Vehicle Embedded System Power Optimization 2026 — PatSnap Eureka
Vehicle Embedded System Power Optimization: 2026 Patent Landscape
From SoC-level dynamic voltage scaling to AI-driven hybrid energy management, this report maps the active patent signals, key assignees, and emerging IP clusters shaping vehicle embedded power optimization from 2009 to 2025.
Two Interlocking Domains of Vehicle Power Optimization
Vehicle embedded system power optimization sits at the intersection of automotive electrification, real-time control software, and silicon-level energy management. Driven by tightening global CO₂ regulations, the proliferation of software-defined vehicle architectures, and the growing computational demands of advanced driver-assistance and autonomous systems, the field has expanded well beyond simple powertrain efficiency.
The first domain is chip- and SoC-level power management: techniques such as dynamic voltage and frequency scaling (DVFS), dynamic power management (DPM), event-driven workload modelling, and imitation learning-based runtime controllers that govern energy consumption within embedded processors and microcontrollers deployed throughout the vehicle. Key technical mechanisms include hardware abstraction layer (HAL)-based energy modelling, hypervisor-mediated hardware parameter alteration, event-based workload prediction, and hierarchical machine-learning policies for DPM.
The second domain is vehicle-level energy management systems (EMS): software and control architectures running on embedded platforms that determine how power is split, stored, and consumed across propulsion sources, energy storage devices, and ancillary loads. Mechanisms span rule-based strategies, equivalent consumption minimization strategies (ECMS), dynamic programming (DP), reinforcement learning (RL)-based real-time optimizers, model predictive control (MPC), and lookahead/predictive approaches leveraging GPS and V2X data.
A notable emerging sub-domain is the centralized vehicle HPC power management architecture, where a single high-performance compute domain replaces distributed ECU networks — creating a new class of embedded power optimization challenge directly evidenced by filings from Chinese assignees addressing PNC-based wake/sleep orchestration. For more on automotive software architectures, see PatSnap’s IP analytics platform and external standards work at ISO on functional safety.
Three Phases of Patent Activity: 2009–2025
Publication dates across the retrieved dataset span 2009 to 2025, enabling a three-phase characterization of maturity and innovation focus.
Foundational Phase: Energy Abstraction and Processor Power Delivery
Early filings focused on embedded system energy abstraction layers and processor-level power delivery. Arizona Board of Regents (Northern Arizona University) filed the foundational HAL/eHAL hypervisor architecture for embedded energy optimization as early as 2016 (US, active). NXP B.V. introduced its event-based power manager in this era. Avago Technologies (now Broadcom) filed multiple Power over Ethernet power delivery architecture patents from 2009 to 2011.
HAL · eHAL · Hypervisor · PoEDevelopment Phase: Lookahead EMS, RL Controllers, and PHEV Optimization
The most concentrated activity period. Cummins Inc. filed its lookahead-based hybrid fuel economy optimization system (WO, 2020; US, 2022). UT-Battelle filed its real-time PHEV powertrain efficiency optimization system (US, 2021). Ningbo Geely filed its RL-based adaptive real-time power/torque split method in multiple jurisdictions. The Arizona State University / University of Texas HiLITE imitation learning framework was filed internationally (WO, 2022). IIT Kharagpur filed its predictive energy management and drive advisory system (IN, 2022).
RL · MPC · ECMS · Lookahead EMSEmerging Phase: AI-Native Controllers and Centralized HPC Architectures
The most recent signals include the University of Texas System’s US grant of HiLITE (2025), UT-Battelle’s continued active PHEV control patent (US, 2025), ePropelled Inc.’s AI-driven electronic magnetic gearing inverter firmware (US, 2024; IN, 2024), Hydrogen Vehicle Systems Ltd.’s GB grant (2025), and two Chinese filings from AutoCore Intelligence Technology (Nanjing) for centralized vehicle HPC power management (CN, active, 2024). US and CN jurisdictions dominate recent grants.
HiLITE · HPC · PNC · eDTSUniversity-Origin IP Remains Commercially Active Across All Phases
Innovation in this dataset is moderately concentrated: approximately 4 assignees account for the majority of active patent results, but the field is not yet monopolized. University-origin IP from the University of Texas, Arizona Board of Regents, and IIT Kharagpur remains commercially active. The SoC-level power optimization IP space is university-dominated but increasingly being transferred to commercial entities. R&D teams should assess the HiLITE patent family — it covers hierarchical DPM with imitation learning for heterogeneous SoCs. See PatSnap customer case studies for IP transfer examples.
University IP · Commercial TransferFour Patent Clusters Shaping Embedded Vehicle Power
The retrieved dataset resolves into four distinct technology clusters, each representing a different layer of the vehicle power optimization stack — from silicon to powertrain to AI-driven control.
Top Assignees by Filing Count
University of Texas / Arizona Board of Regents, Hydrogen Vehicle Systems, and Avago Technologies each hold 4 filings — the highest count in this dataset.
Technology Cluster Distribution
Four clusters span the vehicle power optimization stack — from SoC DPM to vehicle-level EMS, RL-based controllers, and centralized HPC architectures.
From Passenger EVs to Centralized HPC Platforms
The retrieved dataset covers five distinct vehicle application domains, each presenting unique power optimization constraints and IP opportunities.
IP Risks, Filing Opportunities, and Freedom-to-Operate Signals
Five strategic signals emerge from the most recent filings and assignee positioning in this dataset.
HiLITE: Foundational SoC DPM Claim Scope
The University of Texas HiLITE patent family (US grant, 2025) covers hierarchical DPM with imitation learning for heterogeneous SoCs — a broadly applicable foundational claim affecting embedded platform designers across ADAS, infotainment, and domain controller segments. R&D teams entering this space should assess this family carefully. PatSnap analytics can help map claim scope.
Ningbo Geely RL Power Split: FTO Risk for OEMs
Ningbo Geely’s RL-based adaptive power split patent family (WO, EP, US) represents a potential freedom-to-operate risk for any OEM or Tier 1 deploying reinforcement learning controllers for real-time HEV/BEV energy management. The broad vehicle-type scope — cars, trucks, buses, rail, marine, and off-road — makes this a high-priority family for IP strategists to monitor and design around. See PatSnap solutions for FTO workflows.
Five Directional Signals from 2023–2025 Filings
Based on the most recent filings in this dataset, four directional signals are identifiable — plus one V2X/cloud direction visible in literature but not yet dominant in patent filings.
AI-Native Embedded Power Controllers
The University of Texas HiLITE grant (US, 2025) and ePropelled’s AI-embedded inverter firmware (US, 2024) represent a shift from rule-based or lookup-table-driven power management to policies trained via imitation learning or reinforcement learning that execute as lightweight firmware on embedded hardware. The key innovation is keeping the inference footprint small enough for real-time embedded deployment without cloud dependency. Standards bodies such as IEEE are actively developing embedded AI deployment standards relevant to this cluster.
HiLITE · eDTS · Imitation Learning · FirmwareCentralized Vehicle HPC Power Architecture Management
AutoCore Intelligence Technology (Nanjing)’s 2024 CN patents specifically address PNC-based wake/sleep management for centralized domain controllers — a direct technical response to the industry-wide shift from 70–100 ECU distributed architectures to 3–5 domain or zonal controllers. This is a nascent but rapidly developing IP cluster. The AUTOSAR consortium’s adaptive platform specification underpins the software architecture this cluster targets.
PNC · Domain Controller · Zonal · HPCFuel Cell Multi-Source Optimization with Simulation-in-the-Loop
Hydrogen Vehicle Systems Ltd.’s 2025 GB grant for active powertrain control integrating a simulation module signals that simulation-in-the-loop embedded control is moving from R&D to production-intent IP. The scope covers ancillary power (HVAC, parasitic loads) alongside propulsion — expanding the optimization boundary beyond the drivetrain. For FCEV system standards, see the IEC TC69 electric vehicle standards work.
FCEV · MPC · Simulation-in-the-Loop · HVACRL Multi-Vehicle Generalization and V2X Cloud-Augmented EMS
Ningbo Geely’s US continuation (2024) of its RL-based power split controller explicitly covers cars, trucks, buses, rail, marine, and off-road vehicles from a single patent family — indicating an assignee strategy to establish a foundational RL-for-power-split IP position across transport modes. Separately, literature signals (2020–2022) on asynchronous cloud update for predictive hybrid EMS and V2X-based speed prediction point to an emerging cloud-edge hybrid EMS architecture. Patent filings in this specific V2X sub-area were not yet retrieved in sufficient volume to identify dominant assignees — suggesting a relatively open IP space. See PatSnap’s technology solutions for landscape gap analysis.
RL · Multi-modal · V2X · Cloud-Edge EMSJurisdiction Distribution and Assignee IP Positions
| Assignee | Filings (Dataset) | Jurisdictions | Key Technology | Status |
|---|---|---|---|---|
| Board of Regents, Univ. of Texas / Arizona Board of Regents | 4 | US, WO | HiLITE hierarchical imitation learning DPM for heterogeneous SoCs | Active |
| Hydrogen Vehicle Systems Ltd. | 4 | GB, US, WO | Multi-source FCEV active powertrain control with simulation-in-the-loop MPC | Active |
| Avago Technologies International Sales Pte. Limited | 4 | US | Power over Ethernet power delivery architecture (embedded power delivery) | Granted |
| Ningbo Geely Automobile Research & Development Co., Ltd. | 3 | WO, EP, US | RL-based adaptive real-time power/torque split across vehicle types | Active |
Vehicle Embedded System Power Optimization — key questions answered
Vehicle embedded system power optimization encompasses two interlocking domains: chip- and SoC-level power management (techniques such as DVFS, DPM, event-driven workload modelling, and imitation learning-based runtime controllers) and vehicle-level energy management systems (EMS) that determine how power is split, stored, and consumed across propulsion sources, energy storage devices, and ancillary loads.
HiLITE is a hierarchical imitation learning framework developed by Arizona State University and granted to the University of Texas System. It trains DPM policies offline and applies lightweight regression policies at runtime on heterogeneous SoC platforms. Claimed improvements include a 40% reduction in energy-delay product and up to 76% reduction in deadline misses versus prior art.
Among the retrieved patent results, the top assignees by filing count include the University of Texas System / Arizona Board of Regents (4 filings, US and WO), Hydrogen Vehicle Systems Ltd. (4 filings, GB, US, WO), Ningbo Geely Automobile Research and Development (3 filings, WO, EP, US), Avago Technologies (4 filings, US), Cummins Inc. (2 filings, WO, US), and NXP B.V. (2 filings, US).
Centralized vehicle HPC power management replaces distributed ECU networks (70–100 ECUs) with 3–5 domain or zonal controllers, creating a new class of embedded power optimization challenge. AutoCore Intelligence Technology (Nanjing) has filed two CN patents (2024, active) specifically addressing PNC ID-based application and peripheral wake/sleep orchestration for these central computing domains.
Four directional signals are identifiable from 2023–2025 filings: (1) AI-native embedded power controllers using imitation learning or reinforcement learning as lightweight firmware; (2) centralized vehicle HPC power architecture management using PNC-based wake/sleep orchestration; (3) fuel cell multi-source power optimization with simulation-in-the-loop control; and (4) RL-based adaptive torque/power split with multi-vehicle-type generalization covering cars, trucks, buses, rail, marine, and off-road vehicles.
Ningbo Geely’s RL-based adaptive power split patent family (WO, EP, US jurisdictions) represents a potential freedom-to-operate risk for any OEM or Tier 1 deploying reinforcement learning controllers for real-time HEV/BEV energy management. The broad vehicle-type scope covering cars, trucks, buses, rail, marine, and off-road vehicles makes this a high-priority family for IP strategists to monitor and design around.
PatSnap Eureka searches patents and research literature to answer instantly.