n.Tec‑5 — BACK‑END

WORK IN PROGRESS BY: H. Pandit

n.Tec-5 Back-End Documentation

n.Tec-5 is a modular, AI-driven engine tuning and control platform. It functions like a custom Electronic Control Unit (ECU) that manages engine performance through adaptive calibration updates based on real-time sensor data. Inspired by MoTeC-style systems, this back end forms the brain of the tuning architecture.

---

1. Overview and Goals

• Adaptive Engine Calibration: AI-guided updates to boost/fuel/ignition maps.
• Modular Structure: Independent modules for tuning, detuning, lambda control, and aerodynamics.
• Configuration via YAML: Tuning parameters, thresholds, and targets are externally configurable.
• Diagnostics & Logging: Safe detuning on part degradation, with full traceable logs.
• Cross-Platform Deployment: Runs on PC, Raspberry Pi, or embedded Linux.

---

2. Architecture

Major Modules

• BaseMap: Holds calibration parameters (fuel_map, boost_map, etc.).
• Tuner: Applies gradient updates (+/-ε) based on AI tuner decisions.
• AITuner: Neural net decision engine; predicts tuning direction (+1, 0, -1).
• Detuner: Applies safe detune steps based on part health (e.g. reduce boost if turbo degraded).
• ActiveLamdaController: Adjusts lambda (AFR) setpoint based on sensor feedback.
• AeroController: Controls DRS and brake stability from speed/wheel data.

Real-Time Data Loop

1. Collect sensor input
2. Predict adjustment via AITuner
3. Tune or detune parameters
4. Update aero and lambda targets
5. Log state & write back updated map

---

3. Real-Time Tuning Algorithms

Gradient Tuning

new_val = tuner.apply_gradient_increment("fuel_map", direction=+1)

Detuning

part_status = {'turbocharger': True}
params = detuner.check_part_degradation(part_status)

AI Tuner

input_vec = [steering, throttle, accel, brake, lambda_err]
direction = ai_tuner.predict_adjustment(input_vec)

# • +1 = enrich/increase
# • -1 = lean/decrease
# • 0  = no change

---

4. Base Map & Lambda Controller

Base Map (YAML Example)

fuel_map: 1.0
boost_map: 1.0

Lambda Control

lamda_ctrl = ActiveLamdaController(target_lambda=1.0, adjustment_step=0.01)
new_target = lamda_ctrl.update_lambda(current_lambda=1.1)

• If AFR too lean → enrich
• If AFR too rich → lean out

---

5. Configuration (YAML)

configs/tuning_config.yaml:

gradient_step: 0.01
detune_gradient_step: 0.01
lambda_target: 1.0
lambda_adjustment_step: 0.01
tolerance: 0.05

Load Config Example:

with open("configs/tuning_config.yaml", 'r') as f:
    cfg = yaml.safe_load(f)

---

6. Sensor Input Integration

Simulated Example

sensor_data = {
  'steering': 0.5,
  'throttle': 0.7,
  'lambda': 1.05,
  'vehicle_speed': 100.0,
  'wheel_speeds': [90, 91, 89, 92],
}

Real Sensor Integration (via CAN)

msg = can_bus.recv()
if msg.arbitration_id == SPEED_CAN_ID:
    sensor_data['vehicle_speed'] = decode_speed(msg.data)

---

7. Aero Control System

DRS Logic

drs_active = aero_controller.update_drs(vehicle_speed=100, lap_time=60, race_mode=True)

Brake Stability Logic

adjustments = aero_controller.update_braking_stability([90, 92, 88, 91])

---

8. Logging and Output

All modules log decisions and map changes. Example output:

[Tuning] 'fuel_map' adjusted to 1.01  
[Detune] 'boost_map' detuned to 0.95  
[Lambda] target lambda=0.990  
[Aero] DRS=Active, BrakeAdjust=0.97  

---

9. System Integration

The n.Tec-5 Back-End can integrate with:

• MoTeC dash or ECU
• Python front-end GUI
• Embedded control loop on vehicle
• Logging servers (via REST/WebSocket)

The back end operates headless and exposes internal states via logs or API. It is designed for robustness in motorsport or simulation environments.

---

10. Credits

WORK IN PROGRESS BY: H. Pandit
Part of the n.Tec-5 Performance AI Tuning Suite.

---

https://github.com/user-attachments/assets/66abe772-5801-432f-b7e9-d87fa1156551 https://github.com/user-attachments/assets/2193e625-aed2-4b76-ac46-6499485f27ee