RESULTS AND TECHNOLOGIES

Learn more about the SAFEXPLAIN-developed technologies working to enable the use of AI in any type of safety-critical system.

DEMOS

See the CoreDemo walkthrough showing SAFEXPLAIN capabilities in action. Useful for early users in all the transportation and mobility value chains.

CORE DEMO

The Core Demo is an open, complete SW platform that is able to demonstrate critical functionalities such as Temporal Consistency, Supervision to Decision, Diagnostic & Monitoring, Safety Control, and more on selected “toy” examples from space, rail and automotive cases.

For critical autonomous AI-based systems (CAIS) developers and decision-makers in transportation and mobility sectors

CAR DEMO

Provides a flexible and high-fidelity environment with ODD-compliant safety scenarios aligned with functional safety standards, allowing rigorous testing of the agent’s capabilities.

RAIL DEMO

Based on ROS2 architecture and consisting of four main nodes: the video player (or the camera output simulator), the object and track detection node, the stereo depth estimation node and finally the safety function node.

SPACE DEMO (proprietary)

 A critical Guidance, Navigation, and Control (GNC) Space demo able to detect the target and provide accurate information on its position and distance from the agent, operates at a safe distance to avoid crashing or damaging  assets.

Open Source SW

PWCET-AI

pWCET-AI is a novel probabilistic timing analysis tool that allows for the characterization of the timing behaviour and for probabilistic Worst-Case Execution Time (pWCET) estimates of AI-based solutions. The library allows for trustworthy and tight execution time bounds capturing the specific non-deterministic traits of ML and DL software solutions running on complex SoCs such as, for example, the NVIDIA® Jetson Orin™.

For embedded software developers and validation and verification engineers (V&V) in automotive, railway and aerospace sectors

SAFE YOLO

The Safe YOLO library is a basic software re-design/implementation of a subset of YOLO functions in compliance with functional safety standards requirements against SW systematic errors. It provides a safety software design and implementation, that integrates a structured and layered software architecture, for the deployment of DL-components.

For dependable and CAIS developers in the automotive, railway, industrial and aerospace sectors

SEMDRLIB

SemDRlib is a dedicated DL library that generates diverse redundant versions of image-based DL models and applies a number of user-selected transformations in input images to perform multiple diverse inferences intended to provide semantically-identical, yet not bit-identical, results.

For embedded software developers and verification engineers in automotive, railway and aerospace sectors

ROSGUARD

ROSGUARD is a run-time bandwidth monitoring mechanism that controls bandwidth memory usage of non-critical applications and temporarily stops them whenever they exceed the allocated quota. It builds on ROS2 modular architecture and is easily portable across platforms and setups.

For embedded software developers in automotive, railway and aerospace sectors

DLETLIB

DLETLib is a dedicated DL Explainable and Traceable library, incorporating a strongly structured and layered software architectural design that allows for the development of DL components following the requiremetns from functional safety standards like ISO 26262, ISO 21448 (SOTIF), IEC 61508 and others. 

For embedded  software developers and V&V engineers in automotive, railway and aerospace sectors

XAI FOR SAFETY/ EXPLib

EXPLib is a research platform in the form of an opensource repository (GPL3 license or equivalent) containing methodology and tools enabling the use of XAI techniques. It provides a knowledge base and approach for using XAI to support the application of AI-based components in safety critical systems, including a Satellite docking Toy Model.

For industry and research communities interested in safe and explainable AI

Additional Open Content

AI-FUNCTIONAL SAFETY MANAGEMENT (AI-FSM 2.0)

AI-FSM 2.0 development is a ‘safety lifecycle’ defined in compliance with existing AI-safety standards (e.g., ISO 5469) that provides the required basic procedures, guidelines and templates to support the development of DL-components for CAIS systems, with a technical focus on safety and XAI.

For dependable- and CAIS developers in the automotive, railway, industrial and aerospace sectors

AI-V&V SCENARIOS

AI V&V-SCENARIOS is a catalogue of ODD-described scenarios that extends the traditional Functional Safety (FuSa) V&V model to AI techniques and methods. It is is ready to be used, based on public data, available as open source content, integrated with test cases specification.

For embedded software developers in automotive, railway and aerospace sectors

Proprietary SW

ORIN-PMULIB

Orin-PMULib is a dedicated performance monitoring unit libarary that allows for configuration on target performance monitoring counters and debug devices. It is specifically adapted to the platform and providers a lightweight but accurate way to configure and retrieve precise information on traceable hardware events.

For embedded software developers and V&V engineers in automotive, railway and aerospace sectors, who need to use, configure and collect information on hardware events on the NVIDIA® Jetson Orin™

JANE

The JANE autonomous navigation application is an AI software implementing algorithms for navigation that makes space assets more autonomous and reactive and reduces the effort of ground staff. It enables space critical systems with safe and explainable AI for their navigation operations, compliant with ECSS standards for space software verification and validation (ECSS-E-ST-10-02C), dependability and safety (ECSS-Q-HB-80-03A).

For space industry companies employing assets that require navigation and control (Earth Observation, Space Debris Collection and Removal, Telecommunications, In-Orbit Servicing, etc.)

Proprietary Content

AI-V&V-TESTSPECS

The V&V-TESTSPECS is a full set of tests specification that extends the traditional FuSa V&V model to AI techniques and methods.

For embedded software developers in automotive, railway and aerospace sectors

SAFETY PATTERN LIBRARY (SPL)

Safety Pattern Library (SPL) is a basic technical reference foundation that provides a set of documented exemplary safety-case(s) and exemplary safety-concept(s), with a technical focus on safety and XAI that describe common safety design approaches (solutions) to common design requirements (recurrent problems).

For dependable and CAIS developers in the automotive, railway, industrial and aerospace sectors

More project results