Deterministic Full-Stack Autonomy
FieldSpace is a deterministic end-to-end autonomy stack (perception → prediction → planning → control), engineered for OEM adoption with contracts, validation suites, and audit-ready traces.
Deterministic by design
FieldSpace is a complete deterministic stack, validated by suites and regression gates. It can run as your primary autonomy stack, or it can be deployed as a safety observer (fused objects → hazards) during phased OEM rollout.
FieldSpace Safety Observer
Deterministic safety prediction engine that consumes fused objects and outputs hazards + risk surfaces with traceable triggers.
- ✓Works with existing sensor suite—no retraining
- ✓Deterministic: same input → bit-identical output
- ✓Structured hazard outputs for your planner constraints
- ✓Regulator-friendly: replayable, explainable traces
Validation + Virtual Testing
Deterministic test harnesses and virtual suites (compliance + OEM integration + performance regression) to ship safely and prove it.
- ✓Safety Suite scenarios (cut-in, sudden braking, pedestrian step-out…)
- ✓Virtual testing (CARLA, NHTSA checks, OEM interface tests)
- ✓Performance + stability gates (FPS, latency, memory growth)
- ✓Continuous regression detection and reporting
Deterministic, low-latency safety
Under a 40 Hz budget, FieldSpace stays in the single‑digit millisecond range while providing explainability and strict contracts.
Reference autonomy stack (engineering harness)
For end‑to‑end validation, the repo includes a reference autonomy backend (planning + control + safety monitoring) with real-time WebSocket updates.
Autonomy API (8001)
Initialize sessions, engage/disengage autonomy, request trajectories, and stream updates over WebSocket.
- • HTTP: /api/l5/*
- • WS: /ws/l5/{session}
- • Safety: violations + scores
Perception API (8000)
Video analysis, autoscore, live analysis, fusion mode, and full-stack orchestrator endpoints.
- • /api/analyze, /api/autoscore(_async)
- • /api/live-analysis/*, /api/fusion/*
- • /api/fullstack/*
Deterministic testing
Unit/integration/perf tests plus virtual suites to validate OEM readiness before road testing.
- • `run_tests.py` + regression gates
- • `run_virtual_tests.py` suite
- • Safety Suite v1 scenarios
Proven Results
Nexus Platform delivers measurable operational improvements across diverse mobility operations.
Operational cost reduction
Fleet utilization increase
Concurrent users supported
p95 response time
Integration & Requirements
Everything you need to integrate FieldSpace into your autonomous stack.
Required Inputs
- ●Ego poseVehicle position and orientation
- ●Ego velocityLinear and angular velocity
- ●Camera frames or fused objectsRaw images or perception outputs
- ●System timeSynchronized timestamp
Optional Inputs
- • Perception confidences
- • Camera calibration
- • Object tracks / IDs
System Requirements
- CPU4+ cores @ 2.0 GHz
- GPUOptional (CPU-only mode)
- Memory4 GB RAM
- Latency Budget<4ms p99.9
- Update Rate40 Hz
- NetworkROS 2 / gRPC
HazardObjects API
Structured output for your planner at 40 Hz.
{
"timestamp": 1701234567.89,
"hazards": [
{
"id": "h_001",
"type": "moving_object",
"risk_score": 0.85,
"ttc": 2.3,
"position": [12.5, -3.2],
"velocity": [5.0, 0.5],
"cone": {...}
}
],
"cost_map": {
"grid_size": [256, 64],
"resolution": 0.25,
"data": "base64..."
}
}• Frequency: 40 Hz
• Latency: ~1.7ms avg
• Cost maps compatible with existing planners
Ready to evaluate a deterministic full stack?
Run the full deterministic stack on your logs and scenarios, and produce replayable metrics and traces for safety reviews. Deploy as the primary stack or stage adoption via observer/shadow mode.
Engineering-first: determinism, contracts, validation suites.