SafeCLAI

Safety Concept for Low-speed AV Implementations

SafeCLAI

SafeCLAI is developing a re-usable safety concept for autonomous vehicles, enabling more efficient assessment procedures. The SafeCLAI safety layer, to be published in open source, serves as an example for autonomous vehicle OEMs to create safe solutions allowing functional updates over their lifetime.

Challenging

Safety Concept for Low-speed AV Implementations

Our country has a very dense and challenging transport and mobility system. National research agendas and roadmaps of various sectors, such as HTSM, Logistics and Agri&food, promote vehicle automation as a means to increase transport safety and efficiency.

Companies developing and applying vehicle automation must comply with application and/or sector-specific standards and legislation. An important aspect is the safety of the automated vehicle within its operational design domain, which must be demonstrated by manufacturers and assessed by authorities. The different standards and procedures have many similarities, but also lead to significant differences in application experience and available safety-related solutions. For example: Industrial AGVs (Automated Guided Vehicles) have been around for many years, while autonomous road vehicles can only be found in limited test environments and pilots. Companies face an increasing need to cover multiple application environments, such as restricted access areas and public roads, leading to complex technical choices and parallel certification and homologation procedures.

SafeCLAI addresses this challenge by developing a concept for a generic safety layer in the control of autonomous vehicles that can be reused in different applications and sectors. The reusable concept reduces development time and enables more efficient assessment procedures. SafeCLAI focuses primarily on low-speed applications because these are the most desirable and technically feasible. Nevertheless, higher-speed aspects will be considered to enable future expansion.

Re-usable safety control layer

The SafeCLAI safety concept

The SafeCLAI safety concept consists of the addition of a separate safety layer between the functional control layer and the lowest control level of the vehicle. At this lowest level, safe 'x-by-wire' control is assumed to execute commands for steering, acceleration and deceleration. During normal operation, the commands of the functional layer are executed. The SafeCLAI safety layer monitors the functional layer’s behaviour with respect to safety and intervenes if the vehicle exhibits unsafe behaviour. This intervention may consist of executing a 'minimal risk manoeuvre'. In its simplest form, this is an emergency stop.

Unique feature

Open source

In principle, developing such a concept is nothing new. What makes SafeCLAI unique is the fact that a number of parties with relevant knowledge and experience are collaborating on a use case that does not contain company secrets, which means that the process followed (in accordance with standards such as ISO26262 and ISO21448) including the associated documents will be published. The prototype security concept developed in SafeCLAI will first be virtually verified and validated through a scenario-based simulation environment. Then, the prototype will be practically integrated and demonstrated in a scenario with a pedestrian crossing. The documentation and the source code resulting from the project has been published here on Gitlab:

Code on Github

RAAK

Funding

SafeCLAI project is a RAAK-mkb project, funded by SIA (www.regieorgaan-sia.nl).

Project Partners