Integrated Quantum Systems Combine Sensing and Computation with Indefinite Order

Researchers are increasingly exploring methods to transcend conventional limitations in information processing, and a new study details a significant advance in this field. Ivana Nikoloska from Eindhoven University of Technology, along with colleagues, demonstrate a novel scheme for integrated sensing and computation utilising operations with indefinite causal order (ICO). This work investigates whether both sensing and computation can be performed simultaneously within an ICO framework, where the order of these tasks becomes indefinite, a departure from traditional paradigms where sensing always precedes computation. By representing a state as an agent experiencing a superposition of orderings, the team achieved promising results in a magnetic navigation task, showcasing the potential for ICO to enhance information processing capabilities and paving the way for more flexible and powerful computational systems.

Their work introduces quantum operations with indefinite causal order (ICO), a framework where the sequence of sensing and computation can be genuinely uncertain at a fundamental level. This research demonstrates that an agent can, in effect, process information before fully acquiring it, diverging from established paradigms in both classical and quantum information theory. The innovative framework leverages a quantum SWITCH to create a superposition of operational orders, enabling the agent to experience both possibilities concurrently. Unlike traditional machine learning where information acquisition always precedes analysis, this ICO-based system explores a more fluid relationship between these stages. The agent’s ability to learn from a perturbed quantum state, resulting from the initial observation, is central to the process. Optimising the parameters of a parametric model then minimises the discrepancy between predicted and actual targets, demonstrating a functional learning capability within the ICO framework. Extensive experimental validation using a representative task in magnetic navigation confirms the viability of the approach. Training and testing losses achieved 0.032 and 0.035 respectively, indicating successful performance and a strong correlation between predicted outputs and actual target values. The research showcases minimal training and testing losses, suggesting potential for practical applications in areas demanding efficient information processing. The core of this work lies in representing the quantum state as an agent capable of both state observation and learning a predictive function via a parametric model. Under ICO, this agent doesn’t simply choose an order; it exists in a superposition, effectively trying both simultaneously. This causal non-separability, a unique quantum resource, offers advantages over systems with definite ordering, potentially unlocking improvements in quantum communication, metrology, and the inversion of complex dynamics. The study not only advances theoretical understanding of quantum mechanics but also suggests tangible engineering implications for intelligent systems. By formulating quantum integrated sensing and computation, the researchers have created a system where a shared quantum state acts as both a sensing probe and a computational engine. Crucially, the same quantum state serves as the basis for both sensing and computation, streamlining the process and potentially enhancing efficiency. This approach avoids the need for separate, dedicated states for each task, a common limitation in traditional architectures. The agent experiences a superposition of two operational pathways: one where observation precedes computation, and another where computation precedes observation. This contrasts with conventional information processing, which rigidly enforces a strict order where sensing always precedes computation. To implement this ICO framework, the research leveraged a solid-state quantum sensor for state observation, enabling precise measurements of the surrounding environment. The sensor’s output then informed a parametric model, a mathematical representation of the system’s dynamics, used to generate predictions based on the observed state. The successful navigation results suggest that this integrated approach is viable and offers potential advantages over conventional, strictly ordered information processing paradigms. Specifically, the agent successfully learned to predict headings for navigation based on sensed magnetic field data. Furthermore, the study establishes a framework for quantum integrated sensing and computation where both tasks are performed using the same quantum state. This shared quantum state, represented as an agent, undergoes a state-dependent perturbation during observation, imprinting the sensed information onto its quantum properties. The parametric model, trained using a dataset of K states and corresponding targets, then extracts predictions from this perturbed state. Scientists are increasingly exploring information processing paradigms that move beyond strict linear causality. While seemingly abstract, this approach tackles a longstanding challenge in complex environments where the optimal sequence for gathering and interpreting data is often unclear. Traditional systems demand a definitive order, potentially missing crucial information or incurring delays as they adapt to unforeseen circumstances. The implications extend beyond incremental improvements in existing algorithms; the demonstrated success in a magnetic navigation task hints at a pathway towards more robust and adaptable systems, particularly in scenarios demanding real-time decision-making with imperfect or ambiguous data. However, significant hurdles remain. The current experiments are limited to a specific task and rely on relatively simple models. Scaling this approach to more complex scenarios and demonstrating its resilience to noise and errors will be critical. Future research will likely focus on developing more efficient methods for harnessing this principle, perhaps through hybrid quantum-classical architectures, and exploring its potential in other domains like medical diagnostics or financial modelling.