Internal Performance Benchmarks
Internal benchmarks and Safety Suite results from our reference hardware. Pilot KPIs are measured on your platform and your data.
Live System Performance
Real-time metrics from our production system analyzing Tesla Autopilot footage and live traffic scenarios.
Live System Status
Internal R&D Benchmarks
Performance metrics from our reference hardware (256×64 grid, CPU-class compute, 40 Hz). Pilot KPIs are measured on your platform.
| Metric | Measured Value | Conditions | Notes |
|---|---|---|---|
| Core PDE Step | ~0.20ms | 256×64 grid, Numba | Traffic field update |
| Safety Observer E2E (avg) | ~1.7ms | 40 Hz, CPU | Full pipeline |
| Safety Observer p99 | <3ms | 40 Hz, CPU | 99th percentile |
| Safety Observer p99.9 | <4ms | 40 Hz, CPU | 99.9th percentile |
| Mean Lead Time | +0.62s | 5 synthetic scenarios | vs baseline |
| Prediction Horizon | 0.5–2.0s | Configurable | Short-horizon risk |
Internal Scenario: Debris Detection
Example from our Internal Safety Suite showing earlier hazard detection in a synthetic debris scenario.
Scenario: Garbage Truck Debris
A garbage truck ahead loses debris into the ego lane. The physics engine detects the hazard field before the debris is fully visible to camera-based object detection.
SCENARIO PARAMETERS
- • Speed: 65 mph highway
- • Debris type: Large cardboard box
- • Weather: Clear daylight
- • Source: Synthetic Safety Suite
Detection Timing
Object appears in camera frame
Hazard field detects motion anomaly
Note: This is an internal synthetic scenario (1 of 5 in our Safety Suite). Mean lead time across all 5 scenarios: +0.62s. Pilot programs measure on your data.
Latency & Performance Metrics
Internal benchmarks from our reference hardware and Safety Suite validation.
Safety Suite Scenarios
Internal synthetic Safety Suite validation
Latency Distribution
@ 40 Hz, 256×64 grid, CPU-class hardware
Core PDE Performance
Numba-accelerated on CPU
Comprehensive Virtual Testing Infrastructure
Complete validation framework proving FieldSpace is ready for OEM partnerships and regulatory submission with industry-standard testing protocols.
🧪 FieldSpace Virtual Testing Suite
✅ PERFORMANCE BENCHMARKS: PASSED
⚠️ OEM INTEGRATION: API COMPATIBILITY ISSUES DETECTED
CARLA Simulation
NHTSA Compliance
OEM Integration
Performance Testing
Technical Advantages
Physics-based approach delivers measurable advantages over neural network competitors.
Why Physics Wins
Deterministic Inferences
Same inputs produce the same HazardObjects and cost maps every time
Long-Tail Focus
Designed to improve robustness in debris-like, sliding cargo, and occluded scenarios
Explainable Outputs
Structured HazardObjects and cost maps that are straightforward to log and audit
Low Latency
Under 4ms p99.9 at 40 Hz on CPU-class hardware, leaving compute for your main stack
Internal Performance Metrics
Reference hardware: Intel i7 / NVIDIA Orin
Measure KPIs on Your Data
Our 4-week pilot program uses a standard Safety Suite and evaluation harness to compute metrics like lead time, false positives, and added latency on your logs.
~1.7ms avg latency. +0.62s mean lead time. Deterministic outputs.