Frequently Asked Questions

ORTO is not AI-based. It uses deterministic, model-free autonomous agents that learn directly from the live process and write setpoints in closed loop, without building physics models, surrogate models, neural networks, or reinforcement-learning policies. ORTO converges on and tracks the true operating optimum because its “model” is the actual plant, making it inherently robust to drift, nonlinearity, and changing conditions, and simple to deploy (configured much like PID). ORTO runs fully on-premises, installs in a manner consistent with standard control system software, and does not require cloud connectivity.

AI approaches require training across the full operating envelope, which takes time, specialized expertise, and often deliberate movement of plant operating points purely for training purposes, disrupting production. Alternatively, models are required to provide the operating envelope coverage required for training. Once deployed, these systems must also manage drift and often require periodic retraining. When the plant enters regimes outside the training domain, AI-based optimizers can behave unpredictably because (i) they are forced to extrapolate beyond learned experience (ii) they are non-deterministic, in other words, the exact same inputs do not always generate the same outputs. Many also depend on cloud infrastructure, introducing additional latency, cybersecurity, and operational constraints.

In short: ORTO optimizes directly on the plant itself. This avoids training campaigns, eliminates model mismatch, removes the need to disturb operations for data collection, supports secure on-prem deployment using familiar control-system practices, and delivers deterministic behavior with minimal maintenance while continuously tracking the true optimum in real operating conditions.

ORTO Agents do not rely on injected dither or periodic perturbations. Instead, multiple agents work cooperatively to navigate the optimization surface using a deterministic, model-free gradient search based on observed process response. This avoids artificial excitation of the plant and eliminates the time-scale separation requirements typical of dither-based methods.

Yes. ORTO can optimize any objective function provided there is a measurable cause-and-effect relationship between the manipulated variables and the optimization variable. ORTO converges to a local optimum and continuously adapts as the process evolves, which reflects real industrial conditions where objectives are typically nonlinear and time-varying.

No. ORTO does not require deliberate perturbation or dither signals. It learns directly from natural process behavior, so there is no need to impose artificial excitation. ORTO sits above and works in unison with the regulatory controls. 

ORTO treats the live process as its model. Assuming the objective function has been defined and the agents have been configured correctly, ORTO converges to the true local optimum and continuously tracks it as conditions change.

ORTO Agents are always seeking improvement, so when the optimum shifts due to disturbances, equipment aging, or changing operating conditions, ORTO automatically adapts.

Convergence time depends on initial conditions, process dynamics, tuning parameters, and measurement quality. Because ORTO optimizes a real physical system rather than a mathematical model, there is no fixed analytical bound on convergence time. Instead, ORTO prioritizes safe, bounded moves and continuous improvement.

Yes. The ORTO algorithm was originally developed in MATLAB/Simulink, and the standalone ORTO software was built using MATLAB App Designer, making Simulink a natural environment for testing and integration.

ORTO supports on-prem, containerized deployment and OPC connectivity, enabling closed-loop testing against Simulink models, co-simulation during development, and deployment alongside ARM/Debian control stacks.