Quantum Integrated Sensing and Computation with Indefinite Causal Order

Quantum operations with indefinite causal order (ICO) represent a framework in quantum information processing where the relative order between two events can be indefinite. In this paper, we investigate whether sensing and computation, two canonical tasks in quantum information processing, can be carried out within the ICO framework. We propose a scheme for integrated sensing and computation that uses the same quantum state for both tasks. The quantum state is represented as an agent that performs state observation and learns a function of the state to make predictions via a parametric model. Under an ICO operation, the agent experiences a superposition of orders, one in which it performs state observation and then executes the required computation steps, and another in which the agent carries out the computation first and then performs state observation. This is distinct from prevailing information processing and machine intelligence paradigms where information acquisition and learning follow a strict causal order, with the former always preceding the latter. We provide experimental results and we show that the proposed scheme can achieve small training and testing losses on a representative task in magnetic navigation.

💡 Research Summary

The paper investigates whether the two fundamental tasks of quantum information processing—information acquisition (sensing) and information processing (computation)—can be performed within the framework of indefinite causal order (ICO). The authors first formulate a conventional quantum integrated sensing and computation (QISC) scheme in which a shared N‑qubit quantum state |ψ⟩ acts as an “agent”. The environment imprints a physical parameter x onto the state via a unitary U(x), yielding the perturbed state |ψ(x)⟩ = U(x)|ψ⟩. A parametrized quantum circuit Uθ then processes this state, and a final measurement produces a prediction ŷ. The loss Lθ(x) = (y – ŷ)² is minimized by gradient descent, with gradients obtained either by the parameter‑shift rule on hardware or by automatic differentiation in simulation.

To introduce ICO, the authors employ a quantum SWITCH, a higher‑order operation that coherently controls the order of two gates. An auxiliary “order qubit” is prepared in the superposition |+⟩T = (|0⟩ + |1⟩)/√2. Conditional on the order qubit, the system evolves either as Uθ ∘ U(x) (if T=0) or U(x) ∘ Uθ (if T=1). The overall evolution is therefore a coherent superposition of the two possible causal orders, embodying causal non‑separability as a computational resource. The output state after the SWITCH is measured to obtain ŷ, and the same loss‑minimization procedure is applied to update θ.

The authors validate the ICO‑enhanced QISC on a concrete task: magnetic navigation. Three qubits encode the three components of the Earth’s magnetic field (Bx, By, Bz) via single‑qubit rotations Rx, Ry, Rz. The target heading y is defined as y = 2·arctan(Bx, By) (degrees). A hardware‑efficient ansatz consisting of 15 layers of parametrized single‑qubit rotations and cyclic CNOTs (N=3) is used for Uθ, with an additional qubit serving as the order qubit. Training uses 200 labeled examples, batch size 32, learning rate 0.01, and 200 epochs; gradients are computed by back‑propagation in a simulator, and results are averaged over five independent runs.

Experimental results show that the ICO scheme achieves low training loss (≈0.02) and low test loss (≈0.03), indicating both accurate fitting and good generalisation. Moreover, the ICO approach converges faster than the baseline where sensing precedes computation in a fixed order, requiring roughly 30 % fewer epochs to reach comparable loss levels. These findings demonstrate that superposing causal orders can provide practical advantages in learning‑based quantum sensing tasks.

The discussion acknowledges several limitations. Implementing a reliable quantum SWITCH remains experimentally challenging; stability, decoherence of the order qubit, and noise in the underlying hardware can degrade performance. The authors suggest exploring alternative parametrized circuits (e.g., symmetry‑preserving ansätze), advanced optimizers, and regularisation techniques to mitigate barren plateaus. They also call for rigorous theoretical analyses of the ultimate performance bounds that ICO can offer for integrated sensing and computation, analogous to the established advantages of ICO in quantum communication and metrology.

In summary, the paper introduces a novel conceptual framework that merges quantum sensing and learning into a single quantum process with indefinite causal order, provides a concrete experimental demonstration on a magnetic navigation task, and highlights both the promise and the open challenges for future research in this emerging direction.