Data Types Reference

Source: fes_framework/data/types.py

BiomechanicalData#

29 joint angles in degrees. Updated each cycle by DataStreamer (when set_biomech_collection(True) is in effect). Read via self.joints in your strategy.

PascalCase fields mirror the Teslasuit SDK enum names; the framework's parse_ctypes_to_dataclass matches by name.

All fields default to 0.0. Sign convention follows the Teslasuit SDK; check the SDK docs for details.

StepDetectorData#

Foot contact state.

Source: SDK's built-in step detector (foot-IMU acceleration threshold). Read via self.contacts.

EMSParamData#

Single muscle's stimulation parameters.

IsMuted defaults to True so unwritten EmsData() slots are silent — see the docstring in types.py for why.

EmsData#

EMS parameters for all 20 muscles. Each field is an EMSParamData. Default factory yields a muted instance. Write via self.ems_output.<muscle> = EMSParamData(...).

Lower body (10)#

Upper body (10)#

These names come from muscle_map_4R.json. To add a muscle, edit both this dataclass and the JSON config in lockstep.

EMSCalibrationData#

Same 20 muscle slots as EmsData, each holding an EmsMultiplierData(minimum: float, maximum: float). Used by applications that calibrate per-user comfortable amplitude ranges.

The base framework instantiates one (engine.ems_calibration_data) but does not populate or consume it; applications can use it as a shared store.

ControlMessage (subclass to use)#

Empty base class. Application defines fields by subclassing:

@dataclass
class MyControlMessage(ControlMessage):
    threshold_deg: float = 15.0
    quad_amplitude: int = 40

Carries an _on_change callback (private, not serialised across processes) for auto-send patterns. See Messaging.

UtilityMessage#

Framework-defined system flags. Do not subclass.

_on_change callback is supported, same as ControlMessage.

ProcessedData#

20 bones, each with position and rotation. Used by visualisation / LSL outlets. Most strategies don't read this directly.

ProcessedData(
    Hips:           MocapBoneData,
    LeftUpperLeg:   MocapBoneData,
    RightUpperLeg:  MocapBoneData,
    LeftLowerLeg:   MocapBoneData,
    RightLowerLeg:  MocapBoneData,
    LeftFoot:       MocapBoneData,
    RightFoot:      MocapBoneData,
    Spine:          MocapBoneData,
    Chest:          MocapBoneData,
    UpperChest:     MocapBoneData,
    Neck:           MocapBoneData,
    Head:           MocapBoneData,
    LeftShoulder:   MocapBoneData,
    RightShoulder:  MocapBoneData,
    LeftUpperArm:   MocapBoneData,
    RightUpperArm:  MocapBoneData,
    LeftLowerArm:   MocapBoneData,
    RightLowerArm:  MocapBoneData,
    LeftHand:       MocapBoneData,
    RightHand:      MocapBoneData,
)

MocapBoneData has position: BonePosition(x, y, z) and rotation: BoneRotation(w, x, y, z).

RawData#

20 bones, each with raw IMU readings. Plus a timestamp: int = 0.

Each segment is a SensorData:

Same 20 bones as ProcessedData.

HeartRateData#

Populated only when DataStreamer.ppg_available is True.

HRVData#

Heart-rate variability metrics, populated only when PPG is available.

RawPPGData#

Raw photodiode readings. PPG runs at ~200 Hz; the framework keeps the most recent 2 samples per cycle.

CustomPlayable#

A single named slot in a HapticLibrary. Holds the SDK playable ID, mute state, looped flag, and three live-modulation multipliers.

Populated via SuitHandler.create_haptic_touch() or SuitHandler.load_haptic_asset().

HapticLibrary (subclass to use)#

Empty base class. Subclass to declare named CustomPlayable slots.

@dataclass
class ArmHaptics(HapticLibrary):
    biceps_burst:  CustomPlayable = field(default_factory=CustomPlayable)
    reach_forward: CustomPlayable = field(default_factory=CustomPlayable)

HapticLibrary.iter_playables() yields (name, CustomPlayable) pairs for every slot. Used internally by LibraryStimulator.

Vector dataclasses#

Used inside SensorData and MocapBoneData. Mostly internal — you read them off self.suit_handler.streamer.processed_data if you need them; strategies usually don't.

shared_memory_frame (numpy dtype)#

Pre-built numpy structured dtype for SharedRingBuffer. Includes timestamp + the 29 BiomechanicalData fields + the 2 step detector flags + backend_sample_rate. Total: 33 fields.

Use it for the standard biomech + step-detector layout. Build your own if you need more fields (see examples/walking_fes/walking_types.py).

import numpy as np
from fes_framework.data.types import shared_memory_frame
buffer = SharedRingBuffer(name="fes", frame_dtype=shared_memory_frame, capacity=1000)

See also#

• Data Types — concept-level overview
• API Reference — class signatures
• Source: fes_framework/data/types.py