Tire Wear Aware Trajectory Tracking Control for Multi-axle Swerve-drive Autonomous Mobile Robots

Fen Liu; Lihua Xie; Shenghai Yuan; Thien-Minh Nguyen; Tianxin Hu; Xinhang Xu

REVIEW 4 major objections 6 minor 52 references

Tire Wear Aware Trajectory Tracking Control for Multi-axle Swerve-drive Autonomous Mobile Robots

T0 review · 4 major / 6 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read Adding tire wear to the MPC objective cuts simulated tire wear by up to 65%.

desk verdict A coherent simulation study of a reasonable idea, but the headline wear reductions are partly by construction because the same model defines the MPC cost and the evaluation metric. read the letter →

arxiv 2506.04752 v1 pith:G3MEEECN submitted 2025-06-05 cs.RO math.OC

Tianxin Hu , Xinhang Xu , Thien-Minh Nguyen , Fen Liu , Shenghai Yuan , Lihua Xie This is my paper

classification cs.ROmath.OC

keywords multi-axleswerve-driveAGVmodelpredictivecontroltirewearminimizationMagicFormulatrajectorytrackinghierarchicalsimulatedannealingslipratioandangle

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

This paper claims that tire wear on multi-axle swerve-drive autonomous guided vehicles can be reduced substantially by putting a tire-wear term directly into the objective of a model predictive controller, rather than treating wear as an afterthought. The authors express tire wear as three power losses—longitudinal slip, lateral slip, and steering—derived from a simplified wear model together with the Magic Formula tire model, and include them in the MPC cost alongside tracking error. In simulation, the wear-aware controller reduces total tire wear work by 19.19% on a winding curve and by 65.20% in a 60-degree heading-offset maneuver compared with the kinematic baseline, while keeping tracking error within what the authors call acceptable limits. The same simplified wear formulas define both the cost being optimized and the reported savings, so the numbers measure wear as the model defines it rather than as physical tread loss.

What carries the argument

The load-bearing object is the simplified tire wear model of Eqs. (13)-(16), which defines the three wear power components, slip-ratio wear, slip-angle wear, and steering wear, that form the tire wear vector $P_{tw}$. This same vector enters the MPC cost function (Eq. 22) as $P_{tw}^T W_P P_{tw}$, and its time-integrated versions (Eqs. 28-31) are the evaluation metric, so the mechanism the controller optimizes is exactly the mechanism used to measure success. Around it, the Pacejka Magic Formula tire model supplies the forces $f(\cdot)$ that connect slip quantities to wear power, while the hierarchical split, velocity commands at the MPC level and torque tracking delegated to lower-level controllers, keeps the control vector at dimension $4n$ and lets simulated annealing solve the resulting nonconvex problem in real time.

What would settle it

Run the wear-aware and wear-unaware trajectories on an instrumented vehicle or in a high-fidelity wear simulator and measure actual tread depth or rubber mass loss: if the predicted 19% to 65% wear reductions do not appear as physical abrasion, the savings are an artifact of the simplified model. A cheaper check is to feed both recorded control trajectories into an independent abrasion model, such as an Archard-type wear law, and compare the predicted wear ratio.

Watch

Extended reading notes

Core claim

The paper's central claim is that tire wear can be minimized concurrently with trajectory tracking by formulating the control problem as a nonlinear MPC whose cost function is the weighted sum of tracking error and a tire-wear power vector $P_{tw}$ computed from wheel slip ratio, slip angle, and steering motion. Using the Magic Formula tire model to map slip to forces, the authors reduce each wheel's wear power to $P_{sij}=|{}^x_W F_{ij}(\omega_{ij} r_w - {}^x_W v_{ij})|$, $P_{\alpha ij}=|{}^y_W F_{ij}\,{}^y_W v_{ij}|$, and $P_{tij}=|k_t F_{Nij}\,\omega_{sij}|$, and they drive these terms down through velocity-level control inputs, namely wheel angular velocity and steering angular velocity. They claim this hierarchical, velocity-level formulation keeps the problem small enough for simulated annealing to solve in real time on an ordinary PC, and that the resulting controller reduces tire wear by 19.19% on a curve and 65.20% under a 60-degree heading offset while preserving tracking accuracy.

Load-bearing premise

The load-bearing premise is that the simplified wear-power formulas for slip ratio, slip angle, and steering accurately capture real tire wear, because the same formulas define both the optimized cost and the reported savings, and no physical tread-loss measurement is provided.

Editorial extensions

If this is right

On curved paths, the wear-aware controller cuts total tire wear work by 19.19% at 35 km/h compared with the same dynamic model without the wear term, at the cost of a mean tracking-error index increase from 9.60 to 18.44.
In a 60-degree heading-offset recovery maneuver, the wear-aware controller reduces tire wear work by 65.20% versus the kinematic baseline and by 16.58% and 14.43% versus the two dynamic baselines, while keeping the best performance-balance index.
Because the MPC operates on wheel angular velocities and steering angular velocities rather than torques, the control dimension stays $4n$, and the optimization can be solved in real time by simulated annealing on a standard PC.
Across speeds from 5 to 35 km/h and gross vehicle weights from 8 to 20 tons, the wear-aware controller consistently achieves a lower performance-balance index, favoring lower modeled wear and lower tracking error, than the baselines.

Reading between the lines

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

The paper leaves implicit that the 65.20% figure is partly a modeling comparison: the kinematic baseline assumes zero slip, so some of that reduction reflects the dynamic model's ability to predict slip rather than the wear-aware objective alone; the cleaner isolation of the wear term is the 19.19% NTWO-versus-TWO comparison.
A natural testable extension is to sweep the weights $W_Y$ and $W_P$ and map the full Pareto frontier between tracking error and modeled tire wear, since the paper reports only one fixed operating point.
The same velocity-level wear formulation could transfer to other over-actuated wheeled robots, such as forklifts or port tractors, where slip-induced wear dominates maintenance costs, and to online replanning if the solver latency remains acceptable.
Relaxing the constant-vertical-load assumption, which the authors name as their own limitation, would show whether the modeled wear savings persist under acceleration and braking when load transfer changes each tire's contact force.

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.

Desk Editor's Note

read the letter

The one thing to know: the paper is a coherent simulation study of a reasonable idea, but the headline wear reductions are partly by construction—the same simplified wear model defines both the MPC cost and the evaluation metric.

What's new: it puts a Da Silva-style wear model and a velocity-level Pacejka tire model into an MPC for multi-axle swerve-drive AGVs. That specific integration is not in the cited literature, and the hierarchical decomposition (velocity commands with low-level trackers) is sensible. The simulation framework is thorough: three scenarios, five baselines, speeds 5-35 km/h, loads 8-20 t. The equations are mostly consistent, and the authors explicitly flag Assumption 2 (constant vertical load) as a limitation. They also report values for all components of wear.

The soft spots are real but not fatal. The circularity is the main one: Eqs. (13)-(16) appear in the cost (Eq. 22) and in the metrics (Eqs. 28-31), so minimizing J necessarily lowers W_tw. No calibration of B, C, D, E, k_t against abrasion or a higher-fidelity tire model is provided. In Case 2 the 65.20% reduction is against the kinematic baseline, not against their own NTWO; the fairer comparison is around 10%. The abstract says 'maintaining tracking accuracy' but RMSE roughly doubles (e.g., 9.60 to 18.44 cm at 35 km/h); later text says 'slightly compromising', which is closer but still understated. The trade-off is favorable only under their Ω index. Minor: the real-time claim has no timing data, and the free weights Q and L are not discussed.

Bottom line: the idea is plausible and the paper is honest in structure, but the central claims need re-framing. Send it to peer review with expectation of major revision: report timing, use NTWO as the primary baseline, and either validate the wear metric physically or call it a surrogate. For the MS-AGV control subfield, this is worth a serious referee; not for a broad robotics audience.

Referee Report

4 major / 6 minor

Summary. The paper proposes a hierarchical model predictive control (MPC) scheme for trajectory tracking of multi-axle swerve-drive autonomous ground vehicles. The high-level controller commands wheel angular velocities and steering angular velocities, while a lower level tracks these velocity commands. Tire forces are computed from the Pacejka Magic Formula, and a simplified tire wear power model is appended to the MPC cost function. The authors report simulations at speeds of 5–35 km/h and gross vehicle weights of 8–20 t, claiming a 19.19% tire wear reduction in a curve-tracking scenario and a 65.20% reduction in an offset-heading scenario compared with a kinematic baseline. The manuscript also claims real-time solvability using simulated annealing on a normal personal computer.

Significance. If the physical wear claims were established, the work would be a useful contribution to heavy-duty AGV operations, where tire maintenance cost is indeed a practical concern. The hierarchical decomposition, the use of velocity-level control inputs, and the systematic speed/load sweeps against multiple baselines are reasonable ingredients. However, the headline quantitative claims currently rest on a wear surrogate that is both the optimization objective and the evaluation metric, and no independent validation is provided. In addition, the paper's own tables show that tracking accuracy degrades substantially when wear optimization is enabled, which contradicts the abstract's 'maintaining tracking accuracy' claim. The contribution as written is therefore a demonstration that optimizing a model reduces that model's output in simulation, not a validated reduction in physical tire wear. The control formulation may still be valuable after reframing and additional evidence.

major comments (4)

[§3.2, §3.5, §4.3 (Eqs. 16, 22, 28–31)] The wear metric used for evaluation is exactly the quantity being minimized. Equations (13)–(16) define the wear powers P_s, P_alpha, P_t and their sum P_tw; Eq. (22) places P_tw in the MPC cost; Eqs. (28)–(31) compute the reported total wear W_tw from the same P_s, P_alpha, P_t. Minimizing J therefore reduces the reported W_tw relative to NTWO by construction whenever the tracking penalty permits it. The 19.19% and 65.20% reductions are reductions in the optimized surrogate, not independent evidence about physical tire abrasion. The authors should either calibrate or validate the wear model against physical measurements or a higher-fidelity tire/abrasion model, or explicitly restrict all claims to 'reductions in the model-defined wear metric.'
[§4.4.1, Table 1 and Table 3] The claim that tracking accuracy is 'maintained' is not supported by the reported errors. In Table 1 at 35 km/h, average RMSE increases from 9.60 cm (NTWO) to 18.44 cm (TWO); at 15 km/h it increases from 0.62 cm to 2.24 cm. In Table 3 at 20 t it increases from 0.98 cm to 5.40 cm. These are roughly twofold to fivefold increases. The statement 'these values remain within acceptable limits' is not quantified, and no tracking-error constraint is imposed in the optimization. The accuracy claim should be revised or a formal bound on the tracking error should be included.
[Abstract, §4.1] The real-time claim is unsupported by any timing evidence. No solver computation times, hardware specifications, prediction/control horizons (N_p, N_c), or simulated-annealing hyperparameters are reported. The statement that the method 'can be solved by simulated annealing in real-time on a normal personal computer' cannot be verified. The authors should provide wall-clock timing distributions for the solver and specify all optimizer and MPC parameters, or remove the real-time claim.
[§4.2, §4.4.2, Eq. (33)] The headline 65.20% reduction in Case 2 is computed against the Kinematic baseline, whose average RMSE is about 111 cm, far worse than all dynamic baselines; against NTWO, the reduction is only about 9.8% in that scenario (1.84e9 J to 1.66e9 J). The abstract's wording 'compared to the kinematic model without considering the tire wear optimization' is therefore comparing to a much weaker controller, and the contribution of the wear term itself is much smaller. In addition, the performance balance index uses e_bar, the arithmetic mean of errors measured in centimeters and degrees, which is dimensionally inconsistent. Comparisons against NTWO should be reported separately, and unit-consistent metrics should be used.

minor comments (6)

[§3.4, Eq. (10)] The slip-ratio definition divides by the magnitude of the wheel longitudinal velocity x_W v_ij, which can be zero when a wheel is not moving longitudinally; the paper should state how this singularity is handled.
[§3.5, Theorem 1] Theorem 1 is not a theorem in the usual sense: the justification restates that the wheel angular velocities and steering rates influence the forces and wear, but it does not prove feasibility of the velocity-level control inputs under the stated constraints.
[§3.5, Eq. (22)] The notation is inconsistent: Eq. (16) defines P_tw as a 3-by-1 vector, while Eq. (22) uses P_tw as a stacked horizon vector. Please distinguish P_tw(k) from the horizon-stacked P and write the cost as a sum over the horizon or define the stacked vector explicitly.
[§4.4, Figure 9] The caption 'Track better / Less worn-out' is contradicted by Table 1, where the Kinematic model has lower total wear than TWO (e.g., 3.16e8 J versus 7.37e8 J at 35 km/h). The caption should qualify that TWO has the best balance index, not the lowest wear.
[§2.3 and Conclusion] Assumption 2 (constant vertical load) is acknowledged as a limitation, but no sensitivity analysis is given. Since the steering wear term in Eq. (15) is linear in F_N, a brief sensitivity test with perturbed load distributions would help show robustness of the reported conclusions.
[Abstract and text] The manuscript uses 'experiment' to describe MATLAB simulations, which is potentially misleading. It should clearly state that all results are simulation-based unless hardware experiments are performed.

Circularity Check

1 steps flagged · score 7.0 of 10

Reported tire-wear reductions are measured with the same surrogate model that the MPC objective minimizes, so the headline numbers are in-sample consequences; the 'maintaining tracking accuracy' claim is not supported by the tables.

fitted input called prediction [Section 3.2 Eq. (16); Section 3.5 Eq. (22); Section 4.3 Eqs. (28)-(31); Abstract]
"The tire wear power P_tw can be calculated as: P_tw = sum_i sum_j [P_sij; P_alphaij; P_tij] ... the cost function is defined as: J=(Y−Y_t)^T W_Y (Y−Y_t)+P^T_tw W_P P_tw ... Total Tire Wear Work W_tw: The sum of work done by wear caused by slip angle, slip ratio, and steering, defined as follows: W_tw = W_s + W_alpha + W_t."

Eq. (16) defines P_tw as the stacked per-step wear powers P_s, P_alpha, P_t. The MPC cost in Eq. (22) penalizes P^T_tw W_P P_tw, which is exactly these same modeled quantities (weighted and summed over the horizon). The evaluation metric in Eqs. (28)-(31) defines W_tw as the sum over the same P_s, P_alpha, P_t along the trajectory. Minimizing J therefore must reduce the same modeled power terms that are later reported as physical tire-wear reductions (19.19% and 65.20%). No physical abrasion measurement or higher-fidelity tire model is used to validate the surrogate, so the headline reductions are in-sample consequences of the objective rather than independent evidence about physical tire wear.

full rationale

The central controller derivation—MPC for trajectory tracking using a Pacejka-based dynamic model—is self-contained and not circular: the dynamics (Eqs. 1-18) are standard, the baselines are external models, and the tracking-error part of the cost is independent of the wear metric. However, the paper's headline wear reductions are computed with the same surrogate wear model that is optimized: Eq. (22) penalizes P_tw from Eq. (16), and Eqs. (28)-(31) sum the same P_s, P_alpha, P_t to define W_tw. Thus TWO lowering W_tw relative to NTWO is an in-sample consequence of adding the wear penalty, not an independent validation of physical tire abrasion. The authors honestly acknowledge the constant-load limitation (Assumption 2 and the Conclusion), but no higher-fidelity wear model or physical measurement is provided. Separately, the abstract's claim of 'maintaining the tracking accuracy' is not supported by the tables (e.g., at 35 km/h the average RMSE rises from 9.60 cm to 18.44 cm; at 15 km/h from 0.62 cm to 2.24 cm), which is a consistency issue rather than a circularity. No load-bearing self-citation chain or imported uniqueness theorem was found.

Assumptions & free parameters 5 free parameters · 6 assumptions · 0 invented entities

No new physical entities, forces, or conserved quantities are introduced. The central burden comes from the assumed accuracy of the simplified wear model and the unstated numerical parameters required to reproduce the experiments.

free parameters (5)

MPC weight matrices Q and L = not reported
Q_X, Q_Y, Q_Phi and L_s, L_alpha, L_t determine the trade-off between tracking error and tire wear; values are hand-tuned and not given (Section 3.5).
Magic Formula coefficients B, C, D, E = not reported
Longitudinal and lateral force coefficients are required in f(xi) (Section 3.1) and are not specified numerically; they come from empirical fitting to tire test data.
Steering wear loss coefficient k_t = not reported
Appears in P_tij (Eq. 15); no value or source is given.
Prediction and control horizons Np, Nc = not reported
Horizon lengths strongly affect MPC behavior and solver workload; not stated anywhere.
Simulated annealing hyperparameters = not reported
Annealing schedule, number of iterations, and acceptance criteria are not described, so the real-time claim cannot be evaluated.

assumptions (6)

domain assumption The vehicle operates on flat terrain with zero pitch and roll (Assumption 1).
Restricts dynamics to planar motion and is standard for indoor and port AGVs; the paper cites [46].
domain assumption Vertical tire load F_Nij remains constant during motion (Assumption 2).
Neglects load transfer due to acceleration; the conclusion later states uniform load distribution, a stronger claim not derived in Section 3.2.
domain assumption The Pacejka Magic Formula is an adequate tire force model at the velocity level.
Used in Eqs. (11)-(12) and onward; no validation against measured tire data in this paper.
domain assumption The simplified tire wear model of Da Silva et al. [49] correctly represents physical wear.
Equations (13)-(15) define wear power; this is the central surrogate whose accuracy is never checked against real wear.
ad hoc to paper Low-level controllers perfectly track wheel angular velocity and steering angular velocity commands.
Hierarchical design (Fig. 5) removes actuator dynamics from the MPC model; no bandwidth or tracking error analysis is given.
standard math Forward Euler discretization with T=0.01s is sufficiently accurate.
Used in Eqs. (19)-(21); ODE45 is used in evaluation, but the controller model assumes Euler.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Tire Wear Aware Trajectory Tracking Control for Multi-axle Swerve-drive Autonomous Mobile Robots." pith.science (2026). https://pith.science/paper/G3MEEECN

@misc{pith2026250604752,
  author       = {Pith},
  title        = {Pith review of: Tire Wear Aware Trajectory Tracking Control for Multi-axle Swerve-drive Autonomous Mobile Robots},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/G3MEEECN}},
  note         = {Machine review of arXiv:2506.04752}
}

read the original abstract

Multi-axle Swerve-drive Autonomous Mobile Robots (MS-AGVs) equipped with independently steerable wheels are commonly used for high-payload transportation. In this work, we present a novel model predictive control (MPC) method for MS-AGV trajectory tracking that takes tire wear minimization consideration in the objective function. To speed up the problem-solving process, we propose a hierarchical controller design and simplify the dynamic model by integrating the \textit{magic formula tire model} and \textit{simplified tire wear model}. In the experiment, the proposed method can be solved by simulated annealing in real-time on a normal personal computer and by incorporating tire wear into the objective function, tire wear is reduced by 19.19\% while maintaining the tracking accuracy in curve-tracking experiments. In the more challenging scene: the desired trajectory is offset by 60 degrees from the vehicle's heading, the reduction in tire wear increased to 65.20\% compared to the kinematic model without considering the tire wear optimization.

Figures

Figures reproduced from arXiv: 2506.04752 by the authors.

**Figure 1.** Motivation of the Proposed MS-AGV planning solution. (FWS) [28] have been proposed. With the assumption that the steering centers are restricted along the first, last, or middle axles, these methods can use fewer states to describe MS-AGV while limiting its maneuverability. In contrast, the rotation center distance-based steering models (D-based) [29, 30] allow for a more flexible distribution of the steering center… view at source ↗

**Figure 2.** Diagram of MS-AGV Coordinate System validate the proposed method by comparing the performance with previous works and provide a comprehensive performance analysis. Section 5 concludes the paper. 2. MS-AGV Dynamics and Problem Statement 2.1. Preliminary As shown in [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗

**Figure 3.** Diagram of MS-AGV Tire Dynamic Model ⃗𝑧G , ⃗𝑧B and ⃗𝑧W𝑖𝑗 axes, as follows: 𝑹𝑧 (𝜓) = [ cos 𝜓 − sin 𝜓 sin 𝜓 cos 𝜓 ] . (1) Since MS-AGV is commonly employed in indoor environments such as laboratories and factory workshops, as well as outdoor settings like ports, stations, and logistics centers, where the ground is typically well-paved and flat [44, 45]. In this work, we design a controller for the MS-AGV’s planar mot… view at source ↗

Figures from the paper (7 more)

**Figure 4.** Figure 4: Schematic diagram of tire Wear model where 𝑥 𝑊 𝑣𝑖𝑗 and 𝑦 𝑊 𝑣𝑖𝑗 are the velocity components in the wheel coordinate system W. The relationship between each wheel’s slip angle[47], 𝛼𝑖𝑗 ∈ ℝ, and the steering angles 𝛿𝑖𝑗 is given as follows: 𝛼𝑖𝑗 = 𝛿𝑖𝑗 − arctan 𝑦 𝐵 𝑣𝑖𝑗 𝑥 𝐵 𝑣…

**Figure 5.** Figure 5: Controller Structure 3.4. Model Discretization As shown in [PITH_FULL_IMAGE:figures/full_fig_p005_5.png]

**Figure 6.** Figure 6: Tire Longitudinal and Lateral Force Characteristics The cost function combines the trajectory tracking error and the predicted tire wear, weighted by 𝑊𝑌 and 𝑊𝑃 , respectively. By tuning these weights, the controller can slightly reduce tracking accuracy to significant…

**Figure 7.** Figure 7: Trajectory Comparison in Case 1 minimum trajectory tracking error. When tire wear optimization (TWO) is incorporated, tire wear is reduced across all speeds, albeit with a slight increase in tracking error. For instance, at 35 km/h, the total tire wear work 𝑊𝑡𝑤 decrea…

**Figure 11.** Figure 11: Trajectory Comparison in Case 2 Proposed Model Tracks Angle Better In a Step Response Tests Proposed Model Tracks Better With Less Oscillation and Errors Turn and Drive Most Common Task For Delivery Robots [PITH_FULL_IMAGE:figures/full_fig_p009_11.png]

**Figure 10.** Figure 10: Comparison of Steering Center In contrast, the Kinematic Method exhibits substantial tracking errors, especially during high-speed curved trajectories, due to its inability to account for tire dynamics, particularly slip. This highlights the advantage of the TWO, wh…

**Figure 13.** Figure 13: Performance Balance Comparison in Case 3 5. Conclusion This paper proposes a tracking model for Multi-Axle Swerve-Drive AMRs based on dynamic analysis, incorporating a tire wear model and tire dynamics model based on the Pacejka Magic Formula. The problem is formulat…

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

52 extracted references · 50 canonical work pages

[1]

Z. Li, J. Li, Y. Li, An integrated path planning framework for multi- obstacle avoidance of the multi-axle autonomous vehicle with en- hancedsafetyandstability,IEEETransactionsonVehicularTechnol- ogy 73 (5) (2023) 6368–6382

work page 2023
[2]

Beyersdorfer, S

S. Beyersdorfer, S. Wagner, Novel model based path planning for multi-axle steered heavy load vehicles, in: 16th International IEEE Conference on Intelligent Transportation Systems (ITSC 2013), IEEE, 2013, pp. 424–429

work page 2013
[3]

C.Tang,T.Li,T.Li,J.Ma,H.Xiong,All-wheelsteeringpathtracking controlformulti-axlevehiclebasedonadaptivempc,in:Proceedings of the 2024 4th International Conference on Computer, Control and Robotics (ICCCR), IEEE, Shanghai, China, 2024, pp. 183–190

work page 2024
[4]

Y. Zhu, J. Zhu, P. Zhang, Local obstacle avoidance control for multi- axle and multi-steering-mode wheeled robot based on window-zone division strategy, Robotics and Autonomous Systems 183 (2025) 104843

work page 2025
[5]

X.Gan,Z.Huo,W.Li,Dp-a*:Forpathplaningofugvandcontactless delivery, IEEE Transactions on Intelligent Transportation Systems 25 (1) (2023) 907–919

work page 2023
[6]

R. Bai, S. Yuan, K. Li, H. Guo, W.-Y. Yau, L. Xie, Realm: Real-time line-of-sight maintenance in multi-robot navigation with unknown obstacles, in: Proceedings of the IEEE International Conference on Robotics and Automation (ICRA), Atlanta, USA, 2025

work page 2025
[7]

Y. Ma, J. Xu, S. Yuan, T. Zhi, W. Yu, J. Zhou, L. Xie, Mm-lins: a multi-map lidar-inertial system for over-degenerate environments, IEEE Transactions on Intelligent Vehicles - (2024) 1–11

work page 2024
[8]

Nguyen, S

T.-M. Nguyen, S. Yuan, M. Cao, Y. Lyu, T. H. Nguyen, L. Xie, Ntu viral: A visual-inertial-ranging-lidar dataset, from an aerial vehicle viewpoint, The International Journal of Robotics Research 41 (3) (2022) 270–280

work page 2022

Show all 52 references

[9]

Nguyen, S

T.-M. Nguyen, S. Yuan, T. H. Nguyen, P. Yin, H. Cao, L. Xie, M.Wozniak,P.Jensfelt,M.Thiel,J.Ziegenbein,etal.,Mcd:Diverse large-scale multi-campus dataset for robot perception, in: Proceed- ings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, Seattle, Was...

2024
[10]

2905– 2910

H.Zhao,Z.Liu,Z.Li,S.Zhou,W.Chen,C.Suo,Y.-H.Liu,Modelling anddynamictrackingcontrolofindustrialvehicleswithtractor-trailer structure, in: 2019 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), IEEE, Macau, China, 2019, pp. 2905– 2910

2019
[11]

Islam, A

F. Islam, A. Vemula, S.-K. Kim, A. Dornbush, O. Salzman, M. Likhachev, Planning, learning and reasoning framework for robot truckunloading,in:2020IEEEInternationalConferenceonRobotics and Automation (ICRA), IEEE, Paris, France, 2020, pp. 5011–5017

2020
[12]

D. E. Williams, Generalised multi-axle vehicle handling, Vehicle system dynamics 50 (1) (2012) 149–166

2012
[13]

J. Li, X. Xu, J. Liu, K. Cao, S. Yuan, L. Xie, Ua-mpc: Uncertainty- aware model predictive control for motorized lidar odometry, IEEE Robotics and Automation Letters 10 (4) (2025) 3652–3659.doi: 10.1109/LRA.2025.3544978

2025
[14]

K. Cao, M. Cao, S. Yuan, L. Xie, Direct: A differential dynamic pro- grammingbasedframeworkfortrajectorygeneration,IEEERobotics and Automation Letters 7 (2) (2022) 2439–2446

2022
[15]

X. Xu, M. Cao, S. Yuan, T. H. Nguyen, T.-M. Nguyen, L. Xie, A cost-effective cooperative exploration and inspection strategy for heterogeneous aerial system, in: Proceedings of the 2024 IEEE 18th International Conference on Control & Automation (ICCA), IEEE, Reykjavik, Iceland...

2024
[16]

T.-M.Nguyen,Z.Cao,K.Li,S.Yuan,L.Xie,Gptr:Gaussianprocess trajectory representation for continuous-time motion estimation, in: arXiv preprint arXiv:2410.22931, 2024

2024 arXiv
[17]

Zhang, J

J.-X. Zhang, J. Ding, T. Chai, Fault-tolerant prescribed performance control of wheeled mobile robots: A mixed-gain adaption approach, IEEE Transactions on Automatic Control 69 (8) (2024) 5500–5507

2024
[18]

Zhang, T

J.-X. Zhang, T. Yang, T. Chai, Neural network control of underactu- atedsurfacevehicleswithprescribedtrajectorytrackingperformance, IEEETransactionsonNeuralNetworksandLearningSystems(2022)

2022
[19]

Page 11 of 13

M.Cao,K.Cao,S.Yuan,K.Liu,Y.L.Wong,L.Xie,Pathplanningfor multiple tethered robots using topological braids, in: Proceedings of Robotics:ScienceandSystems(RSS),Robotics:ScienceandSystems Foundation, Daegu, South Korea, 2023. Page 11 of 13

2023
[20]

W. Wang, J. Li, X. Li, Z. Li, N. Guo, Multi-objective collaborative control method for multi-axle distributed vehicle assisted driving, Applied Sciences 13 (13) (2023) 7769

2023
[21]

M. Cao, X. Xu, Y. Yang, J. Li, T. Jin, P. Wang, T.-Y. Hung, G. Lin, L. Xie, Learning dynamic weight adjustment for spatial- temporal trajectory planning in crowd navigation, arXiv preprint arXiv:2412.00555 (2024)

2024 arXiv
[22]

M. Cao, X. Xu, S. Yuan, K. Cao, K. Liu, L. Xie, Doublebee: A hybridaerial-groundrobotwithtwoactivewheels,in:2023IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), IEEE, Detroit, Michigan, USA, 2023, pp. 6962–6969

2023
[23]

J. Xu, Y. Ma, Y. Li, X. Zhang, J. Zhou, S. Yuan, L. Xie, Dynamic initialization for lidar-inertial slam, IEEE/ASME Transactions on Mechatronics (2025) 1–12doi:10.1109/TMECH.2025.3554878

2025
[24]

M. Cao, Y. Lyu, S. Yuan, L. Xie, Online trajectory correction and trackingforfacadeinspectionusingautonomousuav,in:Proceedings of the 2020 IEEE 16th International Conference on Control & Au- tomation (ICCA), IEEE, Singapore, 2020, pp. 1149–1154

2020
[25]

Chaudhuri, V

S. Chaudhuri, V. Saini, M. Singh, Kinematic analysis of multi-axle steering system for articulated vehicle, Tech. rep., SAE Technical Paper (2009)

2009
[26]

K. Wu, H. Wang, M. A. Esfahani, S. Yuan, Learn to navigate au- tonomouslythroughdeepreinforcementlearning,IEEETransactions on Industrial Electronics 69 (5) (2021) 5342–5352

2021
[27]

Y. Ye, L. He, Q. Zhang, Steering control strategies for a four-wheel- independent-steering bin managing robot, IFAC-PapersOnLine 49 (16) (2016) 39–44

2016
[28]

F.-x.Xu,C.Zhou,X.-h.Liu,Hierarchicalcontrolstrategiesformulti- mode steering system of emergency rescue vehicle, Mechatronics 85 (2022) 102834

2022
[29]

Zhang, L

P. Zhang, L. Gao, Y. Zhu, Study on control schemes of flexible steering system of a multi-axle all-wheel-steering robot, Advances in Mechanical Engineering 8 (6) (2016) 1687814016651556

2016
[30]

P.X.Zhang,L.Gao,Y.Q.Zhu,Thesteeringperformanceanalysisof multi-axle vehicle based on sideslip angle control strategy, Applied Mechanics and Materials 701 (2015) 799–802

2015
[31]

M. A. Esfahani, H. Wang, K. Wu, S. Yuan, Unsupervised scene cate- gorization,pathsegmentationandlandmarkextractionwhiletraveling path,in:202016thInternationalConferenceonControl,Automation, Robotics and Vision (ICARCV), IEEE, Shenzhen, China, 2020, pp. 190–195

2020
[32]

F.Gao,X.-y.Li,Turningcharacteristicstudyofmulti-axlecompound steering vehicle, in: 2014 IEEE Conference and Expo Transporta- tion Electrification Asia-Pacific (ITEC Asia-Pacific), IEEE, Beijing, China, 2014, pp. 1–5

2014
[33]

Zhang, X.-j

Z. Zhang, X.-j. Ma, C.-g. Liu, S.-g. Wei, Dual-steering mode based on direct yaw moment control for multi-wheel hub motor driven vehicles: Theoretical design and experimental assessment, Defence Technology 18 (1) (2022) 49–61

2022
[34]

HU, M.-m

J.-b. HU, M.-m. FU, X.-y. LI, J. Ni, Control strategy of the multi- axle distributed mechanic and electric drive vehicle’s turning radius, TransactionsofBeijinginstituteofTechnology36(11)(2016)1131– 1135

2016
[35]

Z. Wu, C. Chen, On optimizing steering performance of multi-axle vehicle based on driving force control, in: MATEC Web of Confer- ences, Vol. 124, EDP Sciences, Beijing, China, 2017, p. 07005

2017
[36]

J. Li, Y. Liu, T. Xu, P. Qu, Control strategy for all-wheel coopera- tive steering of multi-axle vehicle, Tech. rep., SAE Technical Paper (2023)

2023
[37]

G.Tong,X.Jin,Studyonthesimulationofradialtirewearcharacter- istics, WSEAS Transactions on Systems 11 (8) (2012) 419–429

2012
[38]

Huang, Y

H. Huang, Y. Chiu, C. Wang, X. Jin, Three-dimensional global pattern prediction for tyre tread wear, Proceedings of the Institution of Mechanical Engineers, Part D: Journal of automobile engineering 229 (2) (2015) 197–213

2015
[39]

H.-b.Huang,Y.-J.Chiu,X.-x.Jin,Numericalcalculationofirregular tire wear caused by tread self-excited vibration and sensitivity analy- sis, Journal of Mechanical Science and Technology 27 (2013) 1923– 1931

2013
[40]

K. Pan, H. Zheng, J. Wu, H. Xiao, Research on steering control of multi-axle steering heavy commercial vehicle based on reducing tire wear, SAE International Journal of Vehicle Dynamics, Stability, and NVH 4 (10-04-01-0005) (2020) 67–80

2020
[41]

Huanjie, Y

W. Huanjie, Y. Shihua, X. Jinsong, Research on braking strategy and performance considering the tire model for multi-axle vehicle, in: 2014IEEEConferenceandExpoTransportationElectrificationAsia- Pacific (ITEC Asia-Pacific), IEEE, Beijing, China, 2014, pp. 1–5

2014
[42]

H.Du,X.Zhu,Q.Liu,T.Ren,Y.Wang,Hierarchicalcoordinatedcon- trol of multi-axle steering for heavy-duty vehicle based on tire lateral and longitudinal forces optimization, Proceedings of the Institution ofMechanicalEngineers,PartD:JournalofAutomobileEngineering 238 (9) (2024) 2727–2740

2024
[43]

X. Li, S. Yuan, H. Cai, S. Lu, W. Wang, J. Liu, Ll-localizer: A life- long localization system based on dynamic i-octree, arXiv preprint arXiv:2504.01583 (2025)

2025 arXiv
[44]

S. Yuan, B. Lou, T.-M. Nguyen, P. Yin, M. Cao, X. Xu, J. Li, J. Xu, S. Chen, L. Xie, Large-scale uwb anchor calibration and one- shot localization using gaussian process, in: Proceedings of the 2025 IEEEInternationalConferenceonRoboticsandAutomation(ICRA), Atlanta, USA, 2025

2025
[45]

Nguyen, Y

T.-M. Nguyen, Y. Yang, T.-D. Nguyen, S. Yuan, L. Xie, Uloc: Learning to localize in complex large-scale environments with ultra- wideband ranges, in: Proceedings of the IEEE International Confer- ence on Robotics and Automation (ICRA), Atlanta, USA, 2025

2025
[46]

V.Skrickij,P.Kojis,E.Šabanovič,B.Shyrokau,V.Ivanov,Reviewof integrated chassis control techniques for automated ground vehicles, Sensors 24 (2) (2024) 600

2024
[47]

Pacejka, Tire and vehicle dynamics, Elsevier, 2005

H. Pacejka, Tire and vehicle dynamics, Elsevier, 2005

2005
[48]

H. B. Pacejka, E. Bakker, The magic formula tyre model, Vehicle system dynamics 21 (S1) (1992) 1–18

1992
[49]

M.DaSilva,R.Cunha,A.Neto,Asimplifiedmodelforevaluatingtire wear during conceptual design, International Journal of Automotive Technology 13 (2012) 915–922

2012
[50]

S. Zhou, Y. Tian, P. Walker, N. Zhang, Impact of the tyre dynamics onautonomousvehiclepathfollowingcontrolwithfrontwheelsteer- ing and differential motor torque, IET Intelligent Transport Systems 17 (8) (2023) 1629–1648

2023
[51]

Papaioannou, V

G. Papaioannou, V. Maroof, J. Jerrelind, L. Drugge, Reducing tyre wear emissions of automated articulated vehicles through trajectory planning, Sensors 24 (10) (2024) 3179. Page 12 of 13 Tianxin HuTianxin Hu received his Bachelor’s degree in Vehicle Engineering from Wuhan Uni-...

2024
[2023]

His research interests include modeling and control of aerial and ground robots, motion plan- ning, and multi-agent systems. Thien-Minh Nguyenis currently a Research Assistant Professor at the Centre for Advanced RoboticsTechnologyInnovation(CARTIN),NTU, Singapore.HewastheWall...

2020

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.

[1] [1]

Z. Li, J. Li, Y. Li, An integrated path planning framework for multi- obstacle avoidance of the multi-axle autonomous vehicle with en- hancedsafetyandstability,IEEETransactionsonVehicularTechnol- ogy 73 (5) (2023) 6368–6382

work page 2023

[2] [2]

Beyersdorfer, S

S. Beyersdorfer, S. Wagner, Novel model based path planning for multi-axle steered heavy load vehicles, in: 16th International IEEE Conference on Intelligent Transportation Systems (ITSC 2013), IEEE, 2013, pp. 424–429

work page 2013

[3] [3]

C.Tang,T.Li,T.Li,J.Ma,H.Xiong,All-wheelsteeringpathtracking controlformulti-axlevehiclebasedonadaptivempc,in:Proceedings of the 2024 4th International Conference on Computer, Control and Robotics (ICCCR), IEEE, Shanghai, China, 2024, pp. 183–190

work page 2024

[4] [4]

Y. Zhu, J. Zhu, P. Zhang, Local obstacle avoidance control for multi- axle and multi-steering-mode wheeled robot based on window-zone division strategy, Robotics and Autonomous Systems 183 (2025) 104843

work page 2025

[5] [5]

X.Gan,Z.Huo,W.Li,Dp-a*:Forpathplaningofugvandcontactless delivery, IEEE Transactions on Intelligent Transportation Systems 25 (1) (2023) 907–919

work page 2023

[6] [6]

R. Bai, S. Yuan, K. Li, H. Guo, W.-Y. Yau, L. Xie, Realm: Real-time line-of-sight maintenance in multi-robot navigation with unknown obstacles, in: Proceedings of the IEEE International Conference on Robotics and Automation (ICRA), Atlanta, USA, 2025

work page 2025

[7] [7]

Y. Ma, J. Xu, S. Yuan, T. Zhi, W. Yu, J. Zhou, L. Xie, Mm-lins: a multi-map lidar-inertial system for over-degenerate environments, IEEE Transactions on Intelligent Vehicles - (2024) 1–11

work page 2024

[8] [8]

Nguyen, S

T.-M. Nguyen, S. Yuan, M. Cao, Y. Lyu, T. H. Nguyen, L. Xie, Ntu viral: A visual-inertial-ranging-lidar dataset, from an aerial vehicle viewpoint, The International Journal of Robotics Research 41 (3) (2022) 270–280

work page 2022