Estimating Safety Risk in Autonomous Trucking using Scenario-Based Probabilistic Risk Assessment
Date and Time: Wednesday, July 31, 2024: 1:30 PM - 5:00 PM
Location: Indigo D
Kelly Smith
Lead Systems Engineer for Autonomy, Kodiak Robotics
Presentation Description
As an element of the Kodiak Driverless Safety Case, Kodiak assesses the expected risk of collision across multiple severity classes using Probabilistic Risk Assessment (PRA), a quantitative risk estimation technique heavily relied upon by the human spaceflight and nuclear power industries. Smith will outline the overall PRA technique, approach used by Kodiak, how it enables management of uncertainty of safety claims for ML-based autonomy systems, and how it addresses the key elements of UL4600 & ISO/PAS 21448 (SOTIF). The PRA tool and process is actively used to inform and guide engineering priorities to accelerate responsible deployment towards a safe L4 autonomous system.
Speaker Biography
Kelly Smith is the Lead Systems Engineer for Autonomy Software at Kodiak. Kelly leads the Kodiak efforts to estimate residual safety risk by applying probabilistic risk analysis techniques to driving scenarios that are relevant to the ODD. Prior to joining Kodiak, Kelly spent 3.5 years in the AV industry at Embark Trucks and developed the Embark Driverless Safety Case. Prior to Embark Trucks, Kelly spent 13 years at NASA's Johnson Space Center as an aerospace engineer developing and certifying safety-critical GN&C systems for the Orion spacecraft, serving as a Trajectory Officer in Mission Control for the EFT-1 Flight Test, and led the Orion Backup Flight Software team for the Artemis 2 mission from conception through V&V. Kelly holds a B.S. in Aerospace Engineering and an M.S. in Computer Science.
Presentation File
IP-Panel Discussant
Category
Invited Presenter