Lateral Control Tuning

Overview

Lateral control manages steering to keep your vehicle centered in the lane. sunnypilot offers multiple lateral control methods with extensive tuning capabilities for optimal performance across different vehicles and driving conditions.

Most vehicles work well with default settings. Advanced tuning is optional and should only be attempted if you understand the implications.

Lateral Control Methods

Torque Control (Default)

Torque-based lateral control directly commands steering torque for precise path following.

EnforceTorqueControl

boolean

default:"false"

Force use of Torque lateral control even if other methods are available

Key characteristics:

Direct torque commands to EPS (Electric Power Steering)
Faster response time than angle-based control
Better handling of dynamic conditions
Requires vehicle-specific tuning

Torque Control Versions:

TorqueControlTune

select

default:"Default"

Select torque control algorithm version:

Default - Latest production tune (recommended)
v1.0 - Current stable release
v0.0 - Legacy tune for compatibility

Neural Network Lateral Control (NNLC)

Formerly known as NNFF, this advanced method uses neural networks trained on vehicle-specific driving data.

NeuralNetworkLateralControl

boolean

default:"false"

Enable NNLC for enhanced steering accuracy using learned vehicle behavior

Benefits:

Trained on real driving data for each EPS firmware
Increased control accuracy
Better adaptation to vehicle-specific characteristics
Continuously improving with more data

NNLC requires vehicle-specific models. Contact @twilsonco on Discord to contribute data for your vehicle if unsupported.

Angle Control

Used by some platforms (Ford, Nissan, Tesla) that accept steering angle commands rather than torque.

Torque control cannot be enforced on angle control platforms and will cause “Controls Unresponsive” errors.

Torque Control Tuning

Self-Tune (Automatic)

The recommended approach for most users:

LiveTorqueParamsToggle

boolean

default:"true"

Enable automatic self-tuning of torque parameters

LiveTorqueParamsRelaxedToggle

boolean

default:"false"

Use less restrictive settings for self-tune learning (Beta)

How it works:

Enable Self-Tune in Settings → Steering
Drive normally for several hours
Parameters automatically adapt to your vehicle
Monitor LiveTorqueParameters for learned values

Learning process:

Analyzes steering response patterns
Adjusts friction compensation
Tunes lateral acceleration factors
Refines based on road conditions

Manual Tuning (Advanced)

For experienced users who want direct control:

CustomTorqueParams

boolean

default:"false"

Enable custom torque parameter tuning interface

TorqueParamsOverrideEnabled

boolean

default:"false"

Allow real-time manual parameter adjustments

When enabled, you can adjust:

Friction Parameter

TorqueParamsOverrideFriction

float

default:"0.01"

Friction compensation factor

Range: 0.0 - 1.0
Step: 0.01
Effect: Higher values = more aggressive friction compensation

What it controls:

Steering system mechanical friction
Low-speed steering resistance
Center position holding force

Tuning guidance:

Too low: Sluggish response, poor centering
Too high: Twitchy steering, oscillation
Typical range: 0.01 - 0.05

Lateral Acceleration Factor

TorqueParamsOverrideLatAccelFactor

float

default:"1.0"

Lateral acceleration response scaling

Range: 0.1 - 5.0
Step: 0.1
Effect: Scales torque response to lateral acceleration demands

What it controls:

Steering sensitivity to curves
Turn entry aggressiveness
Lane change responsiveness

Tuning guidance:

Below 1.0: Gentler steering, slower lane changes
Above 1.0: More aggressive steering, faster response
Typical range: 0.8 - 1.5

Steering Delay Tuning

Live Learning Steering Delay

LagdToggle

boolean

default:"true"

Enable automatic learning of vehicle’s steering response delay

Benefits:

Adapts to your vehicle’s EPS characteristics
Compensates for actuator delay
Improves path tracking accuracy
Reduces lag in steering response

Manual Delay Override

LagdToggleDelay

float

default:"0.15"

Manual software delay when live learning is disabled

Range: 0.05 - 0.5 seconds
Step: 0.01 seconds

When to adjust:

Steering feels delayed or laggy
Path tracking oscillates
Lane changes feel imprecise

Tuning process:

Disable LagdToggle
Start with 0.15s
Decrease if steering lags behind path
Increase if steering oscillates
Test on highway and curves

Advanced Features

Camera Offset

CameraOffset

float

default:"0.0"

Virtually shift camera perspective horizontally

Range: -0.35 to 0.35 meters
Step: 0.01 meters
Positive: Shift left
Negative: Shift right

Use cases:

Adjust for off-center camera mounting
Compensate for wide vehicles
Fine-tune lane positioning

Camera offset affects model perception. Only adjust if necessary and in small increments.

Lane Turn Desire

LaneTurnDesire

boolean

default:"false"

Force model to plan turn intent based on blinker activation

LaneTurnValue

integer

default:"10"

Maximum speed for lane turn desire activation

Range: 0 - 20 mph
Step: 1 mph

Behavior:

When blinker is on below threshold speed
Model anticipates turn maneuver
Adjusts path planning accordingly
Useful for complex intersections

Blinker Lateral Control

BlinkerPauseLateralControl

boolean

default:"false"

Pause lateral control when turn signal is active

BlinkerMinLateralControlSpeed

float

default:"32"

Minimum speed to pause lateral control with blinker

BlinkerLateralReengageDelay

float

default:"0.0"

Delay before lateral control resumes after blinker endsAllows driver to complete maneuver before re-engagement

Internal Tuning Constants

These values are defined in latcontrol_torque_ext_base.py and affect NNLC behavior:

# Friction compensation tuning (source/sunnypilot/selfdrive/controls/lib/latcontrol_torque_ext_base.py)
friction_look_ahead_v = [1.4, 2.0]      # Lookahead time in seconds
friction_look_ahead_bp = [9.0, 30.0]    # Corresponding speeds in m/s

lat_jerk_friction_factor = 0.4          # Jerk-based friction scaling
lat_accel_friction_factor = 0.7         # Accel-based friction scaling (0-3 range)

PID Constants

# From latcontrol_torque_ext_base.py:18-21
KP = 0.8                                 # Proportional gain
KI = 0.15                                # Integral gain
INTERP_SPEEDS = [1, 1.5, 2.0, 3.0, 5, 7.5, 10, 15, 30]  # Speed breakpoints (m/s)
KP_INTERP = [250, 120, 65, 30, 11.5, 5.5, 3.5, 2.0, KP]  # Speed-dependent KP

Internal constants require modifying source code. Most users should use the parameter-based tuning options.

Monitoring and Diagnostics

Developer UI

Enable Developer UI to monitor lateral control in real-time:

Settings → Device → Developer UI

Key metrics:

E.T. - EPS Torque (current steering torque)
L.A. - Lateral Acceleration (m/s²)
FRI. - Friction compensation value
Desired/Actual curvature - Path following accuracy

Parameter Values

Check learned torque parameters:

from openpilot.common.params import Params
import json

params = Params()
live_params = json.loads(params.get("LiveTorqueParameters"))
print(f"Friction: {live_params.get('friction')}")
print(f"Lat Accel Factor: {live_params.get('latAccelFactor')}")

Troubleshooting

Steering Oscillates

Causes:

Friction too high
Lat accel factor too high
Steering delay incorrect

Solutions:

Reduce friction by 0.01-0.02
Decrease lat accel factor by 0.1
Check tire pressure
Enable live learning delay

Poor Lane Centering

Causes:

Friction too low
Camera calibration off
Model confidence low

Solutions:

Increase friction by 0.01
Recalibrate camera
Enable self-tune for 20+ drives
Check camera mount stability

Delayed Steering Response

Causes:

Steering delay too high
Lat accel factor too low
EPS lag

Solutions:

Reduce LagdToggleDelay by 0.02s
Increase lat accel factor by 0.1
Enable live learning delay
Check vehicle-specific tuning

Controls Unresponsive Error

Causes:

Torque control enforced on angle platform
Parameter value out of range
EPS communication fault

Solutions:

Disable EnforceTorqueControl
Reset custom parameters
Check panda connection
Review error logs

Best Practices

Start with Defaults

Use stock settings for at least 10-20 drives to establish baseline

Enable Self-Tune

Let the system learn your vehicle automatically before manual tuning

Change One Parameter

Only adjust one parameter at a time to isolate effects

Test Thoroughly

Drive in various conditions (highway, curves, city) after changes

Document Changes

Keep notes on parameter values and their effects

Parameters System

Understanding parameter management

Longitudinal Control

Speed and acceleration tuning

NNLC Setup

Neural network lateral control guide

Vehicle Tuning

Vehicle-specific tuning tips

Configuration

Development

Car Porting

Lateral Control Tuning

Overview

Lateral Control Methods

Torque Control (Default)

Neural Network Lateral Control (NNLC)

Angle Control

Torque Control Tuning

Self-Tune (Automatic)

Manual Tuning (Advanced)

Friction Parameter

Lateral Acceleration Factor

Steering Delay Tuning

Live Learning Steering Delay

Manual Delay Override

Advanced Features

Camera Offset

Lane Turn Desire

Blinker Lateral Control

Internal Tuning Constants

PID Constants

Monitoring and Diagnostics

Developer UI

Parameter Values

Troubleshooting

Best Practices

Parameters System

Longitudinal Control

NNLC Setup

Vehicle Tuning

Build docs developers (and LLMs) love

Configuration

Development

Car Porting

​Overview

​Lateral Control Methods

​Torque Control (Default)

​Neural Network Lateral Control (NNLC)

​Angle Control

​Torque Control Tuning

​Self-Tune (Automatic)

​Manual Tuning (Advanced)

​Friction Parameter

​Lateral Acceleration Factor

​Steering Delay Tuning

​Live Learning Steering Delay

​Manual Delay Override

​Advanced Features

​Camera Offset

​Lane Turn Desire

​Blinker Lateral Control

​Internal Tuning Constants

​PID Constants

​Monitoring and Diagnostics

​Developer UI

​Parameter Values

​Troubleshooting

​Best Practices

​Related Resources

Parameters System

Longitudinal Control

NNLC Setup

Vehicle Tuning

Build docs developers (and LLMs) love

Overview

Lateral Control Methods

Torque Control (Default)

Neural Network Lateral Control (NNLC)

Angle Control

Torque Control Tuning

Self-Tune (Automatic)

Manual Tuning (Advanced)

Friction Parameter

Lateral Acceleration Factor

Steering Delay Tuning

Live Learning Steering Delay

Manual Delay Override

Advanced Features

Camera Offset

Lane Turn Desire

Blinker Lateral Control

Internal Tuning Constants

PID Constants

Monitoring and Diagnostics

Developer UI

Parameter Values

Troubleshooting

Best Practices

Related Resources