Radar Physics Engine v4.2

C-Shenron Deep Architecture

From Raw LiDAR Scents to Synthetic FMCW Realities

LiDAR RCS ADC PCD
01
Scene Ingestion

Converts raw Semantic LiDAR into physical material properties. Vehicles become High-Permittivity Metal, Foliage becomes Diffuse Carbon.

map_carla_semantic_lidar_latest()
→ Recovered Tags via .view(np.uint32)
02
Physics Modeling

Calculates the Fresnel Reflection loss based on incident angles. Handles Specular (mirror) vs Diffuse (scattering) coefficients.

Sceneset.specularpoints()
Pr = (Pt * G² * λ² * σ) / ((4π)³ * R⁴)
03
FMCW Synthesis

Synthesizes raw Analog-to-Digital samples. Models the Beat Frequency ($f_{beat}$) across multiple Rx antennas & chirps.

heatmap_gen_fast.py
signal = exp(j * 2π * f_beat * t)
04
DSP Pipeline

Processes the raw signal through Range/Doppler FFTs and CA-CFAR Detection to generate the final synthetic PointCloud.

RadarProcessor.cal_doppler_fft()
N=256, Np=128 (Configurable)

⚡ Today's Optimization Discoveries

Summary of the hardware-to-model alignment fixes applied in Iteration 05.

Detected Failure

The 'Cos(Cos)' Bug in Sceneset.py. The engine was calculating np.cos(angles) on values that were already cosines. Result: 46% energy loss on every bounce.

Engineering Fix

Removed the double-dampening logic. Physics engine now correctly interprets incident angles from CARLA, restoring realistic metallic RCS signatures.

Detected Failure

The 'Stationary Reality' Bug. Moving targets appeared stationary or dragged behind the car. Result: Radial velocity was being zeroed out due to incorrect bit-interpretations.

Engineering Fix

Synchronized World-Pose and LiDAR-Tick via recorder.py update. High-fidelity Doppler detection now tracks dynamic NPC movement at 64Hz.

🛠️ Hardware Support Matrix

Radar Model Frequency Chirps (Np) Samples (N) Logic Layer
AWRL1432 BOOST 77-81 GHz 128 256 Rich Physics
TI Cascade Array 77-81 GHz 3 (Fast) 256 Phase-Prime
Generic Radarbook 24 GHz 128 256 Standard