Autonomous Systems & Formal Verification Lab (ASFV Lab)

Key Research Questions & Hypotheses

1. How can we compute tight mathematical bounds on the reachable sets of nonlinear dynamical systems controlled by deep neural networks?
2. Can automated theorem provers synthesize real-time safety shields that intervene with zero latency when invariants are violated?
3. What abstractions bridge the gap between continuous physical dynamics and discrete logic verifiers?

Methodologies & Technical Approaches

1. Combining symbolic interval analysis, zonotopes, and Taylor models for high-dimensional neural reachability.
2. Developing differentiable SMT solvers for automated barrier certificate generation.
3. Validating certified controllers on custom physical robotic rovers in our motion-capture testing facility.

Facilities, Compute & Equipment

• 150 m² experimental robotics arena with 24-camera Vicon optical motion capture system.
• Automated theorem proving test suites in Coq, Lean, and Z3 with reproducible verification artifacts.

Key Scientific Publications & Foundation Works

Selected foundation literature and methodology references utilized by the laboratory team.