Geofence ‑ Driven Trailer Management
Revolutionizing trailer logistics with geofence-driven machine learing.
Gain real-time visibility, reduce idle time, and boost fleet efficiency.
ML Transformation: Geofence - Driven Trailer Management at Soulax
Customer Challenge
Soulax, a software services company, also manages its own internal trailer logistics for customer deliveries and yard movements. The company faced operational inefficiencies due to lack of real-time visibility into trailers once dropped at consignee yards and warehouse locations. Without automation, trailers frequently sat idle, cycle times were extended, and dispatchers relied heavily on manual status checks.
To address these challenges, Soulax deployed an ML-powered geofence and trailer state detection system on AWS for its internal logistics operations, aiming to reduce idle time, improve asset utilization, and eliminate manual tracking.
The ML Problem
The central machine learning challenge was classifying trailer states — staging, loading/unloading, or pickup-ready — using only GPS coordinates and IoT sensor data. Rule-based systems failed due to inconsistent yard layouts, irregular dwell times, and noisy sensor data.
ML Architecture
Pipeline Flow: GPS Pings + Sensor Streams → DBSCAN Clustering → Geofence DB → Real-Time State Detection (XGBoost) → Alerts + Dashboards
Solution Approach
Geofence Clustering with AWS Services
- Amazon SageMaker executed DBSCAN clustering on historical GPS trajectories to dynamically generate geofences
- Convex hull geofences were stored in Amazon RDS (PostgreSQL with PostGIS) for spatial queries
- Hyperparameters tuned with Silhouette Score and IoU via SageMaker Processing
- Metrics captured in Amazon CloudWatch for monitoring.
Operational State Detection
- An XGBoost classifier was trained in SageMaker to detect staging, loading/unloading, and pickup states.
- Features included dwell duration, GPS variance, door sensor signals, temperature, time of day, and weather.
- Models registered in SageMaker Model Registry with artifacts stored in Amazon S3.
Real-Time Inference Pipeline
- Amazon Kinesis Data Streams ingested 500K+ GPS/sensor events daily
- AWS Lambda functions triggered predictions and alerts.
- Notifications delivered via SMS, email, and webhooks with <5-second latency.
Visualization & Monitoring
- Grafana dashboards provided real-time views of yard dwell times, geofence overlays, and event history.
- Amazon CloudWatch monitored drift, data quality, and alert performance.
Results & Business Impact for Soulax
Metric | Impact |
Dwell Time Reduction | ↓ 20% |
Detection Accuracy | 96% Precision / 92% Recall |
Trailer Utilization | ↑ 15% more cycles/week |
Alert Latency | < 5 seconds |
AWS Value Proposition
AWS services were critical in enabling Soulax’s ML transformation:
- Scalable ML Infrastructure – SageMaker supported large-scale clustering and classification.
- Real-Time Processing – Kinesis + Lambda enabled sub-second event detection.
- Cost-Effective Storage – S3 and RDS ensured secure and affordable data storage.
- Monitoring & Observability – CloudWatch provided full visibility into operations
Conclusion
By leveraging AWS machine learning services, Soulax successfully transformed its own trailer logistics operations. The geofence-driven ML system reduced idle time, improved trailer utilization, and eliminated reliance on manual tracking. This case study highlights how Soulax, as both a software services provider and an AWS customer, applied AWS ML capabilities to solve a real-world logistics challenge at scale.
Ready to Revolutionize Your Access Control Systems?
Partner with us to implement cutting-edge, real-time facial recognition solutions tailored to your unique needs.
Let’s discuss how we can help enhance your security and operational efficiency.