Autonomous Navigation

Toshiba and MIRISE use embedded optimization and multi-object tracking to keep a mobile robot navigating through moving obstacles in real time.

Business impact

• Processing rate — The system processes 23 frames per second.
• Higher Order Tracking Accuracy (HOTA) — The deployment improved HOTA by 4% over standard benchmarks and by 23% on benchmarks designed for object obscuration.

Data requirements

• Object detection and tracking data (tabular) — The system uses positional confidence, movement direction, and tracked object associations.
• Robot navigation state (tabular) — Path planning uses the tracking output to update moving-obstacle occupancy and future positions.

AI methods and techniques

• Multi-object tracking — Tracks multiple detected objects and re-identifies them after occlusion or crossing paths.
• Combinatorial optimization — Formulates object association as a matching problem to select track assignments.
• Path planning — Uses tracking data to adjust occupancy areas and predicted positions for navigation.

AI models and model families

Simulated Bifurcation Machine

Industries

Real-world evidence

2 documented case studies on record.

Companies using this: Toshiba Corporation and MIRISE Technologies Corporation, Shield AI.