Unlocking the potential of autonomous systems and operations through supervisory risk control

Autonomous systems and operations enable new and challenging operations, such as unmanned land and sea transport, monitoring and mapping of sea and land areas, and inspections of industrial systems, structures, and components that are difficult to access. Autonomy requires advanced control systems and applications for higher intelligence. This can lead to safer and more efficient systems and operations, but also to increased complexities and dependencies that are very challenging to identify, analyze and control in risk management. Manufacturers need to develop safe and reliable autonomous systems. Operators of the systems must be able to plan and carry out safe and robust operations, with an acceptable level of risk. The industry needs new standards, guidelines and tools as a basis for future business development within autonomy. The authorities must be proactive in relation to regulation, follow-up, development, and use of autonomous technology. Approval of intelligent systems with functionality that involves learning and operational optimization means that risk management and online risk control, testing and verification must be a very important element in design, operation and system validation. This research project develops new methods that make autonomous systems both safer and more intelligent. The methodology enables risk control to be incorporated as an autonomous functionality in the control systems. This means that control systems can make risk-based decisions, which enhances safety and increases functionality, thus leading to greater prevalence and use of autonomous systems.

-

This proposal addresses fundamental research challenges related to risk acceptance and supervisory risk control of autonomous systems and operations. The aim is to develop more powerful risk control solutions to achieve safe system performances and allow for widespread use of autonomous systems. The outcomes of this ambitious project will support and enhance the achievement of higher-level autonomy and intelligence in advanced control systems through the integration of online risk modelling, testing and verification of safe responses with model predictive control (MPC). Autonomous systems are emerging and essential for allowing new and challenging operations, such as mapping and monitoring of oceans and areas on land, inspections of structures difficult to access, and autonomous transportation, both land based and at sea. Autonomous functionality may be a step towards safer and more efficient operations, but software and advanced control systems also lead to complexity and interlocks that are extremely challenging to identify, assess, and control. Autonomous systems are used in very different operations with a range of hazards; from major hazards to occupational risk that may cause human fatalities and injuries, environmental and asset damage, and economic losses. The lack of knowledge, standards and limited operational experience, make traditional risk reducing measures, such as component redundancy and existing methods for verification and validation (V&V), ineffective. Acceptance of highly intelligent systems with built-in learning and optimization capabilities require supervisory risk control and online risk modelling, testing and verification to become a driver in design, operation and system validation. Thus, the interdisciplinary research approach in this project builds on the most powerful theories of risk modelling, V&V, control engineering and autonomous systems. The fundamental research results will be applicable to a wide range of autonomous systems.

Publications from Cristin

Online Probabilistic Risk Assessment of Complex Marine Systems. Principles, modelling and applications

Intention modeling and inference for autonomous collision avoidance at sea

Hybrid Tracking Controller for an ASV Providing Mission Support for an AUV

Risk-informed control systems for improved operational performance and decision-making

Towards Risk-Based Autonomous Decision-making with Accident Dynamic Simulation

Continuous and discrete abstractions for planning, applied to ship docking

Human-System Concurrent Task Analysis: An Application to Autonomous Remotely Operated Vehicle Operations

1234

No publications found

A system-theoretic approach to hazard identification of autonomous operation with multiple autonomous marine systems

Special Panel on Autonomous System Safety, Risk and Security

Advances in the safety and assurance of autonomous ships

Long-duration Autonomy for Open Ocean Exploration

Intelligent risk-based under-ice altitude control for autonomous underwater vehicles

Preliminary hazard analysis of a small harbor passenger ferry - Results, challenges and further work

Risk-Based Obstacle Avoidance in Unknown Environments using Scenario-Based Predictive Control for an Inspection Drone Equipped with Range Finding Sensors

12

Proceedings of the International Workshop on Autonomous Systems Safety 2021

1

No publications found