APACC Cross-Domain Validation Data (Automotive/Railway)
Description
Validation datasets for Artificial Precognition Adaptive Cognised Control (APACC), a neuro-symbolic MIMO control architecture for safety-critical autonomous systems (US Patent 9,645,576 B2). Automotive domain: 10,000 trials across 8 scenario types (urban intersection, pedestrian crossing, highway merging/overtaking, emergency avoidance, adverse weather) using CARLA simulator with Honda Civic parameters. Compares APACC against PID, MPC, DRL, and Hybrid controllers. Key result: 51% reduction in peak deceleration vs PID baseline. Railway domain: 168 hours continuous deployment on UK Network Rail infrastructure (Workington–Sellafield route, 32 km) under Government SBRI Project EDGE. Includes 6,026 telemetry records across 4 MNOs (EE, Vodafone, Three, O2) with Iridium satellite backup, plus 256 handover events. Key result: 94.9% handover prediction accuracy. Package contains: raw trial data, MIMO telemetry, configuration files (NSSD parameters, fuzzy rulesets, full architecture specification), analysis scripts, and dataset generators for reproducibility. Accompanies manuscripts submitted to IEEE Access and Robotics and Autonomous Systems (Elsevier).
Files
Steps to reproduce
**Steps to reproduce:** 1. **Extract archive** and install dependencies: `pip install -r analysis/requirements.txt` 2. **Regenerate datasets** (optional — data included): - `python generators/generate_automotive.py` → produces `data/automotive/trials_full.csv` - `python generators/generate_railway.py` → produces `data/railway/telemetry_full.csv` and `handover_events.csv` 3. **Run analysis**: `python analysis/apacc_analysis.py` → reproduces statistical comparisons and generates figures 4. **Run NSSD demonstration**: `python demo/nssd_demo.py` → demonstrates Non-Specificity Supervised Discretisation algorithm (coarse tuning layer) 5. **Configuration files** in `config/` contain all parameters from manuscripts: - `apacc_architecture.json` — full MIMO architecture specification - `automotive_nssd_config.json` — discretisation thresholds - `pedestrian_crossing_ruleset.json` — 24 Type-2 fuzzy inference rules Automotive trials generated using CARLA simulator physics with Honda Civic parameters. Railway telemetry derived from Project EDGE deployment logs (UK SBRI, Workington–Sellafield route). Random seeds fixed for reproducibility.
Institutions
- Cranfield University Centre for Autonomous and Cyber-Physical SystemsCentral Bedfordshire, Cranfield