Scalable Automatic Reasoning in Model-Based Development

Guest Speaker:
Nikos Arechiga — Toyota Research Institute

Wednesday, September 13, 2017
EEB 248

ABSTRACT: Recent progress in vehicle autonomy and robotics has increased the importance of system assurance, ranging from safety to security concerns. These assurances require systems that are able to reason about large and complex system designs, often containing large lookup tables as well as AI components.

This talk presents a general-purpose technique that leverages machine learning to automatically learn logical antecedents and consequents to simplify a complex formal verification task.

We also describe a specialization of this technique that has been used within Toyota to reason about software with large lookup tables, including a public benchmark.

Finally, we look to the future and describe emerging research directions in automatic reasoning.

BIO: Dr. Nikos Arechiga graduated with a Ph. D. in Electrical and Computer Engineering at Carnegie Mellon working with Professor Bruce Krogh. His graduate work touched on automatic inference of barrier certificates to simplify proofs of safety as well as techniques for provably-correct controller synthesis.

He has been working at Toyota for two years, and has been involved with developing scalable reasoning techniques to address complex models with lookup tables, and is recently considering the problem of reasoning about AI components.

Hosted by Paul Bogdan