WorksComputer Science
Feedback-Linearised Trajectory Tracking via Model-Informed Robust Geometric DDPC
Abstract
Data-driven predictive control (DDPC) predicts system responses from recorded trajectories, but nonlinear dynamics can limit the accuracy of fixed linear predictors. This paper studies geometric robust DDPC for trajectory tracking of a fully actuated manipulator using model-based feedback linearisation. A mechanical model maps virtual acceleration commands into joint torques, while recorded data capture residual dynamics arising from imperfect model parameters. The controller minimises worst-case prediction error over an open Grassmannian ball centred at the fitted trajectory subspace. Simulations on a planar 2R manipulator under gravity compare the proposed method with double-integrator model predictive control (MPC) and direct regularised data-enabled predictive control (DeePC). Results show improved tracking over double-integrator MPC despite increased torque effort, while performance relative to DeePC depends on the reference trajectory.
Introduction
Predictive control chooses future inputs by predicting the system’s response, applies the first input, and repeats the optimisation after each new measurement. Model predictive control (MPC) predicts with a mathematical model. Data-driven predictive control (DDPC) predicts from recorded data instead, as in data-enabled predictive control (DeePC), which represents trajectories as linear combinations of recorded input–output data [4]. The Fundamental Lemma justifies this representation for controllable linear time-invariant (LTI) systems with noiseless, sufficiently exciting data [16].
Nonlinear systems with uncertain parameters challenge both: inaccurate mechanical models yield biased predictions, while linearly combining measured trajectories approximates nonlinear dynamics only locally. Model knowledge reduces the data burden. For fully actuated manipulators, feedback linearisation maps uses a mechanical model to map virtual acceleration commands to joint torques. With exact parameters and states, each joint becomes a double integrator from virtual input to position. Parameter, state-estimation and actuator errors produce residual dynamics learnable from virtual-input and position recordings.
Geometric robust DDPC [3] fits a low-dimensional trajectory subspace and handles fitting errors using an open chordal-distance ball of same-dimensional subspaces on the Grassmannian. It minimises worst-case weighted squared deviation from a target combining measured history and desired future trajectory. The mechanical model supplies the feedback-linearising torque map; DDPC predicts the virtual-input-to-position response. We test whether learning this response improves prediction and tracking beyond the known double-integrator model.
Related work. Regularised and robust formulations extend DeePC to noisy data [5, 9]. For nonlinear systems, Berberich et al. [2] obtain closed-loop stability guarantees by updating the data of a linear predictor online. For feedback-linearizable systems, Alsalti et al. [1] build a data-based predictor from input–output data and basis functions that approximate the unknown nonlinearities, with a fully actuated double inverted pendulum as example. Watson [13] incorporates partial model knowledge into DeePC by combining model-based and data-based representations. Tube MPC handles bounded disturbances around a model prediction [12], and learning-based MPC augments a model with learned corrections and uncertainty estimates [8 a].
Contributions. We apply geometric robust DDPC to trajectory tracking of a planar 2R manipulator under gravity. Prediction tests compare the learned feedback-linearised response with the ideal double integrator, for exact and inaccurate mechanical parameters. Transfer tests examine whether fixed parameter tunings remain effective when the measurement noise and arm parameters change. Tracking tests compare accuracy and torque effort with double-integrator MPC and with direct regularised DeePC, which predict from recorded joint torques and positions directly.
Preliminaries
Notation. Operator stacks its arguments vertically. For we write . We use for the Frobenius norm, for the Moore–Penrose pseudoinverse, and for the column space. and are the identity matrix and the all-ones vector, and is the Kronecker product. For , we write . Superscripts , and mark recorded data, initial histories and references. The measured output is , where is measurement noise. A record of length defines, for , the depth- block Hankel matrix
Each column is a window of consecutive samples. The record is persistently exciting of order if this matrix has full row rank, that is, .
Linear Systems and Model Predictive Control
Consider a controllable and observable discrete-time LTI system , , with , and . Its lag is the smallest positive integer for which the observability matrix has rank .
MPC predicts over a horizon of samples. At step , let and denote the th prediction, stacked as and over . With desired sequences , the input-output tracking cost is written as
where and . MPC minimises this cost subject to the dynamics and an initial state estimate, then applies the first input.
Data-Based Prediction
For , the length- behaviour is the set of trajectories that the system generates from arbitrary initial states and inputs. For , it is a subspace of dimension ([10]). Let be a noiseless trajectory of the system with persistently exciting input of order . Willems et al. [16]’s Fundamental Lemma states:
For prediction, choose a history length and a prediction horizon , and set . In every column of a Hankel matrix, the first samples form the past (superscript p) and the remaining samples form the future (superscript f). The input and output Hankel matrices are therefore and . With measured outputs, the output matrix becomes . All blocks have columns, with for a single record. The blocks and have and rows, and the output blocks have and rows.
Define the histories and , and let denote the measured output history. Under the noiseless spanning conditions, is a valid continuation of this history if and only if some satisfies
Since , the history determines the initial state, and the prediction is equivalent to model-based prediction with the same cost and constraints ([4]).
Regularised DeePC
Measurement noise perturbs both the recorded matrix and the online history, so (4) cannot be satisfied exactly. Regularised DeePC therefore allows a residual in the past-output equation and penalises the coefficient vector ([5]). Penalising pulls the prediction towards the zero trajectory, which is unsuitable for a nonzero reference. Following Elokda et al. [6], we centre the penalty at a coefficient vector that reproduces the reference. Define
The minimum-norm least-squares solution is recomputed when the reference or the data change. It reproduces exactly if , which does not imply physical feasibility.
The regularised problem is
where . Known inputs are matched exactly, and the penalised past-output residual is . The future predictions are and .
Geometric Robust DDPC
Geometric robust DDPC [3] accounts for uncertainty in the subspace that represents the trajectories. Let and . The leading left singular vectors of form an orthonormal basis of the fitted subspace . The dimension is motivated by the LTI behaviour dimension . Nonlinear effects and measurement noise raise the rank of the data matrix, becomes a fixed-rank approximation. The data may be centred and scaled before fitting (Appendix § 9.§ 1).
The Grassmannian is the set of -dimensional subspaces of . For an orthonormal basis of , the projector is . The principal angles between two subspaces satisfy , where is the th singular value. The chordal distance and the open ball of radius around are
At update , the target collects the measured past pairs for and the future reference pairs for . For a decision variable , the projection onto a candidate subspace is the corresponding predicted trajectory, with past and future inputs and outputs. The predictive problem is
with and . The past block of penalises deviations from the measured history, with controlling how strongly the prediction is tied to it, and the future block penalises tracking errors. The radius sets the subspace uncertainty. Figure 1 illustrates the projections.
Figure 1. Geometric uncertainty in . The fitted subspace is the horizontal line. The shaded double cone is the ball of lines with , (dashed boundary lines are excluded). Projecting onto and onto an adverse subspace gives different trajectories, the adverse with the larger residual to .
The objective in (8) is convex in as a supremum of convex functions, while the search for the adverse subspace is nonconvex. Because is continuous in , the supremum over the open ball equals the maximum over its closure. For the returned , the adverse subspace maximises over the ball. The controller decodes the trajectory and applies its first future input. Both inputs and outputs vary with the subspace. The ball also contains subspaces without a causal LTI realisation, so its use here is evaluated experimentally.
Planar 2R Manipulator Control
The mechanical model feedback-linearises the arm, and geometric DDPC predicts the resulting virtual-input–position response.
Manipulator Model and Tracking Task
The base angle is measured from the positive horizontal axis; is relative to link 1, so link 2 has absolute angle . Link has length , mass , centre-of-mass distance from its proximal joint, and inertia about its centre of mass. Figure 2 shows this convention and the centre-of-mass positions . For uniform rods, and .
Figure 2. Planar 2R Manipulator.
Let , , and . The joint dynamics are
where is the inertia matrix, contains Coriolis and centrifugal terms, is the gravity torque, is diagonal viscous damping, and is the applied torque. The expressions for and are given in Appendix A. For gravitational acceleration of magnitude in the negative vertical direction,
Positions are measured at as , with independent . Velocity is estimated. The task is to track , assessed by position error and applied torque. Predictive problems impose no joint or torque constraints.
Feedback Linearisation and State Estimation
Write . The supplied model uses and , and its parameters may differ from the plant. For a held virtual acceleration , the inverse dynamics map gives
With exact parameters, state feedback, continuous torque evaluation and inactive actuator limits, substitution gives . The feedback-linearised system then consists of two double integrators, with exact sampled model , , where
The ideal model is controllable, has observability lag two, and has for . A Kalman filter estimates position and velocity from measured positions and commanded virtual inputs [15], using (12). Figure 3 shows the loop; Appendix A gives the filter.
Figure 3. Geometric DDPC with feedback linearisation. The observer uses the current position measurement. The predictor uses completed pairs through .
The difference between the actual and continuously evaluated model torque is . This includes torque holding and actuator limitations. Consequently, , with
With exact state feedback and unrestricted continuous torque application, the dynamics become , where and . The recorded response contains these model errors together with estimation and actuator effects, giving the learned predictor information beyond the ideal double integrator.
Offline Data and Reference Initialisation
A PD tracking controller with acceleration feedforward and random excitation generates offline data. Its sinusoidal reference varies in centre, amplitude, phase and period between episodes (Appendix B).
For the feedback-linearised arm, is held for , while inverse dynamics is evaluated from the propagated estimate every . For direct data, is computed once at and held for . The actuator enforces torque magnitude and slew limits at each update in both cases. The mechanical model and state estimates are used to generate both offline datasets. Windows are concatenated without crossing episode boundaries.
Tracking references are joint sinusoids with a smooth entry ramp. Sampled , , satisfy the ideal dynamics through a piecewise-constant-acceleration interpolant. Torque and torque-rate requirements are checked throughout the motion (Appendix B).
Geometric DDPC for the Feedback-Linearised Arm
A rank- fit to records of defines . This dimension follows the ideal four-state model; the affine fitting coordinates are specified in Appendix § 9.§ 1.
At update , the target contains the preceding measured pairs and the next reference pairs , expressed in the predictor’s coordinates. The controller solves (8) over the open ball in , with fixed and . A scenario-exchange algorithm returns an approximate adverse trajectory, whose first future virtual input is applied through (11). Appendix § 9.§ 3 gives the solve and stopping rules.
Comparison Controllers
Double-integrator MPC. This controller predicts with (12), using the same horizon, observer, feedback-linearising torque map and actuator as geometric DDPC. It minimises (2) with and , using the same future tracking weights as geometric DDPC in physical coordinates. Its first virtual input is applied through the shared torque map.
Direct regularised DeePC. Direct DeePC uses the recorded torque and position pairs to predict future motion [6, 14]. In Figure 4, the desired motion defines the torque reference . Set and in both the tracking cost and (5). At each update, the measured history initialises (6), whose first predicted torque is applied directly. The model supplies references and collection torques; online feedback uses measured input–output history. Appendix § 9.§ 5 gives the quadratic solve.

Figure 4. Direct regularised DeePC. The desired torque and position sequences define the reference, while completed torque–position pairs update the history and online data bank.
Simulation Studies
We examine prediction accuracy (A), transfer of predictive tuning (B), and cyclic tracking (C). Controllers receive noisy positions; true positions are used only to evaluate errors.
Simulation Setup and Evaluation
Table 1 defines the mechanical conditions. Experiments A and C retain a fixed model while changing the plant parameters. Experiment B uses exact models of two known arms. In all cases, gravity acts downward and position noise has covariance during collection and control.
| Condition | Studies | Lengths [m] | True masses [kg] | Model masses [kg] |
| Exact parameters | A, C | (1.0, 0.8) | (0.9, 0.72) | (0.9, 0.72) |
| Parameter errors | A, C | (1.0, 0.8) | (1.0, 0.8) | (0.9, 0.72) |
| Arm A | B | (1.0, 0.8) | (1.0, 0.8) | (1.0, 0.8) |
| Arm B | B | (0.75, 0.55) | (1.3, 0.6) | (1.3, 0.6) |
Table 1. Mechanical conditions. Model lengths equal true lengths. True damping is (0.04, 0.02) N m s rad in the parameter-error condition and zero elsewhere; model damping is zero throughout.
Controller settings are fixed before evaluation. Direct-DeePC weights are selected on separate Arm A trials at baseline noise. The data-driven methods share the collection procedure and budget, with their respective input conventions. Trials use paired measurement-noise sequences across controllers, and predictive computation is instantaneous. Each run covers a smooth entry ramp and two reference cycles. Appendix § 10 gives the settings and evaluation protocol.
Tables report median [minimum, maximum] periodic position RMSE , median whole-run applied-torque RMS , and completed trials . Both measures average over the two joints. Incomplete trials enter completion counts but are excluded from full-duration error and effort statistics.
Prediction Accuracy After Feedback Linearisation
Experiment A compares the double integrator with the predictor fitted to virtual-input–position data. On independent test trajectories, both predict future positions from the same prescribed inputs and past measured positions ([11]. Future positions enter only the error calculation. Table 2 covers exact and inaccurate mechanics at three noise levels.
| Mechanics | [rad] | Double integrator | Learned behaviour |
| Exact parameters | 0.001 | 0.0098 [0.0068, 0.0110] | 0.0058 [0.0051, 0.0070] |
| 0.005 | 0.0157 [0.0118, 0.0191] | 0.0129 [0.0094, 0.0147] | |
| 0.010 | 0.0251 [0.0213, 0.0279] | 0.0200 [0.0191, 0.0242] | |
| Parameter errors | 0.001 | 0.2096 [0.1831, 0.2496] | 0.0153 [0.0131, 0.0194] |
| 0.005 | 0.1801 [0.1510, 0.2207] | 0.0167 [0.0104, 0.0183] | |
| 0.010 | 0.1598 [0.1292, 0.1916] | 0.0234 [0.0187, 0.0270] |
Table 2. Position prediction RMSE [rad]. Entries are medians [minimum, maximum] across independently fitted models, after taking the median error over each model’s test windows.
Learning gives a modest improvement with exact parameters and a much larger improvement with parameter errors, indicating that data capture systematic departures from the ideal dynamics. The learned predictor remains more accurate at each noise level, although its error grows with noise. Noise also changes the feedback-controlled collection trajectories, so comparisons across noise levels include this effect.
Transfer Across Noise and Arm Changes
Experiment B uses the base cycle with unchanged predictive tuning. Noise is increased on Arm A, then a known Arm B with shorter links and different masses is tested. Each condition uses newly collected data and the stated observer noise covariance; the mechanical model is updated for Arm B. The collection procedure and predictive settings stay fixed.
Tracking error increases with noise for both methods (Table 3). Geometric DDPC retains lower median error and completes every trial. At high noise, its torque variation and actuator activity increase (Figure 5); direct DeePC has a wider error spread and one incomplete trial. On Arm B, both methods retain accuracy close to their Arm A baselines with reduced torque effort. Geometric DDPC remains more accurate at similar effort.
| Condition | Controller | |||
| Arm A, | Geometric DDPC | 0.0042 [0.0038, 0.0054] | 2.984 | 6/6 |
| Direct DeePC | 0.0245 [0.0236, 0.0266] | 2.961 | 6/6 | |
| Arm A, | Geometric DDPC | 0.0138 [0.0116, 0.0164] | 2.977 | 6/6 |
| Direct DeePC | 0.0370 [0.0285, 0.0545] | 2.955 | 6/6 | |
| Arm A, | Geometric DDPC | 0.0280 [0.0210, 0.0345] | 3.560 | 6/6 |
| Direct DeePC | 0.0662 [0.0369, 0.2517] | 2.949 | 5/6 | |
| Arm B, | Geometric DDPC | 0.0048 [0.0039, 0.0055] | 1.957 | 6/6 |
| Direct DeePC | 0.0257 [0.0227, 0.0323] | 1.948 | 6/6 |
Table 3. Transfer with fixed predictive settings. Errors are in radians and torque RMS in N m; gives completed/attempted trials.
Figure 5. Noise transfer with geometric DDPC on Arm A: second-joint position and applied first-joint torque in the first trial, after the ramp. Legend values give in radians.
Trajectory Tracking
Experiment C compares geometric DDPC, double-integrator MPC and direct regularised DeePC. The base reference is tested with exact mechanical parameters; all three references in Table 4 are tested with parameter errors. The mechanical model stays fixed, and offline data are reused across references within each plant condition. Sampled reference torques satisfy the actuator limits.
| Reference | Centre [rad] | Amplitudes [rad] | Period [s] |
| Base | 6 | ||
| Shifted | 6 | ||
| Faster | 4.5 |
Table 4. Joint references, with phases and a two-second entry ramp.
Table 5 summarises tracking accuracy and effort. Figure 6 shows the base-cycle trajectories with parameter errors.
| Condition, cycle | Controller | |||
| Exact parameters, base | Double-integrator MPC | 0.0050 [0.0046, 0.0053] | 2.703 | 6/6 |
| Geometric DDPC | 0.0050 [0.0047, 0.0064] | 2.690 | 6/6 | |
| Direct DeePC | 0.0253 [0.0213, 0.0292] | 2.665 | 6/6 | |
| Parameter errors, base | Double-integrator MPC | 0.1091 [0.1090, 0.1093] | 2.810 | 6/6 |
| Geometric DDPC | 0.0650 [0.0482, 0.0749] | 3.030 | 6/6 | |
| Direct DeePC | 0.0826 [0.0636, 0.0848] | 2.530 | 6/6 | |
| Parameter errors, shifted | Double-integrator MPC | 0.1152 [0.1150, 0.1154] | 3.456 | 6/6 |
| Geometric DDPC | 0.0592 [0.0412, 0.0744] | 3.681 | 6/6 | |
| Direct DeePC | 0.1365 [0.1089, 0.1411] | 3.043 | 6/6 | |
| Parameter errors, faster | Double-integrator MPC | 0.1068 [0.1068, 0.1069] | 2.209 | 6/6 |
| Geometric DDPC | 0.0779 [0.0648, 0.0866] | 2.381 | 6/6 | |
| Direct DeePC | 0.0731 [0.0594, 0.0771] | 1.977 | 6/6 |
Table 5. Cyclic tracking at rad. Errors are in radians and torque RMS in N m; gives completed/attempted trials.
Figure 6. Base-cycle tracking with mass and damping errors. Both joint positions are shown for the first trial of each controller, together with the reference.
With exact mechanics, double-integrator MPC and geometric DDPC give similar accuracy and effort, both outperforming direct DeePC in tracking. With parameter errors, geometric DDPC improves on double-integrator MPC in every paired trial at greater effort. It also has lower median error than direct DeePC on the base and shifted cycles, while direct DeePC has lower error and effort on the faster cycle. All tracking trials complete with inactive actuator limits.
Discussion and Conclusions
This study examined geometric robust DDPC for a feedback-linearised planar 2R arm. The results identify a useful role for data when the mechanical model captures the main structure but leaves systematic prediction errors. Learning the resulting virtual-input–position response improves prediction and tracking under parameter errors. With exact mechanical parameters, the known double integrator already provides an effective controller. Performance relative to direct regularised DeePC depends on the reference, with different trade-offs between accuracy and torque effort.
The predictive settings also remained effective across the tested noise levels and arm parameters. This transfer used newly collected data and the corresponding mechanical-model and observer updates. Higher measurement noise increased tracking error and actuator activity, although every geometric-DDPC trial completed.
Generative AI Use
Generative AI tools assisted with literature review, preparation of TikZ illustrations, simulation-code optimisation, and auditing the mathematical exposition/results.
Mechanical Model, State Estimation and Torque Application
Inertia and velocity coupling. For the coordinates and link parameters in § 3, the inertia and velocity-coupling terms are
Here , , and .
The estimator uses the double-integrator matrices , , in (12) and measured joint positions.
Position-based state estimation. A Kalman filter on the ideal double-integrator model estimates position and velocity [15]. Let and be the prior and corrected estimates, with covariances and . The current measurement gives
After selecting , propagate and , where and is a chosen acceleration-error covariance. The covariance models uncertainty in the observer prediction; the plant has no added process noise.
Between measurements, inverse dynamics uses for , with , obtained by replacing in (12) by . The initial estimate uses the reset position and zero velocity. Commanded virtual inputs drive the filter during collection, startup and tracking.
Torque application. Let divide . At , for , compute . Given the preceding applied torque , positive magnitude bounds , and slew bounds , apply
Here truncates each component of to its interval . The torque is held until the next update. Torque holding and actuator modification enter in (13).
Data Collection and Reference Construction
Data collection. A PD tracking controller with acceleration feedforward follows an auxiliary sinusoid and adds random virtual excitation:
The diagonal gains are fixed. The sinusoid varies between episodes, and is independent, zero-mean Gaussian excitation. The complete command also depends on measured feedback.
For the feedback-linearised dataset, the recorded command is . For the direct dataset, it is . Both use the supplied model and estimated state, with the input-holding conventions of § 3.3.
Position is sampled before the new command acts. The recorded pair is , where for the feedback-linearised system and for direct DeePC. Inputs are stored before actuator modification. An episode of samples supplies columns, giving offline windows. All windows remain within their episodes.
Reference construction. The evaluation references are
with centre , amplitudes , phases , angular frequency and ramp duration . The ramp for , extended by zero before and one after this interval, gives continuous position, velocity and acceleration.
The desired position samples are . A virtual-input reference consistent with the ideal sampled dynamics is obtained from and
This defines a piecewise-constant-acceleration interpolant through the desired positions. Its discrete velocity can differ from the analytic reference derivative. Torque and torque-rate demands of this interpolant are checked through the ramp and periodic motion.
Initial history and sample timing. Before predictive tracking, the arm is held near the initial reference for sampling intervals, using the collection controller with a constant reference and zero added excitation. The final plant and observer states of this hold initialise the tracking run.
At update , the current measurement corrects the observer while the predictor uses completed pairs through . The selected command is applied immediately in simulated time. After the interval, replaces the oldest history pair.
For the direct coefficient reference, extend the desired history before time zero by , , and for . At nonnegative samples, use the torque reference in § 3.5 with the sampled velocity and acceleration in (19). The measured history enters the predictive consistency terms separately.
Geometric Representation and Predictive Solves
Affine, Dimensionless Trajectory Coordinates
Let be the concatenated Hankel matrix of measured pairs , ordered by sample. Fix positive scales for virtual input and position, and set . Define the mean window , scaled centred matrix , and fitted basis by
Here returns the leading left singular vectors. Centring limits the data rank to at most , so the fit requires and an adequately excited th direction. This rank follows the ideal four-state model and approximates the implemented response.
Each candidate describes the physical affine set . The mean, scales and fitted subspace remain fixed during a run. Scaling sets both the relative input/output weights and the geometry used to compare subspaces.
Let and . Stack the measured past and desired future, then transform them into the fitting coordinates:
For the returned trajectory parameter and adverse subspace, decode
The selector extracts the first future virtual input. Physical future tracking weights are the inverse squared channel scales.
Prediction for Prescribed Inputs
For the prediction tests, write a fitted trajectory as , with . Let select all input entries and the past output entries. Given the complete prescribed input stack and past measured positions , compute
The future output entries are the forecast. For the known double integrator, use its exact behaviour and zero affine mean in the same solve. Future positions enter only the error calculation.
Sine Chart and Scenario-Exchange Optimisation
The scenario-exchange method alternates a convex master solve with an adverse-subspace search. The master minimises the largest cost over its current subspaces. The adverse search adds a subspace with a larger cost at the returned . The master uses the bound , which contains the sublevel set whose worst-case cost is no greater than the cost at .
If is an orthonormal basis of the complement of , the sine chart is
For , the positive-definite square root exists and the resulting basis is orthonormal. Its chordal distance from the fitted subspace is . Multistart local searches operate on this closed chart ball. Its extremal value equals the open-ball supremum; the selected chart point is restricted to before command extraction, keeping the executed projection strictly inside the open ball.
Warm starts and an invertible preconditioning of the master improve its numerical solution without changing the objective. Let be the numerical upper estimate of the worst-case cost at the returned parameter, the master-based lower estimate of the optimum, and the cost at the returned adverse projection. The recorded relative gaps are
These floating-point estimates assess the numerical solve. A finite candidate can be applied at the iteration limit with an open gap, and its termination status is retained in the results. Table 6 lists the numerical settings.
| Numerical setting | Value |
| Exchange iteration limit | 20 |
| Adverse search | Three starts, up to 80 iterations each |
| Master / spectral-search limits | 500 / 80 iterations |
| Relative outer / adverse gap targets | / |
| Inner / master tolerances | / |
| Extracted chart radius | At most |
Table 6. Numerical settings for geometric DDPC. Warm starts are enabled.
Across the 6540 geometric updates, 404 meet both gap targets and 6136 reach the exchange-iteration limit. The median relative outer gap is 0.453, with maximum 0.765; the median adverse-search gap is , with maximum 0.491. These bound separations leave the suboptimality of the executed decisions unresolved. Every executed subspace lies strictly inside the radius-0.01 ball.
Double-Integrator MPC
Let . Define by its block: for $0 \le j < i < N$, and zero otherwise. The predicted output is . With and , the minimiser satisfies
The positive-definite coefficient matrix is factored once; the reference and observer prior update the right-hand side.
Reference-Centred Direct DeePC
Let be the current direct data matrix, including online windows. The reference centre is , with defined in (5).
For a wide matrix, take a thin QR factorisation , with , and set . Then . Decompose , where . Since belongs to the row space of , it can be written as , with . Hence
The omitted component changes no prediction and is zero at the optimum. Thus the reduction is exact for the reference-centred squared penalty. When compression is unnecessary, use and . The pseudoinverse is evaluated by a singular-value decomposition with relative cutoff .
Partition and set , . Let . The condensed quadratic has coefficient matrix and linear term
The stationarity and equality equations are
where is a scaled equality multiplier. Cholesky factors and a Schur complement solve this system. Scaled equality and stationarity residuals are checked before accepting the command. Data updates change the factors; reference and history updates change the linear term.
Simulation Parameters and Evaluation
Fixed Settings and Data Collection
All links are uniform rods with . The plant and model parameters are given in Table 1. Table 7 lists the shared controller and simulation settings. The observer uses , , and diagonal initial covariance obtained from the listed standard deviations.
| Setting | Value |
| Predictive / torque-update periods | , |
| History, horizon and rank | , , ; in |
| Offline data per method and trial | Eight episodes of 60 samples; 392 windows |
| Collection gains and excitation | , ; excitation standard deviation 0.7 per joint |
| Baseline position noise | rad per joint |
| Observer uncertainty | ; initial position and velocity standard deviations 0.01 rad and 0.1 |
| Geometric settings | , ; position scale 0.1 rad and virtual-input scale 2 per joint |
| Direct DeePC settings | , , , ; centre |
| Torque magnitude / slew limits | N m / N m |
| Plant integration | Fourth-order Runge–Kutta, 0.005 s steps |
Table 7. Shared simulation settings. The transfer study changes the specified arm and sensor-noise covariance while retaining predictive tuning and the collection procedure.
The auxiliary sinusoids have uniformly sampled centres in rad, amplitudes in and rad, periods in s, and phases in . Episodes start at zero velocity with an independent 0.02 rad Gaussian position perturbation. Direct DeePC retains all offline columns and up to 392 completed online windows.
Direct-DeePC weights were selected on four separate Arm A trials with the base reference and baseline noise, favouring completed runs with low cyclic error. The selected weights, torque reference and reference-centred penalty are then used in every evaluation condition. No condition-specific retuning is performed.
References and Performance Measures
The references in Table 4 have phases and a two-second ramp. Evaluation follows a 1.2-second stabilising hold and lasts for the ramp and two cycles: 14 seconds for the base and shifted references, and 11 seconds for the faster reference. Each condition uses six independent data/noise pairs, with measurement noise paired across controllers.
For a completed run of intervals, define and the periodic indices . The reported measures are
The factor two averages over the joints. Torque effort includes the actual held torque after actuator modification. A trial stops when either position error exceeds 2 rad, either speed exceeds , or a numerical failure occurs. The terminal state is checked as well. Incomplete runs enter completion counts but not full-duration error or effort statistics.
Prediction Tests and Numerical Accuracy
For each test window, the error is , where the vectors stack future positions. Each independently fitted predictor uses three separate 60-sample test episodes, giving 15 non-overlapping windows. Take the median over windows for each fit, then the median and range across six fits. Data are recollected for each noise/mechanics condition, and both predictors solve (22) with identical inputs and past measurements.
The dimension is retained throughout. Supporting fits with and 32 gave larger prediction errors and poorer conditioning on the same test records. A numerical-integration check at half the step size changed saved-interval positions by at most , with unchanged actuator paths. Optimisation settings and termination statistics are reported in Appendix § 9.§ 3.
References
- [1]Mohammad Alsalti, Victor G. Lopez, Julian Berberich, Frank Allgöwer, and Matthias A. Müller. Data-driven nonlinear predictive control for feedback linearizable systems. IFAC-PapersOnLine, 56(2):617–624, 2023. doi: 10.1016/j.ifacol.2023.10.1636.
- [2]Julian Berberich, Johannes Köhler, Matthias A. Müller, and Frank Allgöwer. Linear tracking MPC for nonlinear systems—part II: The data-driven case. IEEE Transactions on Automatic Control, 67(9):4406–4421, 2022. doi: 10.1109/TAC.2022.3166851.DOI
- [3]Shreyas Bharadwaj, Bamdev Mishra, Cyrus Mostajeran, Alberto Padoan, Jeremy Coulson, and Ravi N. Banavar. Robust least-squares optimization for data-driven predictive control: A geometric approach. In Gaurav Sukhatme, Lars Lindemann, Stephen Tu, Adam Wierman, and Nikolay Atanasov, editors, Proceedings of The 8th Annual Learning for Dynamics and Control Conference, volume 331 of Proceedings of Machine Learning Research, pages 511–531. PMLR, 2026. URL https://proceedings.mlr.press/v331/bharadwaj26a.html.arxiv.org/abs/2511.09242
- [4]Jeremy Coulson, John Lygeros, and Florian Dörfler. Data-enabled predictive control: In the shallows of the DeePC. In 2019 18th European Control Conference (ECC), pages 307–312, 2019a. doi: 10.23919/ECC.2019.8795639.DOI
- [5]Jeremy Coulson, John Lygeros, and Florian Dörfler. Regularized and distributionally robust data-enabled predictive control. In 2019 IEEE 58th Conference on Decision and Control (CDC), pages 2696–2701, 2019b. doi: 10.1109/CDC40024.2019.9028943.DOI
- [6]Ezzat Elokda, Jeremy Coulson, Paul N. Beuchat, John Lygeros, and Florian Dörfler. Data-enabled predictive control for quadcopters. International Journal of Robust and Nonlinear Control, 31(18):8916–8936, 2021. doi: 10.1002/rnc.5686.DOI
- [8]Lukas Hewing, Kim P. Wabersich, Marcel Menner, and Melanie N. Zeilinger. Learning-based model predictive control: Toward safe learning in control. Annual Review of Control, Robotics, and Autonomous Systems, 3:269–296, 2020b. doi: 10.1146/annurev-control-090419-075625.DOI
- [9]Linbin Huang, Jianzhe Zhen, John Lygeros, and Florian Dörfler. Robust data-enabled predictive control: Tractable formulations and performance guarantees. IEEE Transactions on Automatic Control, 68(5):3163–3170, 2023. doi: 10.1109/TAC.2023.3241282.DOI
- [10]Ivan Markovsky and Florian Dörfler. Behavioral systems theory in data-driven analysis, signal processing, and control. Annual Reviews in Control, 52:42–64, 2021. doi: 10.1016/j.arcontrol.2021.09.005.DOI
- [11]Ivan Markovsky and Paolo Rapisarda. Data-driven simulation and control. International Journal of Control, 81(12):1946–1959, 2008. doi: 10.1080/00207170801942170.DOI
- [12]D. Q. Mayne, M. M. Seron, and S. V. Raković. Robust model predictive control of constrained linear systems with bounded disturbances. Automatica, 41(2):219–224, 2005. doi: 10.1016/j.automatica.2004.08.019.DOI
- [13]Jeremy D. Watson. Hybrid data-enabled predictive control: Incorporating model knowledge into the DeePC. IEEE Open Journal of Control Systems, 4:549–561, 2025. doi: 10.1109/IOCSYS.2025.3625364.arxiv.org/abs/2502.12467
- [14]Felix Wegner. Data-enabled predictive control of robotic systems. Master’s thesis, ETH Zürich, April 2021.DOI
- [15]Greg Welch and Gary Bishop. An introduction to the Kalman filter. Technical report TR 95-041, University of North Carolina at Chapel Hill, November 2000. URL https://www.cs.yale.edu/homes/hudak-paul/CS474S01/kalman.pdf. Revised edition, November 13, 2000.
- [16]Jan C. Willems, Paolo Rapisarda, Ivan Markovsky, and Bart L. M. De Moor. A note on persistency of excitation. Systems & Control Letters, 54(4):325–329, 2005. doi: 10.1016/j.sysconle.2004.09.003.DOI




