Multi-Agent Production Scheduling Patents 2026
Multi-Agent Production Scheduling Patents
Autonomous software agents are replacing centralized batch planners in manufacturing. This dataset spans 60+ patent and literature records from 2005 through 2024, covering negotiation protocols, deep RL dispatching, and digital twin integration.
From Centralized Solvers to Autonomous Agent Networks
Multi-agent production scheduling (MAPS) applies autonomous, communicating software agents — representing machines, jobs, transport vehicles, or planning modules — to decompose and coordinate complex manufacturing scheduling problems in real time. The field has gained urgency as Industry 4.0 demands, mass customization pressures, and supply chain volatility expose the limitations of centralized, batch-oriented planning systems.
The technology spans six interconnected sub-domains: decentralized job-shop and flexible job-shop scheduling, Contract Net Protocol and auction-based task allocation, reinforcement learning integrated with MAS, Digital Twin-synchronized scheduling, distributed multi-site manufacturing scheduling, and AGV and transport agent co-scheduling. The earliest records in this dataset date to 2005; the most recent active patents were filed in 2024.
The 2020–2022 cohort is the densest cluster in this dataset. Digital Twin integration became a dominant theme, with multiple papers linking real-time field data to scheduling agents. Cloud manufacturing agent architectures, supply chain disruption response MAS, and IoMT-based assembly line scheduling all appeared in this window. Deep RL and Asset Administration Shell standardization represent the technical frontier as of 2023–2024.
In this dataset, patent filings are concentrated among 8 distinct assignees across approximately 15 patent records, while the literature base is far broader — reflecting that much MAPS innovation is disseminated through academic publication rather than patent protection, consistent with university-led research dominance in this field in retrieved records.
Technology Cluster Distribution and Filing Activity Over Time
The retrieved records reveal a clear concentration in negotiation/auction-based and deep RL-based scheduling clusters, with digital twin integration emerging strongly from 2020 onward. Commercial patent filings accelerated between 2020 and 2024.
Patent Records by Technology Cluster (Dataset Snapshot)
Negotiation/auction-based scheduling and deep RL-based scheduling are the two most represented clusters in this dataset, together accounting for the majority of retrieved patent and literature records.
↗ Click bars to exploreCommercial Patent Filings by Period (Dataset Snapshot)
In this dataset, commercial patent filings were sparse before 2015 but accelerated markedly in the 2020–2024 period, with the most recent filings from AVEVA, Rafael Advanced Defense Systems, Waymo, and Tata Consultancy Services.
↗ Click bars to exploreKey Application Domains in Multi-Agent Production Scheduling
Retrieved records span six distinct application domains, from discrete job-shop manufacturing and matrix production systems to aerospace mission planning and cloud computing resource scheduling. Each domain presents distinct agent architecture and protocol requirements.
Discrete Job-Shop Manufacturing
The dominant application domain across the entire dataset, where agents represent machines, jobs, and operations to handle real-time disruptions and variable routing. A 2020 study implemented and deployed a MAS in a physical lab testbed for job-shop scheduling. A 2019 knowledge-and-agent-based system uses pair agents to resolve timing conflicts in cascading process plans.
Discrete ManufacturingMatrix and Mixed-Model Assembly
Increasing product variety drives agent-based control in matrix and flexible assembly systems. A 2021 study applied economic bidding to matrix-shaped production cells, enabling individual job routing paths. A separate 2021 work couples IoMT with MAS for dynamic mixed-model assembly line scheduling under real-time demand variation.
Mass CustomizationIntegrated AGV and Production Scheduling
Material handling has become inseparable from production scheduling in retrieved records. A 2021 MARL study addresses combined flexible job shop and AGV scheduling simultaneously. A 2023 paper benchmarks autonomous mobile robot dispatching methods in an agent-based simulation environment validated against a digital twin of the shop floor.
Integrated LogisticsAerospace and Defense Mission Scheduling
A 2017 study models satellites, receiving stations, and observation areas as negotiating agents for Earth remote sensing scheduling. Rafael Advanced Defense Systems filed a PCT application in 2023 and a US patent in 2024 formalizing mission allocation across tactical agent groups using a Mission Agent Language (MAL) framework.
Aerospace & DefenseLeading Patent Assignees in Multi-Agent Scheduling — Dataset Snapshot
In this dataset, Blue Yonder Group, Inc. is the most patent-prolific commercial assignee with 3 active US patents, followed by B.G. Negev Technologies at Ben-Gurion University, Waymo LLC, and Tata Consultancy Services each with 2 records in retrieved records. The overall landscape is concentrated among 8 distinct assignees across approximately 15 patent records.
Top Assignees by Patent Filing Count — Multi-Agent Scheduling (Dataset Snapshot)
↗ Click bars to exploreBlue Yonder Group, Inc.
Blue Yonder Group holds 3 active US patents filed in 2015, 2017, and 2018, making it the most patent-prolific commercial assignee in this dataset. All three patents cover a system and method for solving supply chain campaign planning problems involving major and minor setups — addressing multi-product family scheduling across supply chain horizons. All patents are listed as active in the retrieved records.
United StatesB.G. Negev Technologies — Ben-Gurion University
B.G. Negev Technologies & Applications Ltd. at Ben-Gurion University holds 2 active US patents filed in 2021 and 2023, covering a master-slave deep reinforcement learning architecture for multi-objective scheduling. The master DRL agent manages a queue of item-representations while the slave DRL agent handles task-level execution policy learning. Both patents are listed as active in the retrieved records.
United States / IsraelFive Frontier Directions Visible in 2023–2024 Records
Based on records dated 2022–2024 in this dataset, five frontier directions are identifiable: AAS-based agent interoperability, deep RL displacing rule-based dispatching, AMR co-scheduling, enterprise-level autonomous digital twins, and defense multi-agent mission scheduling.
Asset Administration Shell as Agent Interoperability Layer
A 2023 paper positions AAS as the I4.0-native description language for scheduling agents, enabling plug-and-play interoperability across heterogeneous factory assets. An automatic parser extracts AAS data to initialize agent knowledge bases in the PADE MAS framework. This directly addresses one of the historically critical barriers to industrial MAS adoption identified in multiple retrieved sources.
Deep RL Replacing Rule-Based Dispatching Heuristics
The 2023 MARL paper demonstrates outperformance of FIFO, SPT, and EDD dispatching rules across five dynamic job shop scenarios by minimizing tardiness and flow time. AVEVA’s 2024 IN patent trains agents at discrete plant decision points to learn ranking policies maximizing a global plant scheduling reward. Ben-Gurion’s master-slave DRL patents (2021, 2023) represent the current commercial state of the art for learned dispatching.
Negotiation-Based vs. Deep RL-Based Multi-Agent Scheduling
Click any row to explore further.
| Dimension | Negotiation / Auction-Based MAS | Deep RL-Based MAS |
|---|---|---|
| Contract Net Protocol or combinatorial auctions between job and resource agents | Agents learn dispatching policies through interaction with simulated or real shop floor environments | N/A |
| 2005 — decentralized dispatching within cross-company production networks | 2021 — master-slave DRL architecture (Ben-Gurion University, US patent) | N/A |
| 2021 iterative combinatorial auction achieves near-centralized solution quality while preserving information privacy | 2023 MARL outperforms FIFO, SPT, and EDD across five dynamic job shop scenarios on tardiness and flow time | N/A |
| Tata Consultancy Services EP/IN 2022 — L2-Norm attribute matching for real-time task self-allocation | AVEVA Software IN 2024 — agents trained at discrete plant decision points for global reward maximization | N/A |
| Handles dynamic task arrivals and information privacy without a global optimizer | Replaces hand-coded heuristics with adaptive policies that improve with experience | N/A |
| Auction complexity can approach NP-hard in large combinatorial settings | Requires significant training data and simulation environment; policy transfer to real deployments remains challenging | N/A |
| CNP-based protocols used in 2022 flow-shop digital twin resilient scheduling paper | AVEVA 2024 patent assigns RL agents to discrete plant decision points within a plant scheduling system | N/A |
| ~14 records in this dataset (largest cluster) | ~10 records in this dataset (fastest-growing 2020–2024 cluster) | N/A |
Frequently Asked Questions: Multi-Agent Production Scheduling
Multi-agent production scheduling (MAPS) applies autonomous, communicating software agents — representing machines, jobs, transport vehicles, or planning modules — to decompose and coordinate complex manufacturing scheduling problems in real time, without requiring a single centralized solver.
This dataset identifies four core clusters: negotiation and auction-based scheduling (using Contract Net Protocol or combinatorial auctions), deep RL-based scheduling (agents learn dispatching policies), digital twin-synchronized agent scheduling (real-time field data drives rescheduling), and evolutionary/metaheuristic hybrid MAS scheduling (genetic algorithms, particle swarm, tabu search combined with agent architectures).
Blue Yonder Group, Inc. holds 3 active US patents (2015, 2017, 2018) covering supply chain campaign planning — the most patent-prolific commercial assignee in this dataset. B.G. Negev Technologies at Ben-Gurion University, Waymo LLC, and Tata Consultancy Services each have 2 patent records in the retrieved dataset.
A 2023 MARL study in this dataset demonstrated that a multi-agent reinforcement learning system outperforms FIFO, SPT, and EDD dispatching rules across five dynamic job shop scenarios by minimizing tardiness and flow time. AVEVA’s 2024 patent trains agents at discrete plant decision points to learn ranking policies that maximize a global plant scheduling reward.
A 2023 paper in this dataset positions AAS as the I4.0-native description language for scheduling agents, enabling plug-and-play interoperability across heterogeneous factory assets. An automatic parser extracts AAS data to initialize agent knowledge bases in the PADE MAS framework, addressing a historically critical barrier to industrial MAS adoption.
The strategic analysis in this dataset identifies AGV/AMR co-scheduling as under-patented relative to its commercial importance — simultaneous production-and-transport scheduling remains predominantly in academic literature with very few active patents. AAS-to-agent connector middleware and supply chain disruption rapid replanning are also noted as patenting opportunities.
Data and insights on this page are based on a limited patent and literature dataset and are for reference only. Figures may not represent the complete technology landscape.