Pilot-Quantum: Multifaceted Control

• Pilot-Quantum is a multifaceted concept uniting guiding approaches from pilot-wave ontology to quantum communication and AI-driven experiment control.
• It standardizes diverse techniques by employing auxiliary states, reserved resources, and coordinated deployments to stabilize and enhance fragile quantum processes.
• Applications span from foundational stochastic trajectory modeling and error correction in quantum channels to middleware management in hybrid quantum-HPC systems and manufacturing pilot lines.

Pilot-Quantum does not denote a single standardized doctrine across current quantum research. In the cited literature, the term and its close cognates span several distinct technical uses: pilot-wave formulations that supplement the wave function with beables, pilot states and pilot tones that convey channel or phase information in quantum communication, pilot abstractions for quantum-HPC middleware, pilot lines and pilot deployments for quantum technologies, and AI copilots for autonomous quantum experimentation (Struyve, 2011, Gyongyosi et al., 2012, Laudenbach et al., 2017, Mantha et al., 2024, Coutard et al., 2019, Sha et al., 7 Aug 2025). Taken together, these usages make “pilot” a recurrent control concept in quantum science: it may guide ontology, stabilize phase, reserve resources, validate manufacturability, structure live deployment, or coordinate experiments.

1. Pilot-wave foundations and ontology

In the foundational literature, the pilot component is the guiding structure of de Broglie–Bohm theory. A non-relativistic system of $n$ spinless particles is described by a wave function $\psi(x,t)=\psi({\bf x}_1,\dots,{\bf x}_n,t)$ satisfying the Schrödinger equation

$$i\hbar \frac{\partial \psi(x,t)}{\partial t} = \left( -\sum_{k=1}^n \frac{\hbar^2}{2m_k}\nabla_k^2 + V(x) \right)\psi(x,t),$$

together with actual particle positions ${\bf x}_1(t),\dots,{\bf x}_n(t)$ obeying the guidance law

$$\frac{d {\bf x}_k(t)}{dt} = \frac{\hbar}{m_k}\,\mathrm{Im}\,\frac{\nabla_k \psi(x,t)}{\psi(x,t)}\Bigg|_{x=x(t)} = \frac{1}{m_k}\nabla_k S(x,t)\Big|_{x=x(t)},$$

where $\psi=|\psi|e^{iS/\hbar}$. The associated continuity equation preserves the $|\psi|^2$ distribution, yielding equivariance and quantum equilibrium. In quantum field theory, this program bifurcates into field ontologies and particle ontologies. Bosonic field ontology is natural in the functional Schrödinger representation, for example with Bohm’s transverse electromagnetic field beable ${\bf A}^T({\bf x})$, but fermionic field ontology is technically difficult: Euler-angle constructions are ontologically opaque or empirically inadequate, while Grassmann-field proposals lack a natural equilibrium measure. Alternatives include minimalist bosonic-only ontologies, Bell’s stochastic lattice model, continuum jump models with creation and annihilation, and deterministic Dirac-sea approaches; none of the reviewed pilot-wave QFTs is fundamentally Lorentz invariant, and ultraviolet regularization remains central (Struyve, 2011).

Spin does not require extra spin beables in Bell’s version of Bohmian mechanics. For a spin-$\tfrac12$ particle with spinor $\Psi$, the density is $\psi(x,t)=\psi({\bf x}_1,\dots,{\bf x}_n,t)$0 and the current is

$\psi(x,t)=\psi({\bf x}_1,\dots,{\bf x}_n,t)$1

so the guidance law remains $\psi(x,t)=\psi({\bf x}_1,\dots,{\bf x}_n,t)$2. On this account, Stern–Gerlach apparatuses do not reveal pre-existing spin values; they split the spinor wave into spatial branches, and the actual particle trajectory enters one branch, which then acts as the effective wave function. Contextuality and Bell nonlocality are handled dynamically through the full system-plus-apparatus configuration and, for entangled systems, through conditional wave functions rather than hidden spin vectors (Norsen, 2013).

2. Advanced pilot-wave developments

Later pilot-wave work extends the basic framework in several orthogonal directions. One line replaces deterministic continuum trajectories with a discrete-space, discrete-time $\psi(x,t)=\psi({\bf x}_1,\dots,{\bf x}_n,t)$3-equivariant Markov chain. In that construction, the actual configuration $\psi(x,t)=\psi({\bf x}_1,\dots,{\bf x}_n,t)$4 lives on a discrete configuration space $\psi(x,t)=\psi({\bf x}_1,\dots,{\bf x}_n,t)$5, and the one-step stochastic matrix is selected by a discrete Hamilton principle that minimizes the average action

$\psi(x,t)=\psi({\bf x}_1,\dots,{\bf x}_n,t)$6

For free nonrelativistic particles, this reduces to minimizing total mean-square displacement, so stochasticity is made as small as possible subject to the quantum marginal constraints. The resulting theory is a stochastic analogue of Bohmian mechanics rather than a deterministic lattice transcription (Gluza et al., 2016).

A separate line emphasizes that individual Bohmian trajectories, though real in the theory, are generally not operationally detectable in quantum equilibrium. Weak-velocity measurements reconstruct a phase-gradient field, but they do not uniquely reveal actual Bohmian paths; nor do “surrealistic trajectory” experiments refute them. The deeper point is underdetermination: alternative guidance laws of the form

$\psi(x,t)=\psi({\bf x}_1,\dots,{\bf x}_n,t)$7

preserve the same $\psi(x,t)=\psi({\bf x}_1,\dots,{\bf x}_n,t)$8-statistics, so standard measurement records do not single out the canonical trajectory law. This shifts the interpretive burden from trajectory reality to the significance of measurement outcomes (Fankhauser, 10 Mar 2025).

Relativistic and statistical extensions further diversify the program. A Bohmian treatment of the Dirac equation uses the conserved current $\psi(x,t)=\psi({\bf x}_1,\dots,{\bf x}_n,t)$9 and the covariant guidance law

$$i\hbar \frac{\partial \psi(x,t)}{\partial t} = \left( -\sum_{k=1}^n \frac{\hbar^2}{2m_k}\nabla_k^2 + V(x) \right)\psi(x,t),$$0

and in the Klein-paradox regime reinterprets negative-energy segments via the Feynman–Stueckelberg picture, so a single continuous worldline can represent pair-production-like processes in strong potentials (Dodaro, 2013). At the statistical level, one proposal studies whether a single realized Bohm trajectory can substitute for an ensemble; in a six-rotor model, the long-time empirical coordinate distribution of one rotor closely tracks the subsystem’s time-averaged marginal quantum distribution, suggesting a subsystem-level single-trajectory counterpart to Born’s rule (Avanzini et al., 2015). Another line argues that non-normalizable states are physically meaningful in pilot-wave theory because probabilities attach to a normalized configuration density $$i\hbar \frac{\partial \psi(x,t)}{\partial t} = \left( -\sum_{k=1}^n \frac{\hbar^2}{2m_k}\nabla_k^2 + V(x) \right)\psi(x,t),$$1, not directly to $$i\hbar \frac{\partial \psi(x,t)}{\partial t} = \left( -\sum_{k=1}^n \frac{\hbar^2}{2m_k}\nabla_k^2 + V(x) \right)\psi(x,t),$$2; this motivates a generalized pilot-wave equilibrium on a support $$i\hbar \frac{\partial \psi(x,t)}{\partial t} = \left( -\sum_{k=1}^n \frac{\hbar^2}{2m_k}\nabla_k^2 + V(x) \right)\psi(x,t),$$3 and a new $$i\hbar \frac{\partial \psi(x,t)}{\partial t} = \left( -\sum_{k=1}^n \frac{\hbar^2}{2m_k}\nabla_k^2 + V(x) \right)\psi(x,t),$$4-function $$i\hbar \frac{\partial \psi(x,t)}{\partial t} = \left( -\sum_{k=1}^n \frac{\hbar^2}{2m_k}\nabla_k^2 + V(x) \right)\psi(x,t),$$5 (Sen, 2022). Environment-induced relaxation has also been modeled through a Bohmian Langevin description derived from a Caldeira–Leggett bath, providing a Brownian mechanism for approach to $$i\hbar \frac{\partial \psi(x,t)}{\partial t} = \left( -\sum_{k=1}^n \frac{\hbar^2}{2m_k}\nabla_k^2 + V(x) \right)\psi(x,t),$$6 (Drezet, 2018). Most recently, a non-Hermitian pilot-wave formulation has proposed local Weyl scale invariance via a complexified gauge coupling $$i\hbar \frac{\partial \psi(x,t)}{\partial t} = \left( -\sum_{k=1}^n \frac{\hbar^2}{2m_k}\nabla_k^2 + V(x) \right)\psi(x,t),$$7, with equilibrium density modified from $$i\hbar \frac{\partial \psi(x,t)}{\partial t} = \left( -\sum_{k=1}^n \frac{\hbar^2}{2m_k}\nabla_k^2 + V(x) \right)\psi(x,t),$$8 to the trajectory-dependent ratio $$i\hbar \frac{\partial \psi(x,t)}{\partial t} = \left( -\sum_{k=1}^n \frac{\hbar^2}{2m_k}\nabla_k^2 + V(x) \right)\psi(x,t),$$9 (Sen et al., 7 Jan 2026).

3. Pilot states and pilot tones in quantum communication

In quantum communication, “pilot” refers not to ontology but to an auxiliary state or reference signal that captures a channel transformation or phase relation. In space–earth links, Pilot Quantum Error Correction addresses a channel modeled by an unknown polarization rotation

${\bf x}_1(t),\dots,{\bf x}_n(t)$0

combined with slight depolarization. The channel action is stored in a pilot qubit

${\bf x}_1(t),\dots,{\bf x}_n(t)$1

and the receiver uses Hadamard and CNOT gates to correct a damaged data qubit probabilistically. With an ${\bf x}_1(t),\dots,{\bf x}_n(t)$2-qubit pilot string ${\bf x}_1(t),\dots,{\bf x}_n(t)$3, the success probability is

${\bf x}_1(t),\dots,{\bf x}_n(t)$4

The paper tabulates ${\bf x}_1(t),\dots,{\bf x}_n(t)$5 raw pilot states and ${\bf x}_1(t),\dots,{\bf x}_n(t)$6, and in a medium-Earth-orbit example with ${\bf x}_1(t),\dots,{\bf x}_n(t)$7 transmitted qubits this gives redundancy ${\bf x}_1(t),\dots,{\bf x}_n(t)$8. For the assumed slightly depolarizing channel with ${\bf x}_1(t),\dots,{\bf x}_n(t)$9, the cited classical and quantum capacities are $$\frac{d {\bf x}_k(t)}{dt} = \frac{\hbar}{m_k}\,\mathrm{Im}\,\frac{\nabla_k \psi(x,t)}{\psi(x,t)}\Bigg|_{x=x(t)} = \frac{1}{m_k}\nabla_k S(x,t)\Big|_{x=x(t)},$$0 and $$\frac{d {\bf x}_k(t)}{dt} = \frac{\hbar}{m_k}\,\mathrm{Im}\,\frac{\nabla_k \psi(x,t)}{\psi(x,t)}\Bigg|_{x=x(t)} = \frac{1}{m_k}\nabla_k S(x,t)\Big|_{x=x(t)},$$1, respectively, before pilot overhead (Gyongyosi et al., 2012).

In continuous-variable QKD, pilot tones serve as strong optical references that restore coherence between a weak quantum signal and an independently generated local oscillator. One true-LO CV-QKD system uses a $$\frac{d {\bf x}_k(t)}{dt} = \frac{\hbar}{m_k}\,\mathrm{Im}\,\frac{\nabla_k \psi(x,t)}{\psi(x,t)}\Bigg|_{x=x(t)} = \frac{1}{m_k}\nabla_k S(x,t)\Big|_{x=x(t)},$$2 Mbaud quantum signal and a $$\frac{d {\bf x}_k(t)}{dt} = \frac{\hbar}{m_k}\,\mathrm{Im}\,\frac{\nabla_k \psi(x,t)}{\psi(x,t)}\Bigg|_{x=x(t)} = \frac{1}{m_k}\nabla_k S(x,t)\Big|_{x=x(t)},$$3 GHz pilot generated by carrier-suppressed optical single-sideband modulation, multiplexed in both polarization and frequency. The pilot and quantum branches are detected separately, with low-noise balanced receivers for the quantum channel and higher-bandwidth receivers for the pilot. The pilot is bandpass filtered, used to estimate optical frequency offset and phase drift, and then applied symbol-wise to recover the quantum data. Over $$\frac{d {\bf x}_k(t)}{dt} = \frac{\hbar}{m_k}\,\mathrm{Im}\,\frac{\nabla_k \psi(x,t)}{\psi(x,t)}\Bigg|_{x=x(t)} = \frac{1}{m_k}\nabla_k S(x,t)\Big|_{x=x(t)},$$4, $$\frac{d {\bf x}_k(t)}{dt} = \frac{\hbar}{m_k}\,\mathrm{Im}\,\frac{\nabla_k \psi(x,t)}{\psi(x,t)}\Bigg|_{x=x(t)} = \frac{1}{m_k}\nabla_k S(x,t)\Big|_{x=x(t)},$$5, $$\frac{d {\bf x}_k(t)}{dt} = \frac{\hbar}{m_k}\,\mathrm{Im}\,\frac{\nabla_k \psi(x,t)}{\psi(x,t)}\Bigg|_{x=x(t)} = \frac{1}{m_k}\nabla_k S(x,t)\Big|_{x=x(t)},$$6, and $$\frac{d {\bf x}_k(t)}{dt} = \frac{\hbar}{m_k}\,\mathrm{Im}\,\frac{\nabla_k \psi(x,t)}{\psi(x,t)}\Bigg|_{x=x(t)} = \frac{1}{m_k}\nabla_k S(x,t)\Big|_{x=x(t)},$$7 km, the measured total excess noise with the preferred carrier-suppressed pilot was $$\frac{d {\bf x}_k(t)}{dt} = \frac{\hbar}{m_k}\,\mathrm{Im}\,\frac{\nabla_k \psi(x,t)}{\psi(x,t)}\Bigg|_{x=x(t)} = \frac{1}{m_k}\nabla_k S(x,t)\Big|_{x=x(t)},$$8, $$\frac{d {\bf x}_k(t)}{dt} = \frac{\hbar}{m_k}\,\mathrm{Im}\,\frac{\nabla_k \psi(x,t)}{\psi(x,t)}\Bigg|_{x=x(t)} = \frac{1}{m_k}\nabla_k S(x,t)\Big|_{x=x(t)},$$9, $\psi=|\psi|e^{iS/\hbar}$0, and $\psi=|\psi|e^{iS/\hbar}$1 SNU under averaged calibration; for the $\psi=|\psi|e^{iS/\hbar}$2 km deployed-fiber case, the paper gives an asymptotic secure-key-rate estimate of $\psi=|\psi|e^{iS/\hbar}$3 Mbit/s under a trusted-receiver model and Gaussian-modulation assumptions (Laudenbach et al., 2017).

A related Gaussian-modulated LLO-CVQKD system uses a $\psi=|\psi|e^{iS/\hbar}$4 GHz pilot tone, frequency multiplexing, polarization multiplexing, and heterodyne detection so that $\psi=|\psi|e^{iS/\hbar}$5 and $\psi=|\psi|e^{iS/\hbar}$6 are measured simultaneously without random basis selection. The pilot phase is used to compensate fast drift from two independent lasers, while a disclosed training sequence removes slow differential channel drift. For a $\psi=|\psi|e^{iS/\hbar}$7 km optical fiber, the reported asymptotic secure key rate is $\psi=|\psi|e^{iS/\hbar}$8 Mbit/s and the finite-size rate for a block of $\psi=|\psi|e^{iS/\hbar}$9 is $|\psi|^2$0 Mbit/s, with mean measured excess noise around $|\psi|^2$1 SNU and mean worst-case finite-size excess noise around $|\psi|^2$2 SNU (Wang et al., 2020).

4. Pilot abstractions, operating systems, and runtime management

In quantum-HPC systems, the exact title “Pilot-Quantum” refers to middleware built on the Pilot Abstraction. The central idea is to decouple resource acquisition from task execution by submitting placeholder jobs, or pilots, that reserve heterogeneous resources in advance and then schedule application tasks onto them. The architecture is organized into four layers—workflow, workload, task, and resource—and exposes pilot_description, task_description, and submit_task interfaces. A Pilot-Manager provisions and releases resources, coordinates pilots, and assigns tasks, while Pilot Agents execute tasks directly or through plugins such as Ray and Dask; tightly coupled jobs can use MPI, CUDA, and Slurm srun. The implementation targets task-parallel hybrid workloads such as variational algorithms, circuit cutting, and quantum machine learning, with demonstrated mini-apps including circuit execution on IBM Eagle and simulators, distributed state-vector simulation, circuit cutting, and QML pipelines. On Perlmutter, the reported average pilot setup time was $|\psi|^2$3 s; zero-compute-task throughput ranged from $|\psi|^2$4 tasks/s for $|\psi|^2$5 tasks to $|\psi|^2$6 tasks/s for $|\psi|^2$7 tasks. The evaluation also reports $|\psi|^2$8-qubit distributed simulation on $|\psi|^2$9 GPUs, up to ${\bf A}^T({\bf x})$0 speedup for the QML compression workflow on ${\bf A}^T({\bf x})$1 nodes, and ${\bf A}^T({\bf x})$2 speedup for one circuit-cutting scenario (Mantha et al., 2024).

A related systems use of “pilot” appears in “Origin Pilot,” a quantum operating system for efficient usage of quantum resources. Origin Pilot integrates quantum task scheduling, quantum resource management, quantum program compilation, and automatic calibration. It defines OS-like abstractions—quantum application, quantum task, quantum transaction, quantum thread, quantum processor, and quantum resource—and couples them to a scheduler that uses FCFS for calibration tasks and HRRN for general quantum tasks. Calibration is triggered when single-gate fidelity falls below ${\bf A}^T({\bf x})$3 or double-gate fidelity below ${\bf A}^T({\bf x})$4, with an interval that starts at ${\bf A}^T({\bf x})$5 minutes and is gradually reduced to ${\bf A}^T({\bf x})$6 minutes; block partitioning allows executable and calibration regions to coexist. The system also implements topology-aware qubit mapping and supports multiple quantum processors, parallel execution, and hybrid computing resources such as CPU, GPU, FPGA, and HPC clusters. Its evaluation includes GHZ workloads on superconducting processors, mapping benchmarks against BMT, and qualitative gains in completed task counts under automatic calibration (Kong et al., 2021).

5. Pilot lines and pilot deployments in quantum technology

In translational quantum engineering, “pilot” can denote either a pilot line for manufacturing or a pilot deployment in the field. A prominent manufacturing example is the fabrication of distributed-feedback quantum cascade lasers on a ${\bf A}^T({\bf x})$7 mm CMOS/MEMS pilot line. In that process, a ${\bf A}^T({\bf x})$8-inch InP wafer carrying the active QCL stack is directly bonded to an oxidized ${\bf A}^T({\bf x})$9 mm Si wafer, the substrate is removed, and a top-metal grating DFB structure is fabricated using CMOS-compatible processing. The reported $\tfrac12$0 devices, based on lattice-matched $\tfrac12$1, achieved threshold current density down to $\tfrac12$2, linewidth $\tfrac12$3 under short-pulse FTIR measurement, about $\tfrac12$4 mW optical power at ambient temperature and $\tfrac12$5 duty cycle, and a characteristic temperature $\tfrac12$6 K. Wafer-level testing covered $\tfrac12$7 devices, with about $\tfrac12$8 relative standard deviation in threshold current density and an estimated $\tfrac12$9 yield of functional lasers. The paper positions this as a route toward low-cost co-integration of mid-IR QCL sources with SiGe-based waveguides and planar optical sensors (Coutard et al., 2019).

A field-deployment example is “Quantum Shuttle,” a quantum-optimization-based traffic pilot during Web Summit 2019 in Lisbon. The service combined historical mobility analysis to design temporary bus lines with live route optimization through a custom Android app, a cloud backend, HERE traffic-aware routing, and D-Wave optimization. The live system operated from November 5 to November 7 with a fleet of $\Psi$0 buses, solving $\Psi$1 optimization problems with average response time $\Psi$2 s overall and $\Psi$3 s for the $\Psi$4 harder instances. The largest live QUBO had $\Psi$5 variables. Of $\Psi$6 recorded trips, $\Psi$7 were treated as valid Quantum Shuttle trips. The online objective minimized route overlap across active buses, and the authors present the deployment as, to their knowledge, the first commercial application depending on a quantum processor to perform a critical live task (Yarkoni et al., 2020).

6. AI copilots for quantum experiments

A more recent usage places the pilot function in the software agent itself. QCopilot is a centralized LLM-based multi-agent framework for quantum sensor development and diagnosis. Its architecture comprises a Decision Maker, Experimenter, Analyst, Multimodal Diagnoser, Web Searcher, and Recorder, backed by a vector knowledge base and commercial LLMs with few-shot prompt engineering. The system combines retrieval-augmented access to hardware parameter reports, active learning, Bayesian optimization, multi-objective Bayesian optimization, uncertainty-oriented correlation analysis, multimodal image comparison, and memory accumulation across runs. In the demonstrated cold-atom workflow, the Decision Maker first optimized magneto-optical trapping, then polarization-gradient cooling, and later switched to diagnosis when operation deviated from baseline. The Experimenter used single-objective Bayesian optimization for MOT loading, multi-objective Bayesian optimization for PGC, and optimal Latin hypercube sampling during fault diagnosis. Applied to atom cooling experiments, QCopilot generated $\Psi$8 sub-$\Psi$9K atoms without human intervention within a few hours, representing approximately $\psi(x,t)=\psi({\bf x}_1,\dots,{\bf x}_n,t)$00 speedup over manual experimentation. In the reported PGC study over $\psi(x,t)=\psi({\bf x}_1,\dots,{\bf x}_n,t)$01 setups, the counts of acceptable points below $\psi(x,t)=\psi({\bf x}_1,\dots,{\bf x}_n,t)$02 were $\psi(x,t)=\psi({\bf x}_1,\dots,{\bf x}_n,t)$03 for MBO, $\psi(x,t)=\psi({\bf x}_1,\dots,{\bf x}_n,t)$04 for temperature-only BO, and $\psi(x,t)=\psi({\bf x}_1,\dots,{\bf x}_n,t)$05 for atom-number-only BO; the authors argue that MBO better concentrates points near the Pareto frontier. The same framework is also used to detect anomalous parameters by comparing normal and abnormal CCD images and changes in correlation matrices (Sha et al., 7 Aug 2025).

7. Conceptual commonalities and boundaries

A common misconception is that Pilot-Quantum names a single mature subfield. The cited record does not support that interpretation. The exact phrase denotes a quantum-HPC middleware in one paper (Mantha et al., 2024), whereas closely related literature uses “pilot” to mean a guiding ontology in hidden-variable theory (Struyve, 2011), a quantum pilot state for channel correction (Gyongyosi et al., 2012), a strong pilot tone for phase compensation (Laudenbach et al., 2017), a CMOS pilot line for device industrialization (Coutard et al., 2019), or an AI copilot for experiment control (Sha et al., 7 Aug 2025).

A plausible commonality is functional rather than disciplinary. Across these works, the pilot element carries structured side information that would otherwise have to be inferred expensively or left ambiguous. In pilot-wave theory it is the ontological supplement that makes measurement outcomes definite; in quantum communication it is the state or tone that captures channel action or phase drift; in middleware it is the reserved placeholder that mediates between application workloads and heterogeneous resources; in industrialization it is the staging environment that validates manufacturability; in autonomous laboratories it is the coordinating agent that decomposes and supervises closed-loop experimentation. This suggests a recurring technical pattern: the pilot is introduced to stabilize, guide, or interpret a quantum process without collapsing the entire problem into brute-force estimation or centralized monolithic control (Mantha et al., 2024, Struyve, 2011, Gyongyosi et al., 2012, Laudenbach et al., 2017, Coutard et al., 2019, Sha et al., 7 Aug 2025).

The uses are nevertheless not reducible to one another. Pilot-wave realism is a foundational framework; pilot tones and pilot states are communication primitives; pilot abstractions are runtime systems; pilot lines and pilot deployments belong to translational engineering; and copilots are AI coordination layers. Even within the foundational branch, the presence of a pilot does not imply direct empirical readability of microscopic paths: individual Bohmian trajectories remain generically undetectable in quantum equilibrium, despite weak-measurement and surreal-trajectory debates (Fankhauser, 10 Mar 2025). The term therefore functions best as a polysemous family label for several quantum research programs in which a pilot component guides, regularizes, or operationalizes an otherwise fragile quantum task.