RESEARCH
Research for reliable autonomy in the physical world
STATE16 studies how estimation, planning, simulation and evidence can work together when autonomous systems must act under uncertainty.
Benchmarks
Glossary: Physical AI
State estimation with context
AIME experiments with a 15-state navigation model, uncertainty propagation and indoor positioning assumptions that do not rely on GPS inside structures.
Long-horizon prediction
Future work explores forward behavior windows, trajectory divergence, trust decay and physical admissibility across short and long horizons.
Evaluation that connects to action
The research direction is not a score alone: it is a reproducible path from a detected failure condition to a tested guardrail.