About this role
Modern autonomous ground vehicles (AGVs/UGVs) in defence operations require sophisticated AI/ML-based control systems for perception, decision-making, and adaptive responses in complex, unstructured environments where terrain can change abruptly. However, formally certifying these opaque learning-based components demands impractical resources, presenting critical safety assurance challenges and delaying the adoption of novel technologies. Our prior independent and DSTL-funded research has established foundational safety assurance techniques for autonomous systems:
This PhD will develop an assured runtime safety controller designed to enable autonomous systems to operate safely in dynamic environments. Developed in collaboration with SC Group Ltd., with applications to defence autonomous systems, the approach combines multiple techniques such as onboard safety monitoring, operating environment adaptation and real-time robust learning of uncertainties and nonlinearities within the dynamical system.
In this PhD, the aim is to simultaneously learn control policies and safety certificates—mathematical proofs that control decisions are safe. Data from system operation provides evidence that both the control and the proofs are valid. The proposed controller will prevent unsafe actions during the deployment of AI/ML-enabled functional blocks in the closed-loop control of AGVs/UGVs.
Please apply via the ‘Apply’ button above.