Hardware profiles¶
A hardware profile YAML describes one physical device. Curves and limits in the profile drive the subsystem physics; code reads the profile, never the other way round (ADR-0003).
Numeric values in a profile must trace back to a row in the Bill of Materials. New components land in the BOM first, then in a profile YAML.
Reference profile¶
profiles/jetson-agx-orin.yaml is the canonical Jetson AGX Orin 64GB
reference. Battery capacity, Peukert exponent, thermal limits, compute
power across load fractions, sensor noise stddevs, and comms link
envelopes live there.
Other profiles¶
| Profile | Notes |
|---|---|
jetson-orin-nx.yaml |
Smaller compute module, smaller battery. |
pi5-hailo.yaml |
Raspberry Pi 5 with a Hailo-8L accelerator. |
spot-core.yaml |
Boston Dynamics Spot's compute core. |
Schema¶
schema_version: "0.1.0"
name: jetson-agx-orin
description: ...
power:
battery_wh: 588
voltage_v_nominal: 14.4
voltage_v_min: 12.0
voltage_v_max: 16.8
internal_resistance_ohm: 0.05
rated_current_a: 5.0
peukert_k: 1.04
soc_pct_low_threshold: 20
soc_pct_critical_threshold: 5
thermal_derate_c: 45
thermal_derate_slope_per_c: 0.02
charge_limit_w: 100
apu:
solar:
panel_w_peak: 60
mppt_efficiency: 0.92
panel_temp_derate_per_c_above_25: 0.004
fuel_cell:
continuous_w: 25 # SFC EFOY-class methanol fuel cell, datasheet
fuel_capacity_g: 790 # ~1 L methanol cartridge (density 0.79 g/ml)
efficiency: 0.25 # SFC EFOY system efficiency (post-BOP)
wh_per_g_fuel: 1.4 # derived from efficiency * 5.53 (methanol LHV) if omitted
vehicle:
bus_voltage_v: 28.0
current_limit_a: 5.0
usb_c_pd:
profiles_w: [15, 27, 45, 60, 100]
default_profile_w: 60
thermal:
ambient_c_default: 25
junction_temp_max: 95
junction_temp_throttle: 85
thermal_resistance_c_per_w: 0.3 # junction-to-enclosure
enclosure_to_ambient_resistance_c_per_w: 0.5 # enclosure-to-ambient (optional, default 0.5)
enclosure_mass_kg: 1.2
enclosure_specific_heat_j_per_kg_k: 900
junction_heat_capacity_j_per_k: 5.0 # optional, default 5.0
headroom_threshold_c: 5.0 # SC-2 guard threshold
compute:
draw_w_idle: 8
draw_w_load: 60
load_curve:
- { load_pct: 0, draw_w: 8 }
- { load_pct: 50, draw_w: 35 }
- { load_pct: 100, draw_w: 60 }
sensors:
position:
lat_sigma: 3.0e-5
lon_sigma: 3.0e-5
alt_m_sigma: 5
...
comms:
links:
- id: lte
bandwidth_bps: 20_000_000
rssi_dbm_nominal: -75
loss_pct_nominal: 0.5
The APU section also accepts the legacy flat fields solar_w_peak,
fuelcell_w_continuous, fuelcell_fuel_capacity_g, and
fuelcell_efficiency for backward compatibility; the nested form is
preferred for new profiles.
L1 ships a Pydantic schema model (src/nous/profiles/schema.py) and a
JSON Schema artefact under docs/schema/. The make schema target
regenerates both.