How often now do we read headlines, such as: “programme X delayed by software problems…”? AOS has signed a contract with BAE Systems’ Military Air Systems, on behalf of the Ministry of Defence (MoD), to research how to achieve regulatory approval for autonomous systems software. This is supported by the MoD’s “Software Systems Engineering Initiative”, or SSEI (http://ssei.org.uk/), an innovative MoD-funded strategic initiative intended to enhance "through life" capability management for software intensive defence systems, and thus reduce risks, delays and cost overruns.
The most recent generation of aircraft and engines rely on software for their primary air vehicle control, with no manual reversion. Although these complex systems provide substantial performance and safety benefits, one of the primary limitations of such embedded, safety-critical software-based systems is the cost and timescale required to make changes to the software, once it has been safety assured. Consequently, changes are accumulated and incorporated in infrequent “block” changes, associated with major airframe or weapons systems upgrades. The complexity of the software means that these modifications incur re-certification costs that approach the cost of the original clearance. It would be desirable to have a software certification process that relates the re-certification effort to the scale of the changes introduced.
Concurrent with the move to increasingly complex manned aircraft has been the introduction of remotely controlled Unmanned Air Systems (UAS) and the imminent introduction of autonomous UAS, where there is no, or little, human intervention.
This proposal concerns UAS intended to operate either fully or partly autonomously. This capability is provided by decision-making software, which replaces all (or part of) the functions performed by the human crew of a manned or remotely piloted vehicle. These extend beyond simply flying the vehicle to a wide range of other critical roles, including:
- mission management and re-routing;
- power management;
- failure detection, analysis and resolution;
- Prognostic Health Management;
- centre of gravity and fuel management;
- ATC and communications; and
- ground handling.
Consequently the scope of such software and systems is much greater than on a manned aircraft.
At the same time one of the key benefits of UAS is their ability to perform a range of “dull, dirty and dangerous” missions, and do this cost effectively. A manned aircraft is designed for a certain role, e.g., strategic or tactical transport, ground attack, or surveillance. Within these roles the variety of missions, e.g. for a tactical transport aircraft, can be very wide and many of the most successful designs are recognised for their versatility. Much of this due to the crew, who are able to be trained and re-trained, and can thus take on new missions when correctly briefed and practised for the new mission.
In contrast, at the time of their design, a UAS must be endowed with a range of behaviours to allow it to provide at least some degree of versatility, and the scope of its autonomous behaviours will reflect the envisaged range of missions. If subsequently its airframe and capabilities can be adapted to new or different missions then the on-board Autonomous Mission Management system has to be upgraded with all of the necessary additional or revised behaviours. The challenge is – how can these additional behaviours be safely introduced without requiring complete re-certification, as is the case with current manned aircraft systems?
The research programme will make use of AOS’s C++-based decision-making system as the experimental platform. This is C-BDI™, which is under development to safety-critical standards as defined by the CAA under the new UAS regulation UAS.1309, an extension to the manned aircraft regulations CS23.1309 and CS25.1309. This project will also be complementary to the national ASTRAEA programme, which aims to clear UK airspace to autonomous UAS without restriction (http://www.projectastraea.co.uk/?OBH=354).
AOS’s partners in the programme include the University of York (http://www.cs.york.ac.uk/research/), high assurance software specialists Kestrel Technology of the USA (United Space Alliance) (http://www.kestreltechnology.com/) and the UK Civil Aviation Authority’s Safety Regulation Group, which is contributing advice on how a regulatory authority would regard the proposed approaches.
The project will address the military requirement to improve software systems’ cost-effectiveness, by reducing the cost and timescales involved in certifying modifications and extensions to safety-critical air, land and maritime manned systems.