2024 Industry Abstract
Optimization of 3-D flight trajectory of variable trim kites for airborne wind energy production
arXiv:2403.00382
Information according to § 5 TMG (German Telemedia Act):
Dr. Rafał Noga
Waldstr. 3
94149 Kößlarn
Germany
Contact
Phone: +49 175 617 6792
E-mail:
VAT ID (§ 27a UStG)
DE457893621
Academic title
Dr. (doctorate awarded in Spain)
Responsible for content according to § 55 Abs. 2 RStV: Dr. Rafał Noga (address as above).
This site uses no cookies, no external fonts and no tracking. The server keeps only technical logs (IP, timestamp, requested file) necessary to operate, purged regularly. If you contact me by e-mail or phone, your data will be processed only to answer your inquiry.
Data controller
Dr. Rafał Noga
Waldstr. 3, 94149 Kößlarn, Germany
E-mail:
Independent MPC consulting, development and implementation for process industries, machine building, energy and autonomous systems.
16 design patterns · 5 application domains · worldwide consulting
Controls that no longer fit a new machine or plant generation? Start with the MPC Remote Diagnostic: 5 days, €4,500 fixed, fully remote.
Model Predictive Control (MPC) is an advanced control strategy that uses an internal model of the system to repeatedly solve an optimization problem over a finite prediction horizon. At each time step, the optimal sequence of future inputs is computed — but only the first action is applied, and the horizon rolls forward. This "receding horizon" principle makes MPC uniquely powerful: it anticipates future behavior, respects hard constraints on states and inputs, and handles competing objectives simultaneously.
Physical limits (actuator bounds, safety envelopes, process constraints) are embedded directly in the optimization — no ad-hoc clamping.
MPC natively handles MIMO systems. Interactions between channels are captured by the model and balanced automatically.
When future references or disturbances are known (trajectory, weather, demand forecast), MPC exploits this information proactively.
NMPC extends the framework to nonlinear dynamics. Mixed-integer MPC handles hybrid continuous/discrete logic.
State estimation feeds the MPC with unmeasured states — closing the loop on quantities that cannot be sensed directly.
soft-sensor.comEconomic MPC directly minimizes operating costs, energy consumption, or yield loss — control aligned with business objectives.
Over years of industrial and research projects, 16 recurring design patterns have been distilled. Each covers a specific MPC challenge — from power electronics at µs timescales to whole-body locomotion control — with a problem statement, formulation, implementation notes, and literature references.
Direct cost minimization — no intermediate setpoint. Control decisions are business decisions.
Read pattern → Pattern 02Physics-based NMPC for continuous and batch processes: energy, quality, and throughput in one controller.
Read pattern → Pattern 03Deterministic optimization on DSP, microcontrollers, and FPGAs with hard real-time guarantees.
Read pattern → Pattern 04MPC for power electronics at microsecond timescales, replacing hysteresis and PWM strategies.
Read pattern → Pattern 05One model-based design replaces parameter-dense calibration tables and generalizes across conditions.
Read pattern → Pattern 06Nonlinear MPC for agile flight, disturbance rejection, and multi-vehicle swarm coordination.
Read pattern → Pattern 07Layered MPC for ground vehicles: trajectory planning + model-based tracking with collision avoidance.
Read pattern → Pattern 08Sub-millimeter trajectory tracking through constrained dynamic optimization of manipulator motion.
Read pattern → Pattern 09Treats contact as a variable to optimize — not a disturbance to reject — for safe physical collaboration.
Read pattern → Pattern 10Constraints-first MPC: friction cones, torque limits, and foot placement resolved in real time.
Read pattern → Pattern 11Decouples path progress from tracking error — more speed from the same hardware, tighter contour accuracy.
Read pattern → Pattern 12Two-stage pattern: optimal plans computed offline, executed by real-time MPC online.
Read pattern → Pattern 13Robust MPC that accounts for perception noise and localization error — guarantees safety under uncertainty.
Read pattern → Pattern 14Physics backbone + learned residual closes the model-reality gap at runtime without reidentification.
Read pattern → Pattern 15Why estimator quality and latency directly determine closed-loop performance — and how to co-design both.
Read pattern → Pattern 16Systematic solution for coupled, actuator-saturated multi-output systems in marine and underwater vehicles.
Read pattern →"…we were always extremely satisfied with Mr. Noga."
"Particularly noteworthy are his exceptionally analytical mind and his very quick grasp."
"Even for difficult problems he always found effective solutions …"
"recommend, without any reservation …"
"work independently and with initiative"
Note: Company names are for context on previous positions and do not represent client endorsements.
A free 30-minute call to understand your system, constraints, and objectives. We assess MPC applicability and identify the key challenges — no preparation needed on your side.
Book a 30-min technical fit callA written proposal delivered within one week: MPC structure, state estimation approach, software stack, and implementation roadmap with effort estimates.
Model identification, controller design, code generation, hardware-in-the-loop testing, and commissioning — delivered as working software on your target platform.
Peer-reviewed research underpinning the MPC and state estimation methods applied in consulting projects.
arXiv:2403.00382
IECON 2018 - 44th Annual Conference of the IEEE Industrial Electronics Society
5th IFAC Conference on Nonlinear Model Predictive Control (NMPC), Seville. IFAC-PapersOnLine, 48(23), pp. 440-445
PhD Thesis, University of Valladolid
53rd IEEE Conference on Decision and Control (CDC), pp. 3530-3535
18th International Conference on Process Control, Tatranská Lomnica, Slovakia
18th IFAC World Congress, pp. 3647-3652
19th IEEE International Conference on Control Applications (CCA), Yokohama, Japan, pp. 1654-1659
Technical Report, CERN
MSc Thesis (Joint: Univ. Valladolid, ENSIEG Grenoble, Univ. Karlsruhe, Politechnika Gdańska)
Occasional updates on MPC methods, design patterns, and industrial applications. No spam.
€4,500 fixed · 5 days · fully remote
New machine or plant generation? Ported controls from the previous design that no longer track the new dynamics?
A focused 5-day remote engagement to assess your process-control problem. I work from your historian data and short interviews with your team to deliver a clear, actionable diagnostic — no plant visit, no long sales cycle, no commitment to a bigger contract.
You get: 1-page executive summary + technical report + 60-min Q&A call.
You provide: ≥ 1 week of historian data, P&ID or schematic, 1–2 short interviews with your operators/engineers, NDA signed before any data exchange.
Dr. Rafał Noga is a control engineer specializing in model predictive control, state estimation and soft sensors, and real-time optimization. PhD from the University of Valladolid, with the research carried out at CERN. Advanced model-based control since 2007: LHC cryogenics at CERN, utility-scale wind turbines at IAV, and airborne wind energy at SkySails.
Full profile on noga.es →