Menu
Close
Sign up for NPL updates
Close
Sign up for NPL updates

For people, place, prosperity and planet, we deliver impact with measurement science

Security and resilience

Assured autonomy

Releasing the potential of autonomous systems for the benefit of the UK

Ensuring confidence in autonomous systems

The deployment of machine learning (ML) and artificial intelligence (AI) is accelerating. This presents an opportunity to perform complex tasks better and faster than currently possible, and deliver significant potential benefits. Ensuring public confidence in these emerging autonomous systems will be key to the speed of their exploitation and the delivery of benefits.  

NPL is helping to address the challenge of how to assure the reliability and safety of autonomous systems. We are establishing a national measurement infrastructure to support reliable development, testing, validation and certification. This is part of a highly collaborative programme of activities involving government, academic, regulatory and commercial stakeholders and is focused on the following themes: 

  • development of relevant and usable standards and infrastructure for data quality and provenance that will ensure data used by, and to train, AI systems is fit for purpose  
  • build the measurement infrastructure including the suite of standardised and traceable metrics required for reliable and repeatable simulation testing and sensor performance characterisation 
  • link uncertainty, traceability and risk in order to validate testing, and measure the propagation of uncertainty through ML/AI

Solving the measurement challenge

Traditional approaches used to certify vehicles/vessels/aircraft as safe are not sufficient to assure the technology which is supporting or replacing the human driver/navigator/pilot - the autonomous system.  We need new technologies and approaches to be developed, so NPL provides scientific and measurement capabilities to address the following challenges:

  • new measurement and testing approaches are required, since functional testing alone is not sufficient to assure autonomous systems 
  • current simulation environments cannot provide a reliable link to real world testing because they have insufficiently defined environmental and sensor models
  • how to examine / validate ML/AI systems when they are ‘hidden’ within black boxes
  • data quality standards are needed to enable operational assurance of autonomous systems
Future aviation

Safe, automated operation of future airspace

Air traffic management systems of the future will need to seamlessly manage the increase in traffic volume, density, diversity of airspace users and the demand for fully integrated, non-segregated airspace. These more complex air traffic conditions require solutions, supported by an appropriate regulatory framework, that will maintain a safe separation between aircrafts, identify and resolve potential conflicts dynamically and optimise routes. This mission is safety critical and may eventually move beyond the reach of human operators. Scalable air traffic management (manned or unmanned) ultimately needs a high-level of autonomy to both run the operation and also infer and deduce the rules that aircraft should adhere to.

Requirement: to provide confidence in the management of a more integrated airspace including both manned and unmanned vehicles, including commercial airliners, drones and eVTOL aircrafts and autonomous aircrafts of various sizes.

Our approach: deploy our measurement expertise to help UK companies, authorities and regulators to develop enabling systems that could be used nationally; accelerating commercialisation and supporting the UK in international regulatory developments.

Areas of focus

  • resilient/reliable communications and sensors
  • system of systems assurance
  • technical frameworks to support integration of tests/testing capabilities and system characterisation
  • data quality frameworks and standards
  • supporting decarbonisation
Self-driving vehicles

With the advent of ‘automated’ or ‘self-driving’ vehicles, we require entirely new types of tests for the systems which are replacing the human driver. Once critical challenge is our reliance on a range of sensors for safety critical applications in connected and autonomous vehicles (CAV). How these sensors perform and where they might fail must be clearly understood; failure to do so may lead to serious safety issues.

The weather is a dominant and particularly complex aspect of the changing conditions that may affect sensors. There is an opportunity for the UK to build and demonstrate leadership in this area to regulators and CAV developers around the world.  

Testing in virtual simulation environments will be a vital component - and the only practical approach - for testing the range of complex scenarios that inform the safety of an AV (autonomous vehicles). It is important that virtual test environments should reliably and accurately emulate both the different physical environments the AV will encounter and also how its sensors perform in these different environments. This requires standardised metrics for measuring sensor performance and for defining environmental conditions or 3rd party objects/agents which would be encountered in the virtual simulation. 

NPL has been collaborating with the Met Office on a research project on behalf of the Centre for Connected and Autonomous Vehicles (CCAV) to specify a usable and reliable framework for understanding how well sensors perform in different weather-related conditions. When fully developed, this Framework will support validation, safety assurance and simulation testing of CAV, across the UK.

Read the full Proof of concept report

Read the Sensor performance study produced in partnership with Connected Places 

Marine autonomous systems

Helping the UK lead on standards for marine autonomy

Autonomous maritime platforms and vessels will require new approaches to testing and measuring performance, beyond what is currently possible. These will allow the appropriate evidence for Safety of (Autonomous) Navigation to be generated and enable the required classification of these vessels and the technology to be commercially exploited.

NPL has teamed up with Lloyd's Register (LR) on marine autonomy projects to ensure that appropriate levels of service and competence can be achieved within the maritime industry. As part of the framework agreement, the two organisations are establishing and enhancing the current body of knowledge for marine autonomy. This combination of skills, expertise and experience is bringing clarity to the requirements for the assurance of autonomy and assisting allied stakeholders in realising the potential of these systems in the market. It will allow standards to be set and consistently applied, and will therefore bring surety to risk management and certification for autonomous and unmanned systems and vessels.

Maritime Autonomy Assurance Testbed - NPL

Case study

Improving safety in automated vehicles

Article

Working with NPL

Our researchers are at the forefront of new innovations that promise to deliver new solutions to security and resilience challenges. We play a vital role in providing the standards and assurance to provide confidence in new products and providing reliable data. Learn more about our research areas and working with our experts. 

If you have a research collaboration where our expertise can help, please can contact the team and discuss your challenge.

Recent reports addressing key challenges

Sensor assurance framework project

The safe operation of connected and autonomous vehicles (CAVs) requires that the impact of the weather on their perception sensors is quantitatively understood and represented in the Operational Design Domain (ODD). Since 2020, the Sensor Assurance Framework1 (SAF) project has been developing methods for capturing this weather sensitivity, including an extensive set of measurements of sensor performance in a wide range of comprehensively measured weather conditions.

This discussion paper, produced jointly by the Met Office and NPL, explores some aspects of this problem in order to stimulate discussion and mutual understanding.

Download report

 

A sensor modelling framework for autonomous systems 

Sensor modelling is an important aspect of developing and testing automated vehicles. This document describes a standardised framework for the practice of developing such models. We describe a general approach and illustrate it with examples for different sensors and sensor models.

Download report

Camera and Lidar sensor models for autonomous vehicles

Camera and Lidar are important sensors in the automated vehicle sensing stack. In order to ensure safety of automated vehicles we need to understand the working of these sensors and verify their functioning in different edge cases so we can understand their characteristics in the operational design domain (ODD). In this study we will demonstrate how to apply the sensor modelling framework to model the functioning of these sensors.

Download report

Contact us

Our research and measurement solutions support innovation and product development. We work with companies to deliver business advantage and commercial success.
Contact our Customer Services team on +44 20 8943 7070