Canada Spends Logo

Explore our Budget 2026 analysis

Grant: Discovery Grants Program - Individual

McMaster University

$36,000.00

Rigorous Model Based Assurance of Object Detection Systems for Safety of Autonomous Vehicles

FY 2023-24

Summary
Autonomous vehicles typically make use of a variety of sensors (cameras, radar, LiDAR) to sense the environment. Object detection is typically done using Deep Neural Networks (DNN) but these systems are vulnerable to false negatives (failures to detect objects) resulting in documented accidents causing injury and death. While companies such as Tesla claim the use of these systems is already safer than a conventional driver (a claim that is debatable), it does not replace the need for rigorous engineering of these systems. In earlier work on system safety and certification of software intensive systems, the applicant and his colleagues have advocated for separation of safety and control, i.e., the use of system architectures that separate complex control software from a simpler monitoring system that is used to guarantee safety properties. Formal verification techniques can be applied to the smaller, simpler safety system to provide the additional assurance required for a safety critical application while innovative controller designs can still be employed because of the safety guarantees provided by the safety system. A good example of this in the automotive domain is the E-Gas 3 Level Safety Concept for throttle-by-wire where a more complex "Intended control function" provides the main performance functionality in Level 1, a simpler safety critical Function Monitoring is provided by Level 2 that checks Level 1 for errors and then Controller monitoring is provided by Level 3 that tests reliability of the Level 1 and 2 systems. The goal of the proposed research project will be to develop a similar safety architecture for object detection sensor systems that is formally verified and backed by a rigorous assurance argument. The architecture will incorporate DNN based components as the main "Level 1 - Intended control function". It will be checked by a deterministic, formally verified, object detection algorithm ("Level 2 - Function monitor"). The "Level 3 - controller monitor" will check the correct functioning of the Level 1 and 2 subsystems. This proposed architecture will allow the development of a standardized rigorous argument about the safety and reliability of the object detection system in the form of an Assurance Case Template that precisely documents the assumptions and limitation of the system.
Program
Discovery Grants Program - Individual
Location
Ontario, Canada
Recipient Type
Academia
Type
Grant
Source
NSERC's Awards Data

As published

Application id
RGPIN-2023-05694
Name nom
Lawford, Mark
Department département
Computing and Software
Organization id
27
Institution établissement
McMaster University
Province en
Ontario
Province fr
Ontario
Country en
CANADA
Country fr
CANADA
Fiscal year exercice financier
2023
Competition year année de concours
2023
Award amount
36000
Program id
RGPIN
Program name en
Discovery Grants Program - Individual
Program name fr
Programme de subventions à la découverte - individuelles
Group en
DISCOVERY RESEARCH
Group fr
RECHERCHE AXÉE SUR LA DÉCOUVERTE
Committee code
1507
Committee name en
Computer Science
Committee name fr
Sciences informatiques
Area of application code
700
Area of application group en
Transportation systems and services
Area of application group fr
Services et systèmes de transport
Area of application en
Transportation systems and services
Area of application fr
Services et systèmes de transport
Research subject code
2705
Research subject group en
Information technology
Research subject group fr
Software and development
Research subject en
Technologie de l'information
Research subject fr
Logiciels et développement
Application title
Rigorous Model Based Assurance of Object Detection Systems for Safety of Autonomous Vehicles
Keyword
Autonomous Systems; Cyber Physical Systems; Formal Verification; Model Based Design; Object Detection; Safety Critical Systems
Application summary
Autonomous vehicles typically make use of a variety of sensors (cameras, radar, LiDAR) to sense the environment. Object detection is typically done using Deep Neural Networks (DNN) but these systems are vulnerable to false negatives (failures to detect objects) resulting in documented accidents causing injury and death. While companies such as Tesla claim the use of these systems is already safer than a conventional driver (a claim that is debatable), it does not replace the need for rigorous engineering of these systems. In earlier work on system safety and certification of software intensive systems, the applicant and his colleagues have advocated for separation of safety and control, i.e., the use of system architectures that separate complex control software from a simpler monitoring system that is used to guarantee safety properties. Formal verification techniques can be applied to the smaller, simpler safety system to provide the additional assurance required for a safety critical application while innovative controller designs can still be employed because of the safety guarantees provided by the safety system. A good example of this in the automotive domain is the E-Gas 3 Level Safety Concept for throttle-by-wire where a more complex "Intended control function" provides the main performance functionality in Level 1, a simpler safety critical Function Monitoring is provided by Level 2 that checks Level 1 for errors and then Controller monitoring is provided by Level 3 that tests reliability of the Level 1 and 2 systems. The goal of the proposed research project will be to develop a similar safety architecture for object detection sensor systems that is formally verified and backed by a rigorous assurance argument. The architecture will incorporate DNN based components as the main "Level 1 - Intended control function". It will be checked by a deterministic, formally verified, object detection algorithm ("Level 2 - Function monitor"). The "Level 3 - controller monitor" will check the correct functioning of the Level 1 and 2 subsystems. This proposed architecture will allow the development of a standardized rigorous argument about the safety and reliability of the object detection system in the form of an Assurance Case Template that precisely documents the assumptions and limitation of the system.