Mathematical Models TRIZ Cases
- Roadside Parking Utilization Mapping from Fleet Parking Positions
Fleet parking positions are classified by road distance to estimate roadside space occupancy, helping drivers cut parking search time and distance.
- Charging Unit Guidance Using User-State Correlation
User walking and at-charger states are correlated with position data to locate working charging units and guide other autonomous vehicles.
- Autonomous Driving Path Prediction in Adversity Conditions
Machine learning predicts safe vehicle trajectories from sensor and image data when road markings, work zones, or obstacles disrupt normal driving.
- Safe Driving Path Learning for Autonomous Adversity Conditions
Uses image, inertial, and trajectory data to train models that detect adversity conditions and guide autonomous vehicles along safer paths.
- Vehicle Feature Map Aggregation With Pose Error Correction
Shared vehicle feature maps are aligned using relative and absolute pose correction to improve perception, prediction, and motion planning.
- Autonomous Vehicle Motion Planner Hyperparameter Tuning for Generalization
Bayesian and evolutionary tuning refines motion planner hyperparameters against truth data to improve realistic driving decisions in new scenarios.
- Vehicle Service Dispatch Using Dynamic Queues and Service Complexity
Dynamic queueing based on service complexity dispatches vehicles at the right time to cut wait times and keep service stations flowing.
- Modular VRU Behavior Prediction With Explainable Uncertainty
Probabilistic graphical models combine with ML to predict pedestrian behavior with uncertainty estimates, reducing black-box risk and retraining effort.
- Aircraft Control Classifier for Unforeseen Scenario Autonomy
A trained classifier selects operating procedures from pathway data so aircraft control can handle unforeseen scenarios with less rule coding.
- Accelerator Slot Location Mapping for Workload Allocation
Pre-characterized PCIe/CXL slot locations guide workload placement to cut latency, improve accelerator use, and manage power.
- End-to-End Driving Neural Network Training for Edge-Case Validation
Curated difficult driving tasks, synthetic data, and memory-augmented imitation learning improve autonomous driving validation at scale.
- Vehicle Control Envelope for Uncertainty-Aware Motion Limiting
Future-state uncertainty is checked against a vehicle control envelope to curb heavy-duty motion before skid, jack-knifing, or trailer swing.
- Neural Network Output Monitoring With Fixed Boundary Detection
A second neural network checks controller outputs against fixed boundary conditions, blocking unsafe actions and triggering targeted retraining.
- Vehicle Safety Event Detection With ML Feedback Tuning
Local and backend model outputs are compared to tune vehicle safety event detection in real time, improving alert accuracy without retraining.
- Trajectory Prediction With Evolving Intent and Adaptive Sampling
A deep neural PHA model tracks changing maneuver intent and prunes trajectory samples to improve prediction accuracy and coverage.
- Computational Wafer Inspection for Hot Spot Defect Prediction
Hot spot analysis predicts lithography defects from process parameter ranges, enabling targeted wafer inspection and fewer random checks.