framework globally designed specifically for autonomous machine economic activity. These milestones were achieved on peaq’s initial mono-chain architecture. The omnichain upgrade described in this paper removes the remaining constraint, chain fragmentation, and is designed to unlock the next order of magnitude in machine participation. The system’s modular design means it does not require the Machine Economy to arrive all at once. It generates value at current scale and compounds as machine adoption grows. The traction above demonstrates that the problem is real, the demand exists and the team delivers. The upgrade expands the addressable market from a single chain’s ecosystem to the entire Machine Economy. Settlement-centric blockchain economics are built on a core assumption: value accrues where execution happens. Applications migrate to the chain. Transactions settle locally. System capture takes the form of transaction fees and staking demand. This model holds for human user-centric networks. It becomes structurally constrained when the primary economic actors are machines. Machines do not optimize for ecosystems. They optimize across them. At any moment, a machine selects the most efficient payment rail, liquidity venue, compute provider, storage network, inference market, and money market available to it — based on cost, latency, reliability, and risk-adjusted efficiency. When competition plays out, each chain leans into its strength rather than trying to copy everyone else. They specialize. A machine shouldn’t be stuck on one chain, because no single network will be best at everything. A machine could use Solana for speed, Base for settlement, Filecoin for storage, and Ethereum to become a liquid, tradable asset, access DeFi, and borrow capital through systems like Aave.That only works if there’s a neutral layer that lets the machine move freely across all of them. Locking machines to a single ecosystem comes at a cost. Something is always sacrificed to sustain system control. Over t
프로젝트로 돌아가기