Mastering EV Low-Speed Creep Control: From State Machine Logic to MCU FOC Implementation

• Author: Johnny Liu, CEO at Dowway Vehicle
• Published: July 19, 2026
• Category: Automotive Engineering / EV Powertrain & Control Algorithms
• Estimated Reading Time: 15 minutes

Author Biography & Authority

Johnny Liu is the Chief Executive Officer at Dowway Vehicle, a builder of electric vehicle (EV) powertrain solutions and advanced Vehicle Control Unit (VCU) architectures. With over 15 years of hands-on R&D experience in motor drive systems, automotive embedded systems, and functional safety (ISO 26262), Johnny has led the development of production-grade drive control algorithms deployed in hundreds of thousands of electric vehicles worldwide.

Technical Metadata

• Primary Entities: Vehicle Control Unit (VCU), Motor Control Unit (MCU), Permanent Magnet Synchronous Motor (PMSM), Field-Oriented Control (FOC).
• Core Algorithms: State Machine Control, Slope Compensation, Space Vector Pulse Width Modulation (SVPWM) Dead-Time Compensation, High-Frequency Injection (HFI).
• Target Audience Search Intent: Detailed EV control engineering, low-speed motor calibration, powertrain integration.

1. Introduction: The Physics of Creep — ICE vs. Electric Vehicles

Why do EVs require active creep control?

Unlike Internal Combustion Engine (ICE) vehicles, which crawl forward naturally due to the fluid coupling in a torque converter, Electric Vehicles (EVs) use Permanent Magnet Synchronous Motors (PMSMs) that generate no torque at zero speed when uncommanded. EVs must use software-driven creep control to mimic this forward crawl, prevent rollback on hills, and keep low-speed driving smooth.

+------------------------------------------------------------------------+
|                                                                        |
|  [ICE Passive Creep]                                                    |
|  Engine Idle ---> Torque Converter (Hydraulic Coupling) ---> Residual Torque  |
|                                                                        |
|  [EV Active Creep]                                                     |
|  VCU Sensors ---> State Machine ---> MCU (Active FOC) ---> Creep Torque |
|                                                                        |
+------------------------------------------------------------------------+

The Legacy Experience

In traditional ICE vehicles with automatic transmissions, the engine runs at idle speed to keep from stalling. When the vehicle stops and the gear is in Drive (D), the torque converter keeps spinning against the transmission fluid. This physical link sends a small residual torque to the wheels, making the vehicle crawl forward. Drivers expect this behavior.

The EV Challenge

For battery electric vehicles (BEVs) with PMSMs, the physics change:

1. Zero Command Means Zero Torque: At near-zero motor speed, the motor outputs zero torque unless we inject current.
2. No Passive Holding: Most EVs use a single-speed reduction gear without a mechanical lock outside of the parking pawl.
3. The “Floating” Pedal Hazard: If you release the brake on a slope without active software control, the vehicle rolls back freely. This creates an unsafe feeling for the driver and causes stress in tight city traffic.

The Three Core Creep Scenarios

We divide active creep control into three distinct situations:

1. Flat-Road Creep: The vehicle is on flat ground in D or R, with the driver releasing the pedals. The motor must output a stable, low-speed torque to crawl at $3\text{–}5\text{ km/h}$ (activating when $v < 8\text{ km/h}$).
2. Brake Creep: This is the transition where the driver modulates the brake pedal. The motor torque must scale down as brake pressure rises. This stops the motor from pushing against the physical brakes, preventing driveline vibration and ensuring a smooth stop.
3. Hill-Hold Creep: Holding the vehicle on hills. The VCU must detect the slope and add a gravity-compensating torque to the base creep torque to keep the vehicle still or crawling upward.

2. EV Creep Control System Architecture: VCU vs. MCU

How do the VCU and MCU share creep control tasks?

The Vehicle Control Unit (VCU) is the decision maker. It reads driver intent, merges sensor data, manages state changes, and calculates the target creep torque ($T_{req}$). The Motor Control Unit (MCU) is the fast executor. It receives the VCU’s torque commands and runs high-speed closed-loop $q$-axis current ($i_q$) tracking using Field-Oriented Control (FOC).

+------------------------------------------------------------------------+
|                      VCU - SYSTEM DECISION MAKER                       |
|                                                                        |
|  [Sensors]                                                             |
|  - Accelerator Pedal                                                   |
|  - Brake Pedal / Pressure                                               |
|  - Gear Shift (D/R)                                                    |
|  - IMU (Longitudinal Accel / Slope)                                    |
|                               |                                        |
|                               v                                        |
|                 +----------------------------+                         |
|                 |    Creep State Machine     |                         |
|                 +----------------------------+                         |
|                               |                                        |
|                               v                                        |
|                  Calculated Torque Request (Treq)                      |
+-------------------------------|----------------------------------------+
                                |
                                | CAN Bus (e.g., 10ms cycle)
                                v
+------------------------------------------------------------------------+
|                        MCU - HIGH-SPEED EXECUTOR                       |
|                                                                        |
|                       +----------------------+                         |
|                       | FOC Current Loop     |                         |
|                       | (iq Execution)       |                         |
|                       +----------------------+                         |
|                               |                                        |
|                               v                                        |
|                         [PMSM Motor]                                   |
|                               |                                        |
|         Feedback (Speed w, Rotor Angle theta)                          |
+------------------------------------------------------------------------+

System Roles

1. Vehicle Control Unit (VCU)

Running at a typical cycle of $10\text{ ms}$, the VCU reads multiple inputs:

• Driver Inputs: Accelerator pedal travel, brake pedal position/pressure, and gear (D/N/R).
• Motion Inputs: Wheel speed, Inertial Measurement Unit (IMU) pitch, and longitudinal acceleration.
• System Limits: Battery Management System (BMS) limits and inverter thermal limits.

The VCU processes these values through its state machine to find the target creep torque ($T_{req}$), then sends it over the CAN/CAN-FD bus.

2. Motor Control Unit (MCU)

Running much faster (usually $10\text{ kHz}$ or $100\text{ }\mu\text{s}$ for the current loop), the MCU controls the physical motor. It takes $T_{req}$ from the VCU and converts it to a $q$-axis target current ($i_{q\_ref}$) using motor-specific lookup tables ($T\text{–}I_q$). It uses space vector pulse width modulation (SVPWM) to drive the inverter while sending actual rotor speed ($\omega$) and electrical angle ($\theta$) back to the VCU.

3. How the VCU Creep Control State Machine Works

What are the main states of an EV Creep State Machine?

A standard production creep algorithm uses a Six-State State Machine: Standby (Off), Active (Flat-road creep), Braking (Blending torque with brakes), Hill-Hold (Holding on slopes), Exit (Transitioning to active driving), and Fault (Safety shutdown). Transitions depend on gear, vehicle speed ($v$), pedal travel, and diagnostic status.

       +---------------------------------------------+
       |                  [FAULT]                    |
       +---------------------------------------------+
             ^                  ^              ^
             | Fault            | Fault        | Fault
             |                  |              |
     +---------------+   Brake  |      +---------------+
     |               | --------->----->|               |
     |   [STANDBY]   |                 |   [BRAKING]   |
     |               | <---------<-----|               |
     +---------------+   Release       +---------------+
             |                                 |
             | Speed < Vcreep                  | Brake Released
             | Pedal < 5%                      | Uphill Slope
             v                                 v
     +---------------+                 +---------------+
     |               |---------------->|               |
     |   [ACTIVE]    |   Brake Pressed |  [HILL-HOLD]  |
     | (Flat Creep)  |<----------------| (Slope Comp)  |
     +---------------+                 +---------------+
             |                                 |
             | Pedal > 5%                      | Pedal > 5%
             | Or Speed > Vexit                | Or Speed > Vexit
             v                                 v
       +---------------------------------------------+
       |                  [EXIT]                     |
       |  (Torque Blending / 1st-Order Filter / Ramp) |
       +---------------------------------------------+

Critical State Transitions

State 1: Standby

This is the default state when creep is off. The VCU commands zero creep torque ($T_{req} = 0\text{ N}\cdot\text{m}$).

State 2: Active (Flat-Road Creep)

• Entry Rules:
  1. Gear is D or R.
  2. Vehicle speed $v < v_{creep}$ (usually $10\text{ km/h}$, using a $0.5\text{ km/h}$ hysteresis band to stop the state from jumping back and forth).
  3. Accelerator pedal position (APP) $< 5\%$ (driver is not pressing the accelerator).
  4. No high-level faults (such as BMS over-temperature, low traction voltage, motor over-temperature, or MCU over-current).

State 3: Braking (Brake Creep Transition)

• Entry Rules: Begins when the driver presses the brake pedal past a set start point (usually $15\%$ travel or pressure equivalent) while in the Active state.
• Control: The VCU drops the creep torque based on how far the driver presses the brake. This blending stops the motor from fighting the mechanical brakes, saving battery power and ensuring a smooth stop.

State 4: Hill-Hold (Slope Creep)

• Entry Rules: Activates when speed is near zero ($v \approx 0\text{ km/h}$), the driver releases the brake, and the IMU or estimating algorithm detects a slope ($\theta > \theta_{threshold}$).
• Control: The VCU calculates the gravity counter-torque and adds it to the base creep torque to hold the vehicle in place.

State 5: Exit

• Entry Rules: Activates when:
  • Accelerator pedal position (APP) exceeds $5\%$.
  • Vehicle speed $v$ exceeds the exit threshold $v_{exit}$ (usually $10\text{–}12\text{ km/h}$).
• Torque Blending: To avoid a sudden torque jump (which causes drivetrain shock), the VCU does not cut the creep torque instantly. It uses a 1st-order Low-Pass Filter (LPF) or a rate-limiter (Ramp) over a $200\text{–}500\text{ ms}$ window, fading out the creep torque as the driver’s pedal torque fades in.

State 6: Fault (Failsafe)

• Entry Rules: Activates instantly if there is any diagnostic trouble code (DTC) from the VCU, MCU, or BMS. The creep state machine shuts down, ramping torque down to $0\text{ N}\cdot\text{m}$ to keep the vehicle safe under ISO 26262.

4. Mathematical Design of Creep Torque & Blending

How is EV creep torque calculated and smoothed out?

We calculate base creep torque using a speed-dependent decay formula: $T_{base}(v) = T_{max} \times \max\left(0, 1 – \frac{v}{v_{exit}}\right)$. This drops the peak torque ($T_{max}$, usually $40\text{–}80\text{ N}\cdot\text{m}$) at zero speed down to $0\text{ N}\cdot\text{m}$ at the exit speed ($v_{exit}$). To prevent driveline shocks, we apply different ramp rates for rising and falling torque.

Speed-Dependent Linear Decay Formula

If creep torque stayed constant, the car would feel too jumpy at higher speeds and weak at a dead stop. Instead, we use a speed-dependent decay curve:$$T_{base}(v) = T_{max} \times \max\left(0, 1 – \frac{v}{v_{exit}}\right)$$

Where:

• $T_{base}(v)$ is the calculated base creep torque.
• $T_{max}$ is the peak static creep torque at zero speed ($40\text{–}80\text{ N}\cdot\text{m}$, tuned to vehicle weight and gear ratio).
• $v$ is the real-time vehicle speed.
• $v_{exit}$ is the speed where creep torque reaches zero ($10\text{–}12\text{ km/h}$).

  Creep Torque (Nm)
   ^
   | 
Tmax+----------------\
   |                  \  Flat-Road Creep Curve
   |                   \
   |                    \
   |                     \
   +----------------------+-------------> Vehicle Speed (km/h)
   0                     Vexit (10-12 km/h)

Advanced Torque Blending Logic

Sudden changes in torque cause drivetrain play and passenger discomfort. Time-domain filtering is necessary:

1. Rate Limiting (Ramping)

We pass the calculated torque through a rate-of-change limiter:$$\frac{dT_{req}}{dt} \le \text{Slew\_Rate}$$

2. Asymmetric Ramp Tuning

For best refinement, we tune the entry and exit ramps separately:

• Active Ramp-up (Brake Release): When the driver releases the brake, we keep the torque recovery ramp gentle (e.g., $100\text{–}150\text{ N}\cdot\text{m/s}$ or a filter time constant of $300\text{–}500\text{ ms}$). This avoids a sudden lurch.
• Active Ramp-down (Brake Press): When the driver presses the brake quickly, motor torque must drop fast (e.g., $300\text{–}500\text{ N}\cdot\text{m/s}$ or a filter time constant of $100\text{–}200\text{ ms}$). This stops the motor from driving immediately so the physical brakes can stop the car.

5. Slope Compensation & Hill-Hold Algorithms

How does an EV calculate slope compensation torque to prevent rolling back?

We calculate slope compensation using the balance of forces on an incline: $T_{slope} = \frac{M \cdot g \cdot \sin\theta \cdot r_{wheel}}{i_{gear} \cdot \eta}$. By finding the gravity force ($F_g = M \cdot g \cdot \sin\theta$) from IMU pitch data or dynamic acceleration estimations, the VCU adds this compensation value ($T_{slope}$) directly to the base creep torque ($T_{base}$) to balance gravity.

                  /|
                 / |  Fg = M*g*sin(theta)
                /  v
               /   |
              /    v 
             / [Car]
            /      
           / theta (Slope Angle)
          /------------------------

The Physics of Rollback

When an EV stops on an incline of angle $\theta$, gravity pulls it backward:$$F_g = M \cdot g \cdot \sin\theta$$

Where:

• $M$ is the vehicle mass (including estimated payload).
• $g$ is gravity ($9.81\text{ m/s}^2$).
• $\theta$ is the slope angle.

If the base creep torque $T_{base}$ converted to wheel force is smaller than $F_g$, the vehicle will accelerate backward, resulting in rollback.

Slope Sensing Methods

Method 1: Physical IMU Sensor (Direct Measurement)

Reading a 6-axis IMU to get real-time road pitch angle ($\theta$).

• Pros: Accurate when static or moving slowly.
• Cons: Picks up high-frequency body vibrations and road bumps. To solve this, we use a low-pass Chebyshev or Butterworth filter in the VCU to clean the signal.

Method 2: Dynamic Acceleration Estimation (Software-Based)

We estimate the incline by comparing wheel acceleration ($a_{wheel} = \frac{dv}{dt}$) with the motor-demanded acceleration model:$$a_{estimated} = \frac{F_{motor} – F_{drag}}{M}$$

Any difference between $a_{estimated}$ and $a_{wheel}$ comes from gravity ($g \cdot \sin\theta$).

• Pros: No extra hardware cost.
• Cons: Higher errors during sudden acceleration or braking.

Slope Compensation Formula

Once we have the road angle $\theta$, the VCU calculates the required motor torque:$$T_{slope} = \frac{M \cdot g \cdot \sin\theta \cdot r_{wheel}}{i_{gear} \cdot \eta}$$

Where:

• $r_{wheel}$ is the rolling radius of the tires.
• $i_{gear}$ is the transmission reduction ratio.
• $\eta$ is the drivetrain efficiency (usually $95\text{–}98\%$).

The final total creep torque command is:$$T_{creep\_req} = T_{base}(v) + T_{slope}$$

Engineering Edge Cases: Directional Sign Conventions

Note that the sign of $T_{slope}$ must match the gear and direction to avoid bugs:

Gear Position	Incline State	Gravity Force Direction	Necessary $T_{slope}$ Sign	Physical Behavior
Drive (D)	Uphill (Forward)	Backwards (Downhill)	Positive ($+$)	Adds to $T_{base}$ to climb the hill.
Drive (D)	Downhill (Forward)	Forwards (Downhill)	Negative ($-$)	Subtracts from $T_{base}$ to prevent the car from rolling away too fast down the slope.
Reverse (R)	Uphill (Backward)	Forwards (Downhill)	Positive ($+$)	Adds to $T_{base}$ (acting in reverse direction) to push the vehicle up the hill backward.
Reverse (R)	Downhill (Backward)	Backwards (Downhill)	Negative ($-$)	Subtracts from $T_{base}$ to prevent the car from descending backwards too fast.

6. MCU Low-Speed FOC Engineering Challenges and Fixes

What are the main low-speed FOC issues in an EV motor controller?

Low-speed Field-Oriented Control (FOC) suffers from three main issues: 1) Poor current signal quality because Back Electromotive Force (BEMF) is tiny near zero speed; 2) Inverter dead-time distortion causing voltage errors at low duty cycles; and 3) Rotor position lag from Resolver-to-Digital Converter (RDC) delays. We solve these using high-res ADCs, real-time voltage compensation, and sensored closed-loop FOC.

       +--------------------------------------------------------+
       |               MCU LOW-SPEED FOC LOOP                  |
       +--------------------------------------------------------+
                               |
                               v
                     [Current Sampling] 
                     - Low SNR / High Noise
                     * Solution: 14+ Bit ADC & Dynamic PI Tuning
                               |
                               v
                     [Park/Inverse Park]
                     - Rotor Angle Error
                     * Solution: 12-16 Bit Resolver (RDC Offset Comp)
                               |
                               v
                     [SVPWM Generation]
                     - Non-linear Dead-time Distortion
                     * Solution: Real-Time Vcomp Injection
                               |
                               v
                     [IPMSM Drive Output]

Challenge 1: Signal-to-Noise Ratio (SNR) Degradation

At high speeds, a PMSM generates a large Back Electromotive Force (BEMF) that dominates the system. At near-zero speeds during creep, the BEMF drops to zero. This leads to issues:

• The current feedback loop is weak against current ripple, electromagnetic noise, and Analog-to-Digital Converter (ADC) quantization errors.
• This noise passes through the Clarke and Park transformations, causing oscillation in the $q$-axis current ($i_q$) loop. This shows up as motor hum and steering wheel vibration.

Engineering Solutions

1. High-Resolution ADCs: We use high-precision current sensors with $14\text{-bit}$ or $16\text{-bit}$ ADC registers.
2. Bandwidth Filtering: We apply software low-pass filters to current feedback signals, tuning the cutoff frequency to protect the current loop stability margin.
3. Dynamic PI Gain Scheduling: We lower the current loop PI controller gains ($K_p$ and $K_i$) at low speeds. This stops high-frequency oscillations while letting the torque response slow down slightly (which is fine for slow creep maneuvers).

Challenge 2: Inverter Dead-Time Effect

To prevent a short-circuit across the high-side and low-side switching devices (IGBTs or SiC MOSFETs) of an inverter leg, we program a short physical delay called “dead-time” ($t_{dead}$, usually $1\text{–}4\text{ }\mu\text{s}$) into the gate signals.

Gate HS  ___|~~~~~~~~~|_____                  (High-Side Gate Signal)
            <--t_dead-->
Gate LS  ~~~~~~~~~|_____|~~~                  (Low-Side Gate Signal)

At low speeds, the inverter runs at tiny PWM duty cycles (low output voltage). During this time, the dead-time error voltage is a large percentage of the output voltage, causing non-linear distortion. This distortion creates a 6th harmonic current ripple, causing torque ripple and vehicle vibration.

Solution: Real-Time Dead-Time Compensation

The MCU compensates for this voltage distortion in the SVPWM module. Based on current direction, we add a compensation voltage vector ($V_{comp}$) in real-time:$$V_{comp} = \frac{V_{dc} \cdot t_{dead}}{T_{pwm}} \cdot \text{sign}(i_{phase})$$

Where:

• $V_{dc}$ is the traction battery DC bus voltage.
• $t_{dead}$ is the physical dead-time plus device switching delays.
• $T_{pwm}$ is the switching period of the PWM carrier (e.g., $100\text{ }\mu\text{s}$ for a $10\text{ kHz}$ switching frequency).
• $\text{sign}(i_{phase})$ is the sign of the respective phase current ($+1$ for outward flow, $-1$ for inward flow).

Challenge 3: Rotor Position Estimation Errors

Accurate Field-Oriented Control relies on knowing the exact rotor electrical angle ($\theta$) to align the magnetic fields.

• Sensorless FOC estimation fails at zero speed because there is no BEMF to track.
• Sensored FOC (using a physical resolver) relies on a Resolver-to-Digital Converter (RDC) chip. At low speeds, the tracking loop inside the RDC introduces delays and phase-lag. If uncorrected, this lag causes an angular offset $\Delta\theta$, which couples the $d$ and $q$ axis currents and causes loop oscillation.

Solution

For passenger vehicle creep, sensored FOC with a high-resolution resolver (using a $12\text{-}$ to $16\text{-bit}$ RDC) is standard. The MCU uses a speed-dependent angle phase-lead algorithm to correct for RDC propagation delays at low speeds.

7. High-Frequency Injection for Zero-Speed Control

How does High-Frequency Injection (HFI) enable sensorless low-speed motor control?

HFI injects a high-frequency voltage signal (usually $5\text{–}20\text{ V}$ at $500\text{ Hz–}2\text{ kHz}$) into the $d$-axis of an Interior PMSM (IPMSM). By measuring the resulting high-frequency current response in the $q$-axis, which varies due to the rotor’s magnetic differences ($L_d \neq L_q$), the algorithm finds the exact rotor position down to zero speed without using Back-EMF.

   d-axis HF Voltage Injection (5-20V, 500Hz-2kHz) 
                         |
                         v
                [IPMSM Rotor Saliency] (Ld != Lq)
                         |
                         v
     Analyze q-axis HF Current Response ---> Extract Rotor Position

The Theory of HFI

For vehicles attempting sensorless control at zero speed, standard BEMF estimation does not work. Instead, High-Frequency Injection (HFI) takes advantage of the structural differences (magnetic saliency) of Interior Permanent Magnet Synchronous Motors (IPMSM), where $d$-axis inductance ($L_d$) differs from $q$-axis inductance ($L_q$).

By injecting a high-frequency carrier voltage into the estimated $d$-axis:$$V_{hfi} = V_m \cdot \sin(\omega_h \cdot t)$$

The magnetic variation produces a corresponding high-frequency current in the $q$-axis. Demodulating this current response yields an error signal proportional to the angular difference between the estimated and actual rotor positions:$$\Delta\theta = \theta_{actual} – \theta_{estimated}$$

A tracking observer (using a Phase-Locked Loop) drives this error to zero, allowing position tracking down to $0\text{ rpm}$.

The Performance Trade-Offs

HFI has real-world engineering downsides:

1. Audible Noise (NVH): Injecting a $500\text{ Hz–}2\text{ kHz}$ voltage wave vibrates the motor structure, generating a high-pitched metallic whine that passengers can hear.
2. Added Losses: The high-frequency excitation increases iron losses and copper losses, slightly lowering efficiency.

Hybrid Position Estimation Strategy

To deliver acceptable quietness, production vehicles use a dual estimation strategy:

• Below $50\text{ rpm}$ ($< 5\text{ km/h}$): HFI is fully active to ensure stable torque and prevent rollback.
• Above $50\text{ rpm}$: The controller transitions from HFI to a standard BEMF sliding-mode observer (SMO) or active flux observer.
• Transition Blending: During the transition speed window, a weighted averaging filter blends the position angles from both observers to prevent any sudden jump in estimated angle, which would trigger a torque shock.

8. Real-world Engineering Calibration & Debugging Guide

Common Creep Issues & Solutions Quick-Reference Table

This troubleshooting matrix details the most common low-speed creep issues, their root causes, and calibration fixes:

No.	Issue Symptom	Root Cause	Tuning / Calibration Direction
1	Creep Speed Too High	High base creep torque settings, or the exit speed limit is too high.	• Reduce the peak static torque $T_{max}$ in the VCU parameters.• Lower the creep exit speed threshold $v_{exit}$ to narrow the creep window.
2	Launch Jerk (Activation Shock)	Torque ramp-up is too sharp, or there is a poor transition between Standby and Active states.	• Increase the activation ramp-up filter time constant (e.g., increase from $200\text{ ms}$ to $400\text{ ms}$).• Reduce the maximum allowed torque rate of change ($dT/dt$).
3	Brake-Release Jerk	The recovery slope of the creep torque is too steep when the driver releases the brake pedal.	• Decrease the torque recovery ramp slope (use an asymmetric curve with a gentler upward slope upon brake release).
4	Hill Rollback	• Noisy pitch angle data from the IMU.• Under-compensation of the slope torque ($T_{slope}$).	• Adjust the IMU low-pass filter to remove vibration noise.• Increase the slope compensation multiplier.• Calibrate a small dynamic forward torque offset.
5	Low-Speed Torque Ripple (Vibration)	• Inverter dead-time compensation is off or poorly tuned.• High sensor noise causing current-loop instability.	• Turn on or tune the SVPWM dead-time compensation lookup table ($V_{comp}$ vs $V_{dc}$).• Lower the current loop PI controller proportional gain ($K_p$) at speeds below $100\text{ rpm}$.
6	Tip-In / Tip-Out Shock	Poor torque handover between the VCU creep controller and the primary accelerator pedal map.	• Adjust the transition algorithm.• Pass the difference ($\vert{}T_{creep} – T_{driver}\vert{}$) through a cross-fader function during handover.

The Importance of High-Rate Logging

A common error in low-speed calibration is relying on standard CAN bus logs. Powertrain CAN messages usually transmit at $10\text{ ms}$ or $100\text{ ms}$ intervals. Driveline shocks and current loop oscillations occur in short windows of $50\text{–}200\text{ ms}$.

At a $100\text{ ms}$ CAN sampling rate, these phenomena look like minor noise. To find the source, engineers must use high-rate internal DAQ tools (like CCP/XCP over Ethernet or high-speed USB-to-CAN) to record MCU variables (including phase currents $i_{a,b,c}$, current commands $i_{d,q}$, actual rotor angle $\theta$, and PWM duty cycles) at $1\text{ kHz}$ or higher.

[CAN Bus Log @ 100ms]      *-------------------*-------------------* (Misses transients)

[MCU Internal Log @ 1kHz]  *-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-* (Captures 50ms shocks)

9. Summary of Key Technical Insights

Creep control is an engineering challenge that is simple in concept but complex in practice. Getting a production-grade low-speed creep experience requires optimization across multiple areas:

• The VCU must manage a robust, multi-state state machine and apply precise, slope-aware torque-blending calculations to keep the vehicle stable.
• The MCU must run advanced FOC, applying real-time dead-time voltage compensation and dynamic current loop filtering to maintain stable torque at low speeds.
• The Calibration Engineer must balance these algorithms using high-rate logging tools to eliminate the margin between a smooth ride and a clunky, jerky one.

As vehicle drivetrains move toward centralized domain controllers and drive-by-wire platforms, creep control will become even smarter—adjusting its base torque maps by predicting traffic and reading cloud map terrain data.

10. References & Further Reading

These resources cover the research and patent foundations of low-speed creep and motor control:

• [1] US Patent 9873353: System and method for controlling creep torque of an electric vehicle. Covers VCU brake-to-creep torque transition curves and active decay functions.Link to Patent PDF
• [2] US Patent 9308831: System and method for controlling vehicle creep torque. Focuses on slope-detection logic and anti-rollback safety features.Link to Patent PDF
• [3] EVreporter Technical Analysis: Field-Oriented Control of PMSM in Electric Vehicles. Explains FOC current loops and low-speed coordinate transformations.Link to Article
• [4] ScienceDirect (2022): Development of a PMSM motor field-oriented control algorithm for electrical vehicles. Academic research on low-speed state estimation and motor control loop stability.Link to Paper
• [5] CSDN Engineering Blog: VCU Torque Management and Creep Calibration Strategies. Hands-on software guide covering state transition code.Link to Article