Pith. sign in

REVIEW 3 major objections 4 minor 47 references

Robust Model Predictive Control Design for Autonomous Vehicles with Perception-based Observers

T0 review · 3 major / 4 minor · reviewed 2026-08-05 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read The paper argues that vision-based perception noise is biased and heavy-tailed, so robust MPC should bound it with zonotopes instead of assuming Gaussian noise — and shows the design running on a real robot.

desk verdict Useful integration of zonotope estimation with LP-MPC, but the stability guarantee doesn't cover the implemented terminal set—fixable, but load-bearing. read the letter →

arxiv 2509.05201 v1 pith:WPUJWRDR submitted 2025-09-05 cs.RO cs.CVcs.LGcs.SYeess.SY

Nariman Niknejad , Gokul S. Sankar , Bahare Kiumarsi , Hamidreza Modares This is my paper
classification cs.ROcs.CVcs.LGcs.SYeess.SY
keywords Perceptionstateestimationuncertaintyzonotopeslinearprogrammingmodelpredictivecontrolheavy-tailednoiseobserverdesign
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

Perception modules built on neural networks do not produce zero-mean Gaussian errors; in practice their errors are biased, heavy-tailed, and worse far from the training distribution. The paper's claim is that robust control should treat those errors as unknown-but-bounded sets, and it constructs a complete control loop on that principle: a zonotopic observer bounds the estimation error, a tube-based MPC keeps state and input inside tightened constraints, and a Minkowski-function cost makes the online optimization a linear program. The stability argument runs through Minkowski–Lyapunov inequalities and contractive zonotopic invariant sets, with a maximal-volume terminal ellipsoid (replaced by an inscribed zonotope in the implementation) providing recursive feasibility. Simulations and a hardware experiment with a CNN-based camera perception module on an omnidirectional robot support the claim: the perception-aware MPC keeps estimates inside their bound sets and achieves lower cumulative cost than a Gaussian Kalman/LQR baseline under heavy-tailed noise. If the claim holds, it matters because it removes the Gaussian assumption from a safety-critical component of autonomous vehicles without giving up real-time computation.

What carries the argument

The load-bearing object is the constrained zonotope — a polytope written as $\{c+G\alpha \mid |\alpha|\le 1, F\alpha=\theta\}$ — used for noise bounds, invariant sets, and cost sets. The argument moves through the Minkowski function $g(S,x)$, the smallest $\lambda \ge 0$ with $x \in \lambda S$, which becomes the stage and terminal cost; because containment of one constrained zonotope in another can be written as a linear program (Lemma 1), four design steps remain convex: making the error set and the deviation set $\lambda$-contractive, enforcing the Minkowski–Lyapunov terminal inequality via containment of polar sets, and computing the maximal terminal ellipsoid. The final conversion from that ellipsoid to an inscribe

What would settle it

Recompute the inscribed terminal zonotope $Z_f$ for the paper's hardware design and test $(A+BK_f)Z_f \subseteq Z_f$ and the inclusions $Z_f \subseteq S_{\bar{x}}$, $K_f Z_f \subseteq S_{\bar{u}}$; if any vertex maps outside, recursive feasibility is not certified. Alternatively, run the closed loop under a fixed bias plus Laplace noise and check whether the observed estimation error ever leaves the predicted zonotopic bound — Theorem 1 says it cannot.

Watch

Extended reading notes

Core claim

The central claim is that for a perception-based observer, the correct uncertainty model is a bounded set rather than a Gaussian distribution. The paper models process noise and perception noise as constrained zonotopes, then designs an observer gain L so that the estimation-error set propagates inside a $\lambda_L$-contracted version of itself, and a feedback gain K so the state-deviation set does the same. This gives the tube: the true state is always contained in the nominal state plus the sum of the two invariant sets, so tightening the original state and input constraints by that sum yields robust constraint satisfaction. The MPC minimizes a Minkowski-gauge cost (state and input gauges joi

Load-bearing premise

The stability argument assumes the zonotope used in the algorithm in place of the terminal ellipsoid still satisfies the invariance and constraint-inclusion properties proven only for the ellipsoid; the paper does not prove that inheritance.

Editorial extensions

If this is right

  • Robust constraint satisfaction no longer depends on knowing a Gaussian covariance for perception noise; a zonotopic bound around the learned map is sufficient.
  • The online MPC can be solved as a linear program, which is fast enough for onboard execution on small robots with a ROS2 stack.
  • The design exposes explicit tuning parameters — contraction ratios lambda_L, lambda_tilde{x}, and lambda_f — that trade convergence speed against the size of the invariant tubes.
  • For any perception module whose error is bounded near its training distribution, the same pipeline gives bounded state-estimation error and constraint satisfaction regardless of the noise's mean, color, or tail weight.
  • The hardware comparison indicates that the advantage over Gaussian designs grows when the noise is biased, because a Kalman-type estimator cannot remove the bias while the zonotopic observer encloses it.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • A direct invariance proof for the inscribed zonotope terminal set, rather than the ellipsoid it approximates, would close the gap between the stability theorem and the implemented LP; as written, the certificate applies to the ellipsoid.
  • The zonotopic observer design is modular: it should transfer to other perception modalities (LiDAR, radar, visual odometry) as long as the error admits a bound around the operating region, which the paper motivates but does not test.
  • One stress test the paper leaves implicit is varying the bias-to-variance ratio of the perception error; the zonotopic design should dominate at high bias, while a tuned Gaussian design may remain competitive at low bias.
  • The infinity/infinity slack cost could be reused in other set-based MPC settings as a cheap way to avoid the idle/deadbeat artifacts of 1-norm costs, independent of the perception application.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 4 minor

Summary. The paper proposes a robust tube-based MPC framework for linear systems in which the state estimate is provided by a perception module (CNN) modeled as a noisy sensor with non-Gaussian, possibly biased noise. The uncertainty is represented by constrained zonotopes, and the observer gain and state-feedback gain are designed so that the estimation error and the deviation between observer and nominal state stay in contractive zonotopic sets (Theorems 1 and 2). The terminal cost is based on Minkowski functions of polar sets (Theorem 3), and the terminal set is initially computed as a maximum-volume ellipsoid satisfying invariance and tightened constraint inclusions (Theorem 4). To keep the MPC a linear program, the terminal set is then replaced by an inscribed zonotope (Remark 5, Algorithm 1), and the stage cost is replaced by an infinity/one-norm slack formulation (Remark 6). The method is validated in simulation and on a ROSbot XL with a custom CNN perception module, comparing favorably with a Gaussian-Kalman-based MPC.

Significance. If the gaps identified below are addressed, the paper would be a useful contribution to perception-based control: it tackles a practically important issue (non-Gaussian, biased perception noise), builds a coherent set-membership observer-plus-MPC pipeline, and includes real-hardware validation. Theorems 1-3 are standard containment derivations and appear essentially correct under the stated assumptions. The main value is the integration of constrained zonotopes, Minkowski-Lyapunov terminal costs, and an LP formulation that is fast enough for onboard use. However, the gap between the terminal set for which invariance is proved (an ellipsoid) and the terminal set actually implemented (a zonotope) directly affects the central closed-loop stability claim; until that gap is closed, the theoretical guarantees do not cover the implemented algorithm.

major comments (3)
  1. [§V, Remark 5 and Algorithm 1, Step 5; §VI.B.3] The implemented terminal constraint (15g) uses a zonotopic inner approximation Z_f of the ellipsoidal terminal set S_f from Theorem 4. Theorem 4 proves invariance and constraint inclusions only for S_f: (A+BK_f)S_f⊆S_f, S_f⊆S_bar{x}, and K_fS_f⊆S_bar{u}. Invariance of S_f does not imply invariance of an inner subset Z_f. In particular, the standard recursive-feasibility argument requires that the shifted terminal state, propagated under A+BK_f, remain in Z_f; no such property is proved for the inscribed zonotope. With λ_f=0.95 in Section VI.B.3 and n=3, an inner zonotope that is also inside S_f cannot generally contain λ_fS_f, so the usual sufficient condition for invariance of a subset is not available. The paper should prove invariance and the inclusion conditions for Z_f (e.g., by computing a contractive zonotopic terminal set directly), implement the ellipsoidal terminal set and solv
  2. [§V, Remark 6 and Eq. (33)] The implemented LP replaces the stage cost in (15a) by a single slack variable β subject to (33b)-(33c). The paper asserts that the invariance and stability results of Theorem 3 remain valid because β is an upper bound on the cost, but no proof is given for the modified problem. The value function of the modified ∞/∞-cost problem is not immediately non-increasing; one must construct a candidate β_{k+1} from the optimal β_k, remove the first stage cost, add the terminal stage cost, and use the terminal inequality (22a). This argument is missing. Since Algorithm 1 solves the modified problem, the stability guarantee for the actual controller is incomplete.
  3. [§VI.B.2, Eq. (34)] The measurement-noise zonotope Z_v is computed using 95% of the perception-noise data. Assumption 6 requires v(k)∈Z_v for all k. The remaining 5% of test points lie outside Z_v, so the bounded-noise hypotheses of Theorems 1 and 2 are not satisfied in the experiments. Consequently, the experimental results do not demonstrate the certified bounded-error or closed-loop stability properties for the actual noise distribution. The authors should either enlarge Z_v to contain all observed errors, provide a statistical containment argument compatible with the theory, or explicitly characterize the hardware experiments as heuristic validation outside the assumptions.
minor comments (4)
  1. [Theorem 4, Eq. (25c)-(25d)] The notation diag(q_x) and diag(q_u) is dimensionally inconsistent with the proof: the proof requires ∥P^{1/2}q_{x,i}∥_2 ≤ q_{x,i}, i.e. the diagonal entries should be the squares of the scalar bounds. Please clarify the notation.
  2. [Problem 1, Eq. (15h)] Constraint (15h) fixes the initial nominal state to x0, but online only the estimate ˆx(k) is available. The text later uses ¯x(k|k) as a decision variable. Please clarify how (15h) is initialized in Algorithm 1 and how the estimate enters the optimization.
  3. [Algorithm 1, Step 5] Step 5 says 'Find the maximum size terminal set S_f using Theorem 4 and Remark 5', but Remark 5 replaces the ellipsoid with a zonotopic approximation. This wording is misleading; the set is no longer the maximum-volume ellipsoid and may not be maximal in any stated sense.
  4. [Lemma 1, Section II.C] Typo: 'constrianed zonotpes' should be 'constrained zonotopes'.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the derivation is self-contained; the zonotope terminal-set gap is a correctness issue, not circular reasoning.

full rationale

The paper's derivation chain is not circular. Theorem 1 and Theorem 2 formulate observer-error and state-deviation invariance as set-containment linear programs with proofs from the error/deviation dynamics and zonotope containment conditions. Theorem 3 converts the terminal Lyapunov inequality into a polar-set containment condition and proves sufficiency of (23) for (22a). Theorem 4 constructs a maximum-volume ellipsoidal terminal set under LMIs that directly enforce invariance and inclusion; the proof derives the implications (25b)-(25d). None of these theorems assumes the conclusion. The noise zonotope Z_v in (34) is an input fitted to perception data, but it is not relabeled as a prediction; the Gaussian-MPC comparison is an experimental benchmark, not a consequence of the fitted set. There are no self-citations: none of the references is authored by the present authors. Remark 5 replaces the ellipsoidal terminal set with a zonotopic approximation via [41] without proving that the zonotope inherits the invariance or inclusion properties of Theorem 4. This is a genuine correctness gap affecting recursive feasibility of Algorithm 1, but it is not circular: the missing property is neither assumed nor derived from itself; it is simply unproven. Likewise, fitting Z_v to 95% of the data bears on the validity of Assumption 6, not on circularity.

Assumptions & free parameters 6 free parameters · 11 assumptions · 0 invented entities

The central result rests on a set of domain assumptions (bounded state/input/noise sets represented as constrained zonotopes, and perception error bounded by a known zonotope) plus standard results from set-based control (zonotope containment, Minkowski/polar properties). The perception noise set is fitted to experimental data, and several design sets and contraction ratios are hand-chosen. No new physical or mathematical entities are introduced.

free parameters (6)
  • Perception noise zonotope Z_v (center c_v and generator matrix G_v) = c_v=[0.0011, -0.0051, 0]^T, G_v=diag(0.0490, 0.0667, 0)
    Fitted to 95% of the test-set perception errors (Sec VI.B.2, Eq. 34). The robust guarantees require 100% containment, so this is a fitted, not a priori, bound.
  • Initial estimation error set Z_e = Z_e = ⟨[0,0,0], diag(0.1,0.1,0)⟩
    Chosen by hand as a design parameter (Sec VI.B.3); it determines the observer gain L and the tightened state constraint.
  • Initial state deviation set Z_~x = Z_~x = ⟨[0,0,0], diag(0.2,0.2,0)⟩
    Chosen by hand (Sec VI.B.3); determines feedback gain K and tightened input constraint.
  • Contraction ratios lambda_L, lambda_~x, lambda_f = lambda_L=0.85, lambda_~x=0.9, lambda_f=0.95
    Hand-selected tuning parameters governing the conservatism of the invariant sets and terminal set (Sec VI.B.3).
  • Cost sets P, Q, R = P=⟨0,1e-4 I⟩, Q=⟨0,2e1 I⟩, R=⟨0,5e1 I⟩
    Hand-chosen weights for the Minkowski stage and terminal costs (Sec VI.B.3); they influence K_f and the terminal set.
  • Simulation sensor noise bias c_v and spread G_v = c_v=0.5, G_v=0.5, Laplace scale b=G_v/sqrt(2)
    Chosen to create biased, heavy-tailed noise for the observer comparison (Sec VI.A).
assumptions (11)
  • domain assumption Assumption 1: System state x(k) is bounded by a constrained zonotope S_x = ⟨c_x, G_x, F_x, theta_x⟩.
    Used to define the state constraint set and the tightening S_bar{x} = S_x ⊖ (S_e ⊕ S_~x) in Problem 1 (Section III).
  • domain assumption Assumption 2: Control input u(k) is bounded by a constrained zonotope S_u.
    Input constraints; used in the tightened input set S_bar{u} = S_u ⊖ K S_~x (Section III).
  • standard math Assumption 3: The pair (A,B) is controllable.
    Needed for designing feedback gain K and terminal gain K_f; standard in MPC.
  • domain assumption Assumption 4: Process noise w(k) lies in a known constrained zonotope Z_w^c.
    Input to the error propagation (Lemmas 3-4, Theorems 1-2). In the experiments w=0.
  • domain assumption Assumption 5: Perception error is bounded: ||E(o(k))-Cx(k)|| ≤ mu_v for x(k) in B_rho(x_d).
    Central premise for using a bounded set to model perception noise. The experimental noise set covers only 95% of samples, so this is only approximately satisfied (Sec VI.B.2).
  • domain assumption Assumption 6: Sensor noise v(k) lies in a known constrained zonotope Z_v^c.
    Used in the observer design (Theorem 1) and the state-deviation bound (Theorem 2).
  • standard math Assumption 7: The pair (A,C) is observable.
    Needed for designing the Luenberger observer gain L.
  • domain assumption Assumption 8: Estimation error set S_e is a constrained zonotope Z_e^c.
    The error bound used in the tightened state constraint and in Theorems 1-2.
  • domain assumption Assumption 9: State deviation set S_~x is a constrained zonotope Z_~x^c.
    The deviation bound used in the tightened constraints and Theorem 2.
  • standard math Lemma 1 (from [35]): Containment of constrained zonotopes is characterized by existence of Pi, H, gamma satisfying (5).
    Black-box result used in Theorems 1-3 to formulate invariant set conditions as linear constraints.
  • standard math Properties of Minkowski/support functions and polar sets (Remark 2, from [36]).
    Used to convert the Lyapunov inequality into a set-containment condition in Theorem 3.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Robust Model Predictive Control Design for Autonomous Vehicles with Perception-based Observers." pith.science (2026). https://pith.science/paper/WPUJWRDR

@misc{pith2026250905201,
  author       = {Pith},
  title        = {Pith review of: Robust Model Predictive Control Design for Autonomous Vehicles with Perception-based Observers},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/WPUJWRDR}},
  note         = {Machine review of arXiv:2509.05201}
}
read the original abstract

This paper presents a robust model predictive control (MPC) framework that explicitly addresses the non-Gaussian noise inherent in deep learning-based perception modules used for state estimation. Recognizing that accurate uncertainty quantification of the perception module is essential for safe feedback control, our approach departs from the conventional assumption of zero-mean noise quantification of the perception error. Instead, it employs set-based state estimation with constrained zonotopes to capture biased, heavy-tailed uncertainties while maintaining bounded estimation errors. To improve computational efficiency, the robust MPC is reformulated as a linear program (LP), using a Minkowski-Lyapunov-based cost function with an added slack variable to prevent degenerate solutions. Closed-loop stability is ensured through Minkowski-Lyapunov inequalities and contractive zonotopic invariant sets. The largest stabilizing terminal set and its corresponding feedback gain are then derived via an ellipsoidal approximation of the zonotopes. The proposed framework is validated through both simulations and hardware experiments on an omnidirectional mobile robot along with a camera and a convolutional neural network-based perception module implemented within a ROS2 framework. The results demonstrate that the perception-aware MPC provides stable and accurate control performance under heavy-tailed noise conditions, significantly outperforming traditional Gaussian-noise-based designs in terms of both state estimation error bounding and overall control performance.

Figures

Figures reproduced from arXiv: 2509.05201 by the authors.

Figure 1
Figure 1. Block diagram of the closed-loop control architecture featuring a perception-based observer. A camera collects images of the plant, which are processed [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. State estimation comparison between the zonotopic observer and [PITH_FULL_IMAGE:figures/full_fig_p009_2.png] view at source ↗
Figure 4
Figure 4. RobotPerceptionNet architecture. Plus camera (Orbbec Inc., Shenzhen, China) on the master computer captures perception data, and commands are com￾municated to the robot via ROS2 over Wi-Fi. The discrete-time state space kinematics of the omnidirectional robot is [42]  A = I3×3, B =  tsLab tsLab tsLab tsLab tsLab −tsLab tsLab −tsLab ts −ts −ts ts   , where Lab := La +Lb, with La = 135mm and Lb = 85mm. The states … view at source ↗
Figures from the paper (2 more)
Figure 5
Figure 5. Figure 5: (a) Performance of RobotPerceptionNet on the test dataset, showcasing the perception noise and its corresponding zonotopic representa￾tion. (b) A 3D plot illustrating the variation of perception noise magnitude relative to the ground truth locations. A comparative anal…
Figure 6
Figure 6. Figure 6: Control inputs produced by Perception-MPC alongside the imposed [PITH_FULL_IMAGE:figures/full_fig_p011_6.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

47 extracted references · 41 canonical work pages

  1. [1]

    Vision- based safety-critical landing control of quadrotors with external uncer- tainties and collision avoidance,

    J. Lin, Z. Miao, Y . Wang, H. Wang, X. Wang, and R. Fierro, “Vision- based safety-critical landing control of quadrotors with external uncer- tainties and collision avoidance,”IEEE Transactions on Control Systems Technology, vol. 32, no. 4, pp. 1310–1322, 2024. JOURNAL OF LATEX CLASS FILES, VOL. 18, NO. 9, SEPTEMBER 2020 12

    work page 2024

  2. [2]

    An active perception framework for autonomous underwater vehicle navigation under sensor constraints,

    D. Chang, M. Johnson-Roberson, and J. Sun, “An active perception framework for autonomous underwater vehicle navigation under sensor constraints,”IEEE Transactions on Control Systems Technology, vol. 30, no. 6, pp. 2301–2316, 2022

    work page 2022

  3. [3]

    Robust perception under adverse conditions for autonomous driving based on data aug- mentation,

    Z. Zheng, Y . Cheng, Z. Xin, Z. Yu, and B. Zheng, “Robust perception under adverse conditions for autonomous driving based on data aug- mentation,”IEEE Transactions on Intelligent Transportation Systems, vol. 24, no. 12, pp. 13 916–13 929, 2023

    work page 2023

  4. [4]

    Vision-Driven 2D Supervised Fine-Tuning Framework for Bird's Eye View Perception

    L. He, Q. Wang, H. Sun, Q. Xu, B. Gao, S. E. Li, J. Wang, and K. Li, “Vision-driven 2d supervised fine-tuning framework for bird’s eye view perception,”arXiv preprint arXiv:2409.05834, 2024

    work page Pith review arXiv 2024

  5. [5]

    A Multi-Scale Isolation Forest Approach for Real-Time Detection and Filtering of FGSM Adversarial Attacks in Video Streams of Autonomous Vehicles

    R. Abhulimhen, N. Begashaw, G. Comert, C. Zhao, and P. Pisu, “A multi-scale isolation forest approach for real-time detection and filtering of fgsm adversarial attacks in video streams of autonomous vehicles,” arXiv preprint arXiv:2502.16044, 2025

    work page Pith review arXiv 2025

  6. [6]

    Vision-based model predictive control for steering of a nonholonomic mobile robot,

    Z. Li, C. Yang, C.-Y . Su, J. Deng, and W. Zhang, “Vision-based model predictive control for steering of a nonholonomic mobile robot,”IEEE Transactions on Control Systems Technology, vol. 24, no. 2, pp. 553– 564, 2015

    work page 2015

  7. [7]

    Learning a deep neural net policy for end-to-end control of autonomous vehicles,

    V . Rausch, A. Hansen, E. Solowjow, C. Liu, E. Kreuzer, and J. K. Hedrick, “Learning a deep neural net policy for end-to-end control of autonomous vehicles,” in2017 American Control Conference (ACC). IEEE, 2017, pp. 4914–4919

    work page 2017

  8. [8]

    Leader-follower formation tracking control of mobile robots: A visual observer-based approach,

    S. Huang, H. Zeng, W. Lan, and X. Yu, “Leader-follower formation tracking control of mobile robots: A visual observer-based approach,” IEEE Transactions on Control Systems Technology, 2025

    work page 2025

Show all 47 references
  1. [9]

    End-to-end autonomous driving: Challenges and frontiers,

    L. Chen, P. Wu, K. Chitta, B. Jaeger, A. Geiger, and H. Li, “End-to-end autonomous driving: Challenges and frontiers,”IEEE Transactions on Pattern Analysis and Machine Intelligence, 2024

    2024

  2. [10]

    Certainty equivalent perception-based control,

    S. Dean and B. Recht, “Certainty equivalent perception-based control,” inLearning for Dynamics and Control. PMLR, 2021, pp. 399–411

    2021

  3. [11]

    Robust guarantees for perception-based control,

    S. Dean, N. Matni, B. Recht, and V . Ye, “Robust guarantees for perception-based control,” inLearning for Dynamics and Control. PMLR, 2020, pp. 350–360

    2020

  4. [12]

    Accuracy prevents robustness in perception-based control,

    A. A. Al Makdah, V . Katewa, and F. Pasqualetti, “Accuracy prevents robustness in perception-based control,” in2020 American Control Conference (ACC). IEEE, 2020, pp. 3940–3946

    2020

  5. [13]

    Safe adaptive cruise control under perception uncertainty: A deep ensemble and conformal tube model predictive control approach,

    X. Li, A. Girard, and I. Kolmanovsky, “Safe adaptive cruise control under perception uncertainty: A deep ensemble and conformal tube model predictive control approach,”arXiv preprint arXiv:2412.03792, 2024

    2024 arXiv

  6. [14]

    Learning auton- omy: Off-road navigation enhanced by human input,

    A. Nagariya, D. Filev, S. Saripalli, and G. Pandey, “Learning auton- omy: Off-road navigation enhanced by human input,”arXiv preprint arXiv:2502.18760, 2025

    2025 arXiv

  7. [15]

    Robust controller synthesis for vision-based spacecraft guidance and control,

    N. Rahimi, S. Talebi, A. Deole, M. Mesbahi, S. Bandyopadhyay, and A. Rahmani, “Robust controller synthesis for vision-based spacecraft guidance and control,” inAIAA SCITECH 2022 Forum, 2022, p. 2213

    2022

  8. [16]

    Multi-sensor environmental perception and adaptive cruise control of intelligent vehicles using kalman filter,

    P. Wei, Y . Zeng, W. Ouyang, and J. Zhou, “Multi-sensor environmental perception and adaptive cruise control of intelligent vehicles using kalman filter,”IEEE Transactions on Intelligent Transportation Systems, vol. 25, no. 3, pp. 3098–3107, 2023

    2023

  9. [17]

    A comprehensive survey on the integrity of localization systems,

    E. Maharmeh, Z. Alsayed, and F. Nashashibi, “A comprehensive survey on the integrity of localization systems,”Sensors (Basel, Switzerland), vol. 25, no. 2, p. 358, 2025

    2025

  10. [18]

    A survey and framework of cooperative perception: From heterogeneous singleton to hierarchical cooperation,

    Z. Bai, G. Wu, M. J. Barth, Y . Liu, E. A. Sisbot, K. Oguchi, and Z. Huang, “A survey and framework of cooperative perception: From heterogeneous singleton to hierarchical cooperation,”IEEE Transactions on Intelligent Transportation Systems, 2024

    2024

  11. [19]

    Recursive state estimation: Unknown but bounded errors and system inputs,

    F. Schweppe, “Recursive state estimation: Unknown but bounded errors and system inputs,”IEEE transactions on automatic control, vol. 13, no. 1, pp. 22–28, 2003

    2003

  12. [20]

    Zonotopes and kalman observers: Gain optimality under distinct uncertainty paradigms and robust convergence,

    C. Combastel, “Zonotopes and kalman observers: Gain optimality under distinct uncertainty paradigms and robust convergence,”Automatica, vol. 55, pp. 265–273, 2015

    2015

  13. [21]

    An improved model predictive control-based trajectory planning method for automated driving vehicles under uncertainty environments,

    T. Qie, W. Wang, C. Yang, Y . Li, Y . Zhang, W. Liu, and C. Xiang, “An improved model predictive control-based trajectory planning method for automated driving vehicles under uncertainty environments,”IEEE Transactions on Intelligent Transportation Systems, vol. 24, no. 4, pp....

    2022

  14. [22]

    Invariant set planning for quadrotors: Design, analysis, experiments,

    M. Greiff, H. Sinhmar, A. Weiss, K. Berntorp, and S. Di Cairano, “Invariant set planning for quadrotors: Design, analysis, experiments,” IEEE Transactions on Control Systems Technology, 2024

    2024

  15. [23]

    Tube-based stochastic model predictive control with application to wind energy conversion system,

    X. Liu, L. Feng, X. Kong, S. Guo, and K. Y . Lee, “Tube-based stochastic model predictive control with application to wind energy conversion system,”IEEE Transactions on Control Systems Technology, vol. 31, no. 5, pp. 2173–2187, 2023

    2023

  16. [24]

    Robust tube- based mpc for tracking of constrained linear systems with additive disturbances,

    D. Lim ´on, I. Alvarado, T. Alamo, and E. F. Camacho, “Robust tube- based mpc for tracking of constrained linear systems with additive disturbances,”Journal of Process Control, vol. 20, no. 3, pp. 248–260, 2010

    2010

  17. [25]

    Rigid tube model pre- dictive control for linear sampled–data systems,

    F. A. Fontes, S. Rakovi ´c, and I. V . Kolmanovsky, “Rigid tube model pre- dictive control for linear sampled–data systems,”IFAC-PapersOnLine, vol. 50, no. 1, pp. 9840–9845, 2017

    2017

  18. [26]

    The implicit rigid tube model predictive control,

    S. V . Rakovi ´c, “The implicit rigid tube model predictive control,” Automatica, vol. 157, p. 111234, 2023

    2023

  19. [27]

    Elastic tube model predictive control with scaled zonotopic sets,

    S. Diaconescu, F. Stoican, B. D. Ciubotaru, and S. Olaru, “Elastic tube model predictive control with scaled zonotopic sets,”IEEE Control Systems Letters, 2024

    2024

  20. [28]

    Analysis of linear programming in model predictive control,

    D. R. Saffer II and F. J. Doyle III, “Analysis of linear programming in model predictive control,”Computers & chemical engineering, vol. 28, no. 12, pp. 2749–2763, 2004

    2004

  21. [29]

    Set operations and order reductions for constrained zonotopes,

    V . Raghuraman and J. P. Koeln, “Set operations and order reductions for constrained zonotopes,”Automatica, vol. 139, p. 110204, 2022

    2022

  22. [30]

    Constrained zonotopes: A new tool for set-based estimation and fault detection,

    J. K. Scott, D. M. Raimondo, G. R. Marseglia, and R. D. Braatz, “Constrained zonotopes: A new tool for set-based estimation and fault detection,”Automatica, vol. 69, pp. 126–136, 2016

    2016

  23. [31]

    Linear programming and model predictive control,

    C. V . Rao and J. B. Rawlings, “Linear programming and model predictive control,”Journal of Process Control, vol. 10, no. 2-3, pp. 283–289, 2000

    2000

  24. [32]

    Model predictive optimal averaging level control,

    P. J. Campo and M. Morari, “Model predictive optimal averaging level control,”AIChE Journal, vol. 35, no. 4, pp. 579–591, 1989

    1989

  25. [33]

    Minkowski terminal cost functions for mpc,

    S. V . Rakovi ´c and M. Lazar, “Minkowski terminal cost functions for mpc,”Automatica, vol. 48, no. 10, pp. 2721–2725, 2012

    2012

  26. [34]

    Fourier-motzkin elimination and its dual,

    G. B. Dantzig and B. Curtis Eaves, “Fourier-motzkin elimination and its dual,”Journal of Combinatorial Theory, Series A, vol. 14, no. 3, pp. 288–297, 1973. [Online]. Available: https://www.sciencedirect.com/ science/article/pii/0097316573900046

    1973

  27. [35]

    Computing the maximal positive invariant setfor the constrained zonotopic case,

    B. Gheorghe, D.-M. Ioan, F. Stoican, and I. Prodan, “Computing the maximal positive invariant setfor the constrained zonotopic case,”IEEE Control Systems Letters, 2024

    2024

  28. [36]

    Polarity of stability and robust positive invariance,

    S. V . Rakovi ´c, “Polarity of stability and robust positive invariance,” Automatica, vol. 118, p. 109010, 2020

    2020

  29. [37]

    Learn- ing and correcting non-gaussian model errors,

    D. Smyl, T. N. Tallman, J. A. Black, A. Hauptmann, and D. Liu, “Learn- ing and correcting non-gaussian model errors,”Journal of Computational Physics, vol. 432, p. 110152, 2021

    2021

  30. [38]

    Blanchini, S

    F. Blanchini, S. Mianiet al.,Set-theoretic methods in control. Springer, 2008, vol. 78

    2008

  31. [39]

    M. S. Grewal and A. P. Andrews,Kalman filtering: Theory and Practice with MATLAB. John Wiley & Sons, 2014

    2014

  32. [40]

    Minkowski–bellman inequality and equation,

    S. V . Rakovi ´c, “Minkowski–bellman inequality and equation,”Automat- ica, vol. 125, p. 109435, 2021

    2021

  33. [41]

    Scalable zonotope-ellipsoid conversions using the euclidean zonotope norm,

    V . Gaßmann and M. Althoff, “Scalable zonotope-ellipsoid conversions using the euclidean zonotope norm,” in2020 American Control Confer- ence (ACC). IEEE, 2020, pp. 4715–4721

    2020

  34. [42]

    Visual odometry and control for an omnidirectional mobile robot with a downward-facing camera,

    M. Killpack, T. Deyle, C. Anderson, and C. C. Kemp, “Visual odometry and control for an omnidirectional mobile robot with a downward-facing camera,” in2010 IEEE/RSJ International Conference on Intelligent Robots and Systems. IEEE, 2010, pp. 139–146

    2010

  35. [43]

    Rfm-slam: Exploiting rela- tive feature measurements to separate orientation and position estimation in slam,

    S. Agarwal, V . Shree, and S. Chakravorty, “Rfm-slam: Exploiting rela- tive feature measurements to separate orientation and position estimation in slam,” in2017 IEEE International Conference on Robotics and Automation (ICRA). IEEE, 2017, pp. 6307–6314

    2017

  36. [44]

    Pytorch: An imperative style, high-performance deep learn- ing library,

    A. Paszke, “Pytorch: An imperative style, high-performance deep learn- ing library,”arXiv preprint arXiv:1912.01703, 2019

    1912 arXiv

  37. [45]

    Using convolutional neural networks for image recognition,

    S. Hijazi, R. Kumar, C. Rowenet al., “Using convolutional neural networks for image recognition,”Cadence Design Systems Inc.: San Jose, CA, USA, vol. 9, no. 1, 2015

    2015

  38. [46]

    Incorporating state estimation into model predictive control and its application to network traffic control,

    J. Yan and R. R. Bitmead, “Incorporating state estimation into model predictive control and its application to network traffic control,”Auto- matica, vol. 41, no. 4, pp. 595–604, 2005

    2005

  39. [47]

    Stochastic tube mpc with state estimation,

    M. Cannon, Q. Cheng, B. Kouvaritakis, and S. V . Rakovi ´c, “Stochastic tube mpc with state estimation,”Automatica, vol. 48, no. 3, pp. 536–541, 2012. JOURNAL OF LATEX CLASS FILES, VOL. 18, NO. 9, SEPTEMBER 2020 13 Nariman Niknejadreceived his bachelor’s degree in Mechanical E...

    2012

Pith tools

Reviewed August 5, 2026 · model on record in the stance chip above.