In heterogeneous multi-core systems, efficiently allocating tasks to cores is a key challenge due to the complexity of hardware architectures and the dynamic nature of workloads. System states and task behaviors are inherently partially observable due to limited accessible information about interactions among tasks and shared resources, which complicates the prediction and management of execution times and resource contention. Moreover, uncertainties such as unpredictable execution times and variable workloads, combined with the challenge of prioritising high-criticality tasks while maintaining overall system performance and ensuring fairness in the execution of lowcriticality tasks, necessitate the use of sophisticated allocation strategies. We propose a novel approach which embraces the uncertainty of modern systems. Partially Observable Markov Decision Processes (POMDP) and Active Inference are used to minimise uncertainty and optimise allocation decisions. By representing uncertainties within the POMDP framework, our method enables probabilistic predictions of task behaviours and system states. Active Inference refines these predictions and adapts decisions dynamically, handling workload variability and system changes such as adding new tasks. Our methodology emphasises application-layer scheduling as opposed to kernellevel modifications, thereby giving our approach both platform awareness and deployment flexibility. Moreover, the capacity for offline training mitigates the adverse effects of online updates on the accuracy of target platform comprehension. Experimental results demonstrate superior performance over baseline approaches, enhancing system efficiency and resource utilisation by incorporating task criticality awareness, even in the presence of uncertainties and partial observability.

Existing classification frameworks for AI and autonomous systems are being outpaced by recent advancements in AI technologies. This limits their applicability to modern intelligent systems, particularly agentic AI systems (autonomous systems that leverage foundation models to achieve wide-ranging, multi-layered goals). To address this deficiency, we introduce INSYTE, a multi-faceted framework that supports the classification of AI systems ranging from traditional rule-based systems to cutting-edge embodied AI and agentic systems. To that end, INSYTE considers the essential characteristics of an AI system across eight key dimensions grouped into four categories: system design ( underspecification and adaptiveness ); functionality ( breadth and depth ); operating environment ( diversity and dynamism ); and independence from human operational control ( intervention and oversight ). Different AI systems (or versions of systems) yield different ‘patterns’ on an eight-axis radar chart that INSYTE uses to provide an immediate visual summary of an AI system’s overall capability and a detailed representation of its individual characteristics. The INSYTE framework aligns with OECD’s definition of deployed AI systems, which is becoming the standard definition used by legislators and developers worldwide.

Autonomous systems are of high complexity and often regarded as mixed-criticality systems (MCSs) in which functions are allocated criticality levels according to risk assessment based on safety standards. Typically, tasks have different real-time requirements across criticality levels, and the estimated worst-case execution times (WCETs) are distinct. Further, limitations in computational resources increase the difficulty of integrating tasks onto one shared hardware platform. Conventionally, all nonsafety critical tasks must be discarded or suspended to guarantee the execution of safety-critical tasks when facing a timing fault. This typically leads to a considerable decrease in the system’s Quality-of-Service (QoS). Achieving more graceful degradation is critical to minimizing QoS reduction. This work focuses on tackling timing faults and proposes a novel graceful degradation strategy for use in a mixed-criticality context. Thus, when a system has multiple operational modes depending on the environment or an operational task, our approach can give an effective way of managing degradation to maximize QoS, which is currently not sufficiently recognized in MCS. Furthermore, the proposed causality analysis-based degradation process “bridges the gap” so functional dependencies are considered in scheduling design and thus leads to a graceful degradation that is both feasible and reasonable in functional and nonfunctional terms. The evaluations show that QoS can be better preserved using the proposed context-aware degradation process when compared with more conventional MCS scheduling approaches.

Time-sensitive networking (TSN) is being widely investigated to provide Ethernet capabilities for in-vehicle backbone communication. However, the gate control list (GCL), as a simple mechanism for achieving timing determinism for safety-critical traffic (ST) frames with hard deadlines, is too rigid to handle the intrinsic timing uncertainty of automated driving systems (ADSs). Due to the complexity and the unpredictable operating environment, there can be delayed ST frames that can disrupt the solidly fixed timing behavior. In this work, we present one novel approach to effectively use bandwidth resources to deal with delayed ST frames, which cannot be handled by a traditional fixed GCL, and the discarding of them should be reduced to improve the system’s integrity. An acceptance test is implemented to report to the application layer when an ST frame will miss its deadline and hence be rejected, i.e., prevented from entering the switch. To further improve the efficiency of bandwidth usage, we investigate how to improve the performance of the more important Class A frames in an audio-video-bridging (AVB) switch, which adopts a credit-based shaper mechanism, and we propose a constant bandwidth server to replace the credit-based shaper while taking fairness into consideration. Evaluation with extensive experiments shows that both resilience and efficiency of the TSN are significantly enhanced compared with a credit-based shaper, especially when the traffic load is relatively high. For the delayed ST frames and event-triggered traffic, our approach is able to schedule more than the solution using the AVB switch even with a high network load.