Predictive Spray Switching in Actuation Control

• Predictive spray switching is an anticipatory control paradigm using forward models to time spray activation based on planned trajectories and future coverage.
• It applies across domains—agriculture for nozzle control, combustion for discontinuous injection, painting for color prediction, and electrospray for droplet reversal.
• Efficiency trade-offs involve increased switching events versus reduced redundant spraying, enhanced path efficiency, and optimized application performance.

Searching arXiv for the provided paper and closely related work on predictive spray switching and adjacent spray-prediction domains. Search 1: exact title match for agricultural predictive spray switching. Search 2: related work on multi-section control and sensor-free predictive spray switching. Search 3: adjacent uses of predictive spray switching in combustion, nozzle breakup, painting, and electrospray contexts. Predictive spray switching denotes a family of control and modeling strategies in which spray activation, deactivation, or transition behavior is determined with explicit anticipation of future evolution rather than only instantaneous state. In arable farming, the term refers most directly to a predictive logic for the on- and off-switching of boom-mounted nozzles during area coverage along a planned path pattern (Plessen, 1 Apr 2025). In adjacent literature, closely related ideas appear as discontinuous droplet injection in unsteady spray flamelets, pixel-level prediction of spray-color switching in robotic painting, geometry-conditioned prediction of abrupt breakup transitions in nozzle flows, and local criteria for electrospray-assisted reversal of charged droplets (Huenchuguala et al., 29 Jun 2025, Wang et al., 2024, Ramlau et al., 15 Jun 2026, Kahn, 15 Sep 2025).

1. Conceptual scope and domain-specific meanings

The literature does not present predictive spray switching as a single universal algorithm. Instead, it appears as a recurring design principle: a switching decision is made from a model of what will happen later in the path, trajectory, flow, breakup pattern, or droplet motion. In agriculture, this means that the switching logic uses knowledge of the future traversal and of whether an area will be or will have been covered by another part of the overall pattern (Plessen, 1 Apr 2025). In combustion, discontinuous injection is treated as a time-dependent forcing that modifies evaporation source terms, scalar dissipation rate, temperature, and stability in mixture-fraction space (Huenchuguala et al., 29 Jun 2025). In robotic spray painting, spray-switch timing predictions are tied to the final painted color rather than only paint thickness (Wang et al., 2024). In nozzle-breakup surrogates, the phrase is connected to cases where small geometry changes cause abrupt qualitative changes in breakup pattern (Ramlau et al., 15 Jun 2026). In electrospray, the relevant switching event is local stalling or reversal of a charged droplet under an opposing electric field (Kahn, 15 Sep 2025).

Domain	Predictive object	Representative paper
Arable farming	Nozzle on/off switching along planned coverage patterns	(Plessen, 1 Apr 2025)
Spray flamelets	Continuous vs. discontinuous droplet injection	(Huenchuguala et al., 29 Jun 2025)
Spray painting	When and where to switch spray color on/off	(Wang et al., 2024)
Nozzle breakup	Transition or “switching” in breakup pattern across geometries	(Ramlau et al., 15 Jun 2026)
Electrospray	Minimum electric field for droplet reversal or stalling	(Kahn, 15 Sep 2025)

This distribution of meanings suggests that predictive spray switching is best understood as an anticipatory spray-control paradigm rather than as a term restricted to one sector.

2. Predictive nozzle switching in arable area coverage

The most explicit formulation appears in "Predictive Spray Switching for an Efficient Path Planning Pattern for Area Coverage" (Plessen, 1 Apr 2025). The setting is a machinery carrying a boom aligned along a working width, with a set of nozzles that must be switched on and off while the machinery travels along a specific path planning pattern. Concatenation of multiple path patterns and corresponding concatenation of the proposed switching logics enables nominal lossless spray application for area coverage tasks.

For a single pattern, the predictive switching rules are given as a sequence over labeled path transitions. Between A and D the spray is switched off. Between D and E it is switched on, because the mainfield lane section between those points has not been or will not be more efficiently covered. Between E and J it is switched off because that area will be or will have been covered by another part of the overall pattern. Between J and K it is switched on again. Between K and A it is switched off. Between A and M it is switched on for a final headland segment that includes some previously traversed headland (Plessen, 1 Apr 2025).

Several properties distinguish this logic from purely reactive actuation. First, spraying is purposely not activated during turning maneuvers between headland and mainfield lanes, since boom travel speed varies during turns. Second, redundant spraying is avoided especially along headland segments that are traversed multiple times. Third, boundary handling at the start and end of the field uses reactive switching to ensure full coverage at non-symmetric boundaries.

The same paper contrasts this predictive logic with a state-of-the-art reactive switching logic for Boustrophedon-based path planning. Reactive switching turns the nozzle on when entering an area not yet sprayed and off when overlapping with a previously sprayed area, whereas predictive switching uses a precomputed sequence tied to the path plan (Plessen, 1 Apr 2025). A common misconception is therefore that predictive switching is simply delayed reactive switching; the description in the paper is stricter than that, because the logic depends on explicit lookahead over future traversals.

3. Efficiency trade-offs, overlap, and sectionality

The agricultural literature frames predictive spray switching as part of a coupled path-planning and actuation problem. For an odd number of lanes $N$, the path-length expressions reported for the alternative and Boustrophedon methods are

$$L_\text{path}^\text{Alternative}(N) = N(H-4R+2\frac{2R\pi}{4} + 3W) + c_2$$

and

$$L_\text{path}^\text{Boustrophedon}(N) = N(H-4R+2\frac{2R\pi}{4} + 4W) + c_1.$$

The reported savings are proportional to $-NW$, so a larger number of mainfield lanes or a wider machinery yields greater benefit (Plessen, 1 Apr 2025).

The same analysis reports that both methods achieve complete and lossless area coverage in ideal conditions, specifically instant switching and no spray dynamics or swath bleed at edges. The trade-off is an increased number of spray on/off transitions. For odd $N$,

$$N_\text{ON}^\text{Alternative}(N) = \frac{3}{2}(N + 1),$$

whereas

$N_\text{ON}^\text{Boustrophedon}(N) = N + 1.$

The implication stated in the paper is that each transition may introduce a brief non-ideal spray transient, so the cumulative effect of transients grows with the number of switchings (Plessen, 1 Apr 2025).

A subsequent comparison between Automatic Section Control and simpler one- or two-section alternatives broadens this trade-off analysis (Plessen, 15 Aug 2025). Six methods are evaluated on 10 diverse real-world field examples, including non-convex field contours, freeform mainfield lanes and multiple obstacle areas. The three section set-ups control 48 sections, 2 sections, or all nozzles uniformly as one single section. The alternative path method $M_2$ saves 4–12% in overall pathlength compared to Boustrophedon regardless of spray switching logic. Overlap relative to the ideal reference is reported as 1–6% for ASC with 48 sections, 18–60% for the two-section alternative, and often 20–80% for the one-section alternative (Plessen, 15 Aug 2025).

This comparison is important because it shows that predictive spray switching is not synonymous with highly instrumented automation. The one- and two-section predictive methods are described as suitable for manual driving, pre-planned, sensor-free, and low cost, whereas ASC requires high-precision sensors and automation to operate many sections at once (Plessen, 15 Aug 2025). A plausible implication is that predictive spray switching occupies a design space between reactive fully automated section control and purely manual untimed spraying.

4. Unsteady spray flamelets and discontinuous injection

In combustion research, predictive spray switching appears in the form of discontinuous or switched droplet injection. "Unsteady solutions of the spray flamelet equations" extends spray flamelet formulations to time-dependent situations by writing the governing equations for species mass fractions $Y_k$ and temperature $T$ in mixture fraction space with explicit time derivatives (Huenchuguala et al., 29 Jun 2025):

$$L_\text{path}^\text{Alternative}(N) = N(H-4R+2\frac{2R\pi}{4} + 3W) + c_2$$0

and

$$L_\text{path}^\text{Alternative}(N) = N(H-4R+2\frac{2R\pi}{4} + 3W) + c_2$$1

Closure is obtained by a Lagrangian description of the liquid phase in mixture fraction space. The local evaporation source term is accumulated over droplet groups as

$$L_\text{path}^\text{Alternative}(N) = N(H-4R+2\frac{2R\pi}{4} + 3W) + c_2$$2

This formulation readily accommodates any time-dependent spray injection scenario, including both continuous and discontinuous droplet injection. The paper distinguishes cases labeled $$L_\text{path}^\text{Alternative}(N) = N(H-4R+2\frac{2R\pi}{4} + 3W) + c_2$$3 for continuous injection and $$L_\text{path}^\text{Alternative}(N) = N(H-4R+2\frac{2R\pi}{4} + 3W) + c_2$$4 and $$L_\text{path}^\text{Alternative}(N) = N(H-4R+2\frac{2R\pi}{4} + 3W) + c_2$$5 for discontinuous injection with different injection/no-injection duty cycles. In the discontinuous cases, sharp evaporation and dissipation waves pass through mixture fraction space. Figure 1 is reported to show that discontinuous injection leads to higher and oscillatory flamelet temperatures at stoichiometry, $$L_\text{path}^\text{Alternative}(N) = N(H-4R+2\frac{2R\pi}{4} + 3W) + c_2$$6, compared to continuous injection for the same overall fuel load. Figure 2 maps flamelet stability and temperatures at stoichiometry over grids of $$L_\text{path}^\text{Alternative}(N) = N(H-4R+2\frac{2R\pi}{4} + 3W) + c_2$$7, and the discontinuous strategy widens the region of parameter space where stable burning solutions are found (Huenchuguala et al., 29 Jun 2025).

The paper’s conclusion is that discontinuous spray injection can be used predictively to enhance flamelet temperature and stability, increase the robustness of combustion under challenging conditions, and potentially control extinction/re-ignition dynamics. This is a different use of “switching” from agricultural nozzle actuation, but the common structure is still anticipatory control of spray-state transitions.

5. Robotic spray painting and geometry-conditioned breakup prediction

In robotic spray painting, predictive spray switching is tied to the final optical outcome. "Solve paint color effect prediction problem in trajectory optimization of spray painting robot using artificial neural network inspired by the Kubelka Munk model" states that conventional thickness-based optimization can only qualitatively reflect the color distribution and cannot simulate the color effect of spray painting at the pixel level (Wang et al., 2024). The proposed regression model predicts painted color from base color, simulated paint film thickness, and paint class:

$$L_\text{path}^\text{Alternative}(N) = N(H-4R+2\frac{2R\pi}{4} + 3W) + c_2$$8

The final SPCP-ANN incorporates gating and residual structure, is trained with MSE and Adam, and is used to replace the traditional thickness simulation method in the objective function of spray gun trajectory optimization. The optimization includes trajectory height $$L_\text{path}^\text{Alternative}(N) = N(H-4R+2\frac{2R\pi}{4} + 3W) + c_2$$9, velocity $$L_\text{path}^\text{Boustrophedon}(N) = N(H-4R+2\frac{2R\pi}{4} + 4W) + c_1.$$0, and spray confidence $$L_\text{path}^\text{Boustrophedon}(N) = N(H-4R+2\frac{2R\pi}{4} + 4W) + c_1.$$1, where $$L_\text{path}^\text{Boustrophedon}(N) = N(H-4R+2\frac{2R\pi}{4} + 4W) + c_1.$$2 decides spray switching, and the objective combines pixel-level color difference with total spraying time. NSGA-II is used for multi-objective Pareto optimization (Wang et al., 2024).

The paper explicitly states that the model can enable spray-switch timing predictions, controlling when and where to switch spray color on/off for precise edge and region demarcation. Reported prediction errors are 7.743% mean relative error on single base color tests and 6.814% on gradient base color tests. The account in the paper is therefore not about nozzle overlap avoidance, but about predictive switching of color deposition in multi-color combination spraying (Wang et al., 2024).

A different predictive mechanism is reported in "Learning Interface Breakup: A Geometry-Conditioned Latent Surrogate for Spray Formation" (Ramlau et al., 15 Jun 2026). The surrogate is trained on 797 high-fidelity CFD simulations with adaptive mesh refinement and uses the AMR cell-density field as a compact proxy. Full trajectory inference is reported as approximately 0.045 seconds per simulation, with a speed-up of more than $$L_\text{path}^\text{Boustrophedon}(N) = N(H-4R+2\frac{2R\pi}{4} + 4W) + c_1.$$3 relative to Basilisk CFD. The model supports “spray switching” or transition prediction in the sense that small geometry changes can cause abrupt qualitative changes in breakup pattern, and the learned latent structure performs better than geometry-nearest neighbors on such cases (Ramlau et al., 15 Jun 2026). This suggests that predictive spray switching can also refer to anticipating regime changes in transient two-phase breakup rather than directly actuating a spray device.

6. Electrospray criteria, assumptions, and recurring limitations

In electrospray-assisted droplet dynamics, predictive spray switching is formulated as a local threshold problem. "Predictive Criteria for Electrospray-Assisted Droplet Dynamics in Aerodynamic Flow Fields" couples a steady OpenFOAM RANS solution of a NACA 1912 airfoil channel flow with a reduced-order Lagrangian particle model (Kahn, 15 Sep 2025). The droplet dynamics satisfy

$$L_\text{path}^\text{Boustrophedon}(N) = N(H-4R+2\frac{2R\pi}{4} + 4W) + c_1.$$4

From the streamwise balance, the paper derives a predictive criterion for the minimum opposing electric field required to slow or reverse the droplet’s streamwise motion:

$$L_\text{path}^\text{Boustrophedon}(N) = N(H-4R+2\frac{2R\pi}{4} + 4W) + c_1.$$5

It also introduces a control-authority map,

$$L_\text{path}^\text{Boustrophedon}(N) = N(H-4R+2\frac{2R\pi}{4} + 4W) + c_1.$$6

which identifies regions where modest fields can produce strong kinematic response (Kahn, 15 Sep 2025).

Across these literatures, several limitations recur. Agricultural predictive switching is nominal in the sense that complete and lossless coverage is asserted under instant switching and without spray dynamics or swath bleed at edges (Plessen, 1 Apr 2025). The multi-section comparison emphasizes uncertainty factors such as cross wind drift, boom height, and nozzle clogging in open-field conditions (Plessen, 15 Aug 2025). The spray flamelet formulation was introduced precisely because prior solutions had been limited to steady evaporation profiles (Huenchuguala et al., 29 Jun 2025). The geometry-conditioned breakup surrogate is trained on a specific family of NURBS nozzle geometries at single Reynolds and Weber numbers, relies on AMR structure produced by Basilisk, and has no explicit conservation enforcement (Ramlau et al., 15 Jun 2026). The electrospray framework assumes one-way coupling and low particle Reynolds number (Kahn, 15 Sep 2025).

Taken together, these works show that predictive spray switching is a cross-domain research theme linking path planning, actuated deposition, unsteady source-term closure, learned transition prediction, and field-induced droplet control. The shared technical premise is that spray behavior is most effectively managed when switching decisions are tied to a forward model of coverage, color outcome, flamelet response, breakup morphology, or local aerodynamic-electrostatic balance.