Technical Article
Know Your Power
Why State of Power Must Become a Real-Time Constraint for Humanoid Motion
A robot needs a real-time battery power envelope before motion, because remaining energy alone cannot guarantee the next maneuver safely.
Executive Summary
A humanoid can show plenty of remaining battery energy yet be unable to execute the next motion safely. State of charge describes capacity; state of power describes how much power the pack can deliver or absorb for a defined time, without violating current, voltage, temperature, or cell limits. Walking recovery, lifting, jumping, braking, and joint acceleration create power pulses whose feasibility changes with temperature, aging, cell imbalance, impedance, and regenerative charge acceptance. The proposed semiconductor architecture links precise cell voltage, pack current, temperature, battery models, state estimation, and power-limit prediction to motion planning and control. Instead of discovering a weak battery through voltage sag or protection trips, the robot should expose a time-dependent power envelope before committing its body to demanding motion. Battery intelligence therefore becomes an actuator constraint: the robot must know not only energy remaining, but what power is available for the robot's next demanding physical maneuver ahead.Visual pending: Systems concept
Energy Remaining Is Not Power Available
Two batteries can show the same SoC and offer different immediate capability. One may be warm, healthy and balanced. Another may be cold, aged or limited by a single weak cell. Their energy gauges can appear similar while their safe peak currents, terminal-voltage margins and regenerative acceptance differ materially. The 2024 State-of-Power review highlights operating condition, battery model, safe operating area and estimation error as coupled determinants of usable peak power. Guo et al. The time dimension matters too. A pack may sustain a high power pulse for milliseconds, less power for several seconds and still less continuously. Therefore a single fixed power rating is an incomplete interface to robot control. A useful power envelope is indexed by a prediction horizon: the control system should know what is available now for the duration of the contemplated action. This is particularly important in humanoids because the electrical load is tightly coupled to physical state. The same lift can require different power depending on posture, payload, acceleration profile and which joints cooperate. A disturbance can produce an unplanned burst of demand. Regenerative braking can reverse the sign of power within the same maneuver. Physical intelligence therefore needs an electrical capability constraint that changes with the machine.The Battery Has Several States, Not One Percentage
Battery management already operates with multiple estimated states. State of Energy addresses usable energy; State of Health reflects degradation; SoC estimates remaining charge; and SoP estimates short-term deliverable or absorbable power. Infineon's current 48 V BMS architecture explicitly lists SoC, SoH, SoP and SoS calculations in the battery control unit, making SoP an established BMS function rather than a robotics-specific invention. Infineon 48 V BMS For robotics, the architectural step is to expose SoP to the consumers that need it. A battery controller may correctly protect the pack by opening a contactor or limiting operation, but discovering an electrical limit only after a motion command has already propagated through the robot is too late. The planner, whole-body controller or local motion supervisor should receive a capability envelope early enough to reshape the action.State of Power Is a Constraint-Satisfaction Problem
Battery SoP is not merely current multiplied by nominal voltage. Online methods predict terminal voltage and other state variables under candidate current or power trajectories, then determine which constraint becomes active first. The classic multi-constraint formulation by Xiang and colleagues evaluates current, voltage, state-of-charge and design-power limits over a specified interval. Xiang et al. Newer reviews describe a broader field involving equivalent-circuit, electrochemical, data-driven and hybrid approaches, with temperature, aging and model uncertainty influencing accuracy. 2026 model-based SoP review That constraint perspective is useful for humanoid architecture because it produces an actionable output. Instead of sending hundreds of battery variables to motion software, the BMS can calculate an admissible discharge and charge boundary plus confidence. The robot controller then compares requested power against available power and chooses among authorize, reshape, derate, delay or safe degradation.The Semiconductor Chain Starts With Measurement
A trustworthy SoP estimate is only as good as the measurements underneath it. Cell voltages establish proximity to upper and lower voltage limits. Pack current determines instantaneous load and supports coulomb counting. Temperature changes resistance, reaction kinetics, charge acceptance and protection thresholds. Synchronized acquisition matters because voltage and current must describe the same electrical event if model parameters and impedance are to be inferred correctly. Infineon's BMS material illustrates the semiconductor functions behind that chain: cell monitoring and balancing, pack-current measurement, coulomb counting, battery control, isolated communication, protection and diagnostics. Its current-sensing guidance emphasizes accurate current measurement together with fast overcurrent detection, while the battery control unit aggregates voltages, temperatures and currents for state calculation and housekeeping. Infineon current sensing Infineon BCU Functional safety adds another layer. The referenced BMS safety architecture combines current sensing, pack-voltage monitoring, diagnostics, isolated communication and fast overcurrent response. Those mechanisms do not calculate the robot's trajectory, but they establish the trustworthy measurement and protection substrate on which a higher-level power envelope depends. Infineon BMS functional-safety noteVisual pending: Systems concept
Temperature, Aging and Cell Variation Move the Boundary
Internal impedance converts current into loss and voltage sag. As impedance rises, a demanded current can drive terminal voltage toward its lower limit sooner. Temperature changes impedance and allowable operating conditions; aging changes both resistance and capacity. These effects mean SoP must be estimated from present condition rather than treated as a nameplate constant. Cell imbalance makes the weakest series cell important. A pack controller cannot safely use the average cell when one cell reaches its voltage boundary first. This is why cell-level monitoring and balancing are directly relevant to system power capability, even though the robot experiences only the aggregate pack output. The reverse direction is equally important. During negative mechanical work, energy flows from the joints back toward the DC bus. The battery can accept only the regenerative power permitted by its current state. Charge acceptance can narrow at high state of charge, low temperature or other limiting conditions. If the pack cannot absorb the requested regeneration, the power architecture must route energy elsewhere, reduce regeneration or dissipate it safely. This complements rather than duplicates the earlier W4M chapter on bidirectional energy flow: that chapter owns where regenerated energy goes; Chapter 031 owns whether the battery can accept it now.Visual pending: Systems concept
From Estimate to Motion Authorization
The robotics interface should be more useful than a raw SoP percentage. A motion stack needs at least direction, horizon and confidence. Direction distinguishes discharge from regenerative capability. Horizon distinguishes a sub-second recovery pulse from sustained walking or lifting. Confidence indicates whether sensor uncertainty, model mismatch or rapidly changing conditions justify extra margin. A practical message could contain several power limits: maximum discharge power for 100 milliseconds, one second and ten seconds; continuous power; maximum regenerative acceptance over comparable horizons; thermal or cell-limited flags; and a confidence or validity indicator. The exact interface depends on platform architecture, but the principle is stable: capability must be communicated before control commits the body. This turns the BMS from a protective subsystem into a participant in physical decision-making without moving motion control into the battery. The battery controller remains responsible for measurement, estimation, local protection and capability reporting. The robot controller remains responsible for deciding what action to take. The interface between them becomes an explicit contract.Why a Power Envelope Improves Physical AI
Motion planners already respect kinematic limits, collision constraints and torque boundaries. Electrical capability should join that set. A power-aware planner can reduce acceleration, distribute effort across more joints, alter posture, postpone a noncritical lift, lower speed, or choose a recovery strategy whose demand remains within the current envelope. It can also avoid commanding regeneration the battery cannot accept. This matters for learning-based control as well. A policy trained on an abstract actuator model may learn maneuvers that depend on electrical peaks unavailable in a cold or aged production robot. Exposing the power envelope to simulation, policy conditioning and validation can reduce that sim-to-real mismatch. The battery becomes part of the robot's observable physical state rather than an invisible energy tank.The Safe Operating Area Must Become a Planning Boundary
The Safe Operating Area of the battery is normally enforced locally through limits and protection. SoP estimation converts that local safety envelope into a predictive system constraint. Instead of waiting for a cell voltage, current or temperature threshold to be reached, the robot can anticipate when a requested action is likely to cross it. That prediction does not replace hard protection. The BMS must still enforce independent electrical and thermal limits because higher-level software can fail, data can be stale and estimates can be wrong. The architectural objective is layered defense: predictive capability helps the robot avoid invalid commands; deterministic protection remains available when prediction or control is insufficient.A Different Definition of Battery Intelligence
Battery intelligence in humanoid robotics should not be measured by how accurately the display predicts remaining runtime. Runtime matters, but a physical machine also needs to know what its stored energy can become in the next instant. A robot that understands its available power can act with greater confidence because it does not confuse energy inventory with executable strength. The semiconductor opportunity follows directly. Accurate voltage, current and temperature sensing; synchronized data conversion; robust microcontrollers; protected communications; nonvolatile state; diagnostics; isolation and fast protection all contribute to the quality of the capability estimate. Recent research continues to improve State-of-Power algorithms and computational efficiency for embedded use. 2026 Pegasus SoP research The final design rule is simple: before a humanoid commits mass, torque and momentum to a demanding action, it should know whether the electrical system can support that action over the required time. Energy tells the robot how long it may keep working. State of Power tells it what is physically possible now.Glossary
- Battery Management System
- The electronic system that measures, estimates, protects, communicates and controls battery operating state across cells, modules and pack functions.
- Cell imbalance
- Variation among series-connected cells in charge, capacity, resistance or voltage that can cause one cell to reach an operating limit first.
- Charge acceptance
- The amount of charging or regenerative power a battery can safely absorb under its present voltage, temperature, health and state conditions.
- Internal impedance
- The battery's effective opposition to current flow, including resistive and dynamic components that influence voltage response, losses and available power.
- Power envelope
- The time-dependent boundary of allowable battery discharge and regenerative power under the currently active electrical, thermal and state constraints.
- Prediction horizon
- The future time interval over which battery power capability is estimated; allowable peak power usually depends on the selected duration.
- Safe Operating Area
- The combination of permitted electrical, thermal and state conditions within which a battery is intended to operate without exceeding defined limits.
- State of Charge
- An estimate of remaining battery charge relative to its usable charge capacity; it does not directly state instantaneous power capability.
- State of Energy
- An estimate of the energy that remains usable from a battery under defined operating conditions.
- State of Health
- A measure of battery degradation relative to a reference condition, commonly reflecting capacity, resistance, power capability or combined indicators.
- State of Power
- An estimate of the maximum battery power that can be delivered or absorbed over a defined prediction horizon without violating active operating constraints.
- Voltage sag
- A temporary drop in terminal voltage under load caused by current-dependent internal losses and dynamic electrochemical behavior.
Sources
- A review of model-based state of power estimation for lithium-ion batteries · 2026-07-30 · Journal of Energy Storage / Elsevier
Review compares model-based State-of-Power methods, battery electrochemical effects, algorithm trade-offs, hierarchical estimation, digital twins, and cloud-edge deployment directions for batteries.
https://www.sciencedirect.com/science/article/pii/S2352152X26022498 - Automotive battery control unit (BCU) · 2026 · Infineon Technologies AG
Infineon positions the battery control unit as the estimator, communications hub, thermal coordinator, balancing controller, and protection decision node for batteries.
https://www.infineon.com/application/automotive-battery-control-unit - Automotive battery management system (BMS) - 48 V · 2026 · Infineon Technologies AG
Infineon identifies battery-control functions calculating SoC, SoH, SoP and SoS from monitored pack information while communicating with domain controllers securely.
https://www.infineon.com/application/automotive-48v-bms - Automotive current sensing & coulomb counting · 2026 · Infineon Technologies AG
Infineon describes precise pack-current measurement, coulomb counting, fast overcurrent detection, temperature stability, and protection signaling as core BMS measurement functions today.
https://www.infineon.com/applications/automotive/electric-drivetrain/battery-management-system/current-sensing-coulomb-counting - Fast and robust state-of-power estimation for lithium-ion batteries using pegasus algorithm · 2026 · Energy Conversion and Management / Elsevier
Recent research proposes computationally efficient State-of-Power estimation addressing rapid embedded prediction under changing battery conditions and operational constraints for real-time applications.
https://www.sciencedirect.com/science/article/pii/S0196890426005637 - FuSa in a Nutshell: BMS application with AURIX TC4xx · 2025-10-24 · Infineon Technologies AG
Application note links redundant current sensing, voltage monitoring, diagnostics, isolated communication, overcurrent response, secure communication and functional-safety mechanisms in BMS architectures.
https://www.infineon.com/assets/row/public/documents/10/42/infineon-fusa-in-a-nutshell-bms-application-with-aurix-tc4xx-applicationnotes-en.pdf - Lithium-Ion Battery Online Rapid State-of-Power Estimation under Multiple Constraints · 2018-01-24 · Energies / MDPI
Paper estimates lithium-ion peak charge and discharge power over defined horizons while enforcing current, voltage, charge, and design-power constraints online simultaneously.
https://www.mdpi.com/1996-1073/11/2/283 - Recent advancements in battery state of power estimation technology: A comprehensive overview and error source analysis · 2024-12-01 · Journal of Energy Storage / Elsevier
Review maps battery State-of-Power definitions, safe-operating constraints, estimation algorithms, modeling choices, uncertainty sources, and error propagation for lithium-ion systems comprehensively.
https://www.sciencedirect.com/science/article/pii/S2352152X24038805
