Pith. sign in

REVIEW 5 major objections 5 minor 29 references

SMART-Merge Planner: A Safe Merging and Real-Time Motion Planner for Autonomous Highway On-Ramp Merging

T0 review · 5 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

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

Pith's one-line read A lattice-based motion planner with a merge cost, an adaptive collision cost, and a desired-speed heuristic achieves 100% merge success and the shortest merge times in hundreds of simulated highway on-ramp merges.

desk verdict Solid engineering result for forced highway merging, with a genuinely useful cost-shaping combination; the 100% success claim only holds under lane-constrained traffic, which the abstract should say. read the letter →

arxiv 2507.10968 v1 pith:UE2RARZC submitted 2025-07-15 cs.RO cs.SYeess.SY

Toktam Mohammadnejad , Jovin D'sa , Behdad Chalaki , Hossein Nourkhiz Mahjoub , Ehsan Moradi-Pari This is my paper · ORCID
classification cs.ROcs.SYeess.SY
keywords autonomousdrivingmotionplanninglatticeplannerhighwaymergingforcedmergecollisionavoidancedesiredspeedheuristicreal-time
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

Merging onto a highway when the ramp is about to end is a forced, time-limited maneuver. This paper argues that a lattice-based motion planner can handle that maneuver reliably if its cost function is tuned to the merge task. The tuning consists of three additions: a merge cost that pushes the ego vehicle to enter the main lane early, an adaptive collision cost that penalizes time-to-collision and safe-distance violations, and a desired-speed heuristic that sets the target speed from predicted safe distances to surrounding traffic. In 210 simulations spanning eight on-ramp geometries and traffic speeds from 10 to 120 km/h, the planner reports 100% merge success, the shortest average merge time among three compared versions, and smoother accelerations and jerks, all at a 10 Hz update rate on a single CPU core.

What carries the argument

The carrying mechanism is a conformal spatiotemporal lattice: a graph of candidate states whose positions, headings, and curvatures align with the road geometry, with time and velocity sampled at each state. Trajectories are built by connecting states with curvature-continuous splines, and each candidate is scored by a weighted sum of cost terms. Three terms do the paper's work: a lane-centering and merge cost that becomes a high base cost plus a steep slope once the ego vehicle is past the midpoint of the merge lane, encouraging an early lateral move; an adaptive obstacle cost that, in the merge state, combines an inverse-time-to-collision term with exponential safe-distance penalties computed for both the ego and the obstacle at a predicted collision point; and a desired-speed heuristic that derives the ego's target speed from the smallest safety factor among the lead vehicle in the merge lane and the lead and rear vehicles in the main lane. Real-time performance comes from road-aligned sampling, reparameterizing the curvature spline with equally spaced knot points, and restricting acceleration samples based on whether the desired speed is above or below the current speed.

What would settle it

Run the same planner on a closed test track with a human-driven vehicle in the main lane that deliberately changes lanes into the merging car's path once the merge begins; if the ego vehicle collides or touches a boundary, the universal success claim is shown to rest on the no-lane-change assumption.

Watch

Extended reading notes

Core claim

The paper's central claim is that a lattice planner can solve forced highway on-ramp merges safely and quickly, provided the cost function directly encodes the two things that make merging hard: the need to commit to a lane change before the ramp ends, and the need to time the merge relative to the predicted motion of main-lane traffic. The authors compare the full SMART-Merge planner with Planner A, which replaces the desired-speed heuristic with the speed limit, and Planner B, which disables the merge cost. The full planner achieves 100% success on 160 naturalistic cases and 50 dense-traffic cases, while Planner A drops to 88.75% and 76% respectively, and Planner B succeeds 98.75% and 100% but with longer average merge times. The authors attribute the success margin to the desired-speed heuristic and the merge-time margin to the merge cost term, and note that disabling the adaptive collision cost causes frequent collisions.

Load-bearing premise

The load-bearing assumption is that every other vehicle on the road stays in its lane and moves at a constant speed, so the 100% success claim holds only in traffic worlds where no one changes lanes or brakes abruptly.

Editorial extensions

If this is right

  • A standard lattice planner can perform forced merges in real time, at 10 Hz on a single CPU core, without GPU or parallel computation, if the cost terms and sampling are tuned to the merge task.
  • The desired-speed heuristic, rather than the speed limit, is what keeps the planner successful in dense or aggressive traffic: replacing it with the road speed limit drops success from 100% to 88.75% in naturalistic scenarios and to 76% under increasing density.
  • Incentivizing an early merge is what makes merges fast: disabling the merge cost raises average merge time from 13.24 s to 15.96 s in Evaluation A and from 29.29 s to 36.91 s in Evaluation B.
  • The planner produces smoother trajectories than its ablations, with lower maximum longitudinal and lateral acceleration and jerk, implying that the cost adaptations also benefit passenger comfort.
  • The same planner design can be deployed on resource-constrained vehicle computers while keeping a 5-second planning horizon and a 10 Hz replanning rate.

Reading between the lines

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

  • The 100% success rate is established under the paper's stated traffic assumption of no lane changes and constant-speed prediction, so a natural next experiment is to replace the constant-speed model with a probabilistic lane-change or braking model and measure where success drops; this would draw the boundary of the claim.
  • The merge-cost recipe is not merge-specific: any forced gap-acceptance maneuver with a length limit, such as a highway exit, a lane drop, or a roundabout entry, has the same 'must commit before the end' structure, so the same cost terms may transfer with little modification.
  • The desired-speed heuristic chooses the most conservative of three safety factors; a testable extension would be to make the bias toward matching the rear vehicle or the lead vehicle depend on the gap length, which could shorten merge times further in sparse traffic where the conservative choice is unnecessary.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

5 major / 5 minor

Summary. The paper proposes SMART-Merge, a lattice-based motion planner for autonomous highway on-ramp merging. The planner builds on a conformal spatiotemporal lattice, adds a dedicated merge cost term, an adaptive dynamic collision cost, and a desired-speed heuristic, and uses road-geometry-aware sampling to improve kinematic feasibility and run in real time. The method is evaluated in IPG CarMaker on 160 naturalistic on-ramp scenarios from the HOMER dataset and on 50 density-varying scenarios using the MR-IDM traffic model. The reported results show a 100% success rate, the lowest average merge time among two ablated baseline configurations, smoother longitudinal/lateral acceleration and jerk metrics, and a 10 Hz update rate on a single CPU core. The authors acknowledge that the traffic prediction model is lane-constrained and that traffic agents are assumed not to change lanes.

Significance. If the claims hold within the stated simulation environment, the planner offers a practical, interpretable lattice-based solution for forced highway merging, with a sensible ablation study isolating the contributions of the desired-speed heuristic and the merge cost term. The paper's strengths include the use of a high-fidelity simulator, scenario diversity derived from naturalistic data, an explicit single-core real-time implementation in C++/ROS, and attention to comfort metrics such as jerk and lateral acceleration. However, the significance is bounded by the lane-change-free and no-emergency-braking traffic assumption, the absence of any external baseline comparison, and the non-disclosure of all cost weights and safety factors. The headline 100% success and robustness claim therefore overstates the current evidence, even though the underlying algorithmic ideas are promising.

major comments (5)
  1. [II-D, III-A; Abstract] The unqualified 100% success rate and the conclusion that the planner provides a 'reliable and robust solution' are supported only for a restricted traffic class: Section II-D states that 'we assume that the traffic agents do not change lane,' and Section III-A states that the CarMaker and MR-IDM traffic models 'do not account for emergency braking and sudden evasive maneuvers like lane changes.' A main-lane vehicle that changes lanes into or out of the target gap shifts the predicted collision positions in Eq. (18) and invalidates the safe-speed calculation in Eqs. (19)-(20), because those equations assume fixed lane geometry and constant-speed motion. Since discretionary lane changes are common and safety-critical in real on-ramp merging, the abstract's claim goes beyond what the experiments establish. The authors should either soften the abstract and conclusion to state the assumption explicitly, or extend the evaluation to include lane-changing and emergency-braking traffic agents.
  2. [III-B] The note in Evaluation B that 'other road users can influence the situation, e.g., vehicles merging into the main lane or vehicles forced to decelerate in response to their lead vehicle' appears to contradict Section II-D's assumption that traffic agents do not change lane and Section III-A's statement that the traffic models do not account for lane changes. If the traffic actors in Evaluation B were allowed to perform lane changes, the planner was tested outside its stated prediction model and the 100% success claim needs to be reported separately for cases with and without such events. If no lane changes occurred, the sentence is misleading and should be corrected. Please clarify what behavior was actually simulated and whether any lane-change events occurred.
  3. [II-C, Eqs. (6), (14), (16), (20)] The paper states that 'We used the same set of weights for all the scenarios' but never reports the values of the cost weights w_i, the merge cost parameters c and m_merge, the collision cost weights α1 and α2, the safety factors α_lead_merge, α_lead_main, and α_rear_main, or the longitudinal safety constants t_reaction, a_max_acc, and a_max_dec. These parameters are load-bearing: the desired-speed heuristic and the cost-based selection of trajectories determine both the reported merge times and the success rate. Without these values, the results cannot be reproduced or transferred to other scenarios. Please provide a complete table of all parameters and, ideally, a sensitivity analysis over them.
  4. [II-B, III-A] The claim that the planner runs in real time at 10 Hz is not substantiated by measured computation time. Section II-B says 'we interrupt the search process if needed to ensure that the planner runs at a certain frequency,' which means the 10 Hz rate may be enforced by truncating the search rather than by completing the planned computation. No planning-time statistics, truncation rates, or information about how often the search was cut short are reported. The authors should report wall-clock planning times (mean, maximum, percentiles) and state how often and under what conditions the search was interrupted, since truncation can affect trajectory quality and success.
  5. [II-C Eq. (17) vs. II-D Eq. (20)] There is an internal inconsistency in the safety-distance definitions. In Eq. (17), the safe following distance d_safe(s) for the collision cost includes the reaction distance plus the braking distance needed to match the lead vehicle's speed. In Eq. (20), however, the desired-speed heuristic requires only the reaction distance (e.g., e_d_lead_main = α_lead_main v), with no braking-distance term. This means the desired speed may be set using a less conservative gap criterion than the one used for collision checking, which is particularly relevant when the ego vehicle is faster than the lead vehicle. The authors should justify why the reaction-distance-only criterion is safe in the desired-speed heuristic, or align Eq. (20) with Eq. (17).
minor comments (5)
  1. [II-D, Fig. 1] The terms 'hard nose' and 'soft nose' are used to define the behavioral states but are never formally defined. Please add brief definitions or a diagram annotation explaining these geometric reference points.
  2. [II-C, Eq. (18)] The computation of the collision positions s_ego and s_obs_i and the arrival times T_ego and T_obs_i is not described. Please specify how the overlapping bounding-box point is found and how T_ego and T_obs_i are computed from the ego trajectory and the constant-speed traffic prediction.
  3. [III-B, Tables II and III] Tables II and III report only mean values for merge time, accelerations, and jerks. Since success is binary and the reported means can hide outliers, please include standard deviations, percentile values, or per-density-bin results, especially for Evaluation B where traffic density is varied.
  4. [III-A] There is a typo in the phrase 'designed sepcifically for simulating on-ramp merge reactions' — 'sepcifically' should be 'specifically'.
  5. [III-C] The baselines are only two ablated versions of the proposed planner. While this is an appropriate ablation design, the abstract's phrase 'shortest amount of time compared with the baselines' should not be read as a comparison with prior or alternative planners. Adding an external baseline, such as a conventional Frenet or MPC-based merge planner, would strengthen the claim.

Circularity Check

0 steps flagged · score 1.0 of 10

No derivation-level circularity; self-cited simulation model and ablation baselines affect external validity, not the internal derivation chain.

full rationale

The paper's derivation chain is self-contained. The cost function in Eq. (6) is a weighted sum of explicitly defined penalties (Eqs. 7-18), and the desired speed heuristic in Eqs. (19)-(20) is derived from kinematic safe-following and reaction-distance requirements, not from the reported 100% success rate. The success metric is an empirical simulation outcome computed after planning, not a term in the optimization objective, and the ablation baselines (Planner A and Planner B) are used to measure the marginal contribution of the proposed heuristic and merge cost; their failures show that success is not forced by the cost definitions alone. The cited MR-IDM traffic model [29] is authored by overlapping researchers, but it is used as an external simulation environment for traffic reactions rather than as a premise in the planner's derivation; it does not encode the planner's cost terms, search procedure, or success criterion, so its use does not make the planner's success true by construction. The acknowledged lane-constrained constant-velocity prediction assumption (Section II-D and the conclusion) is an external-validity limitation, not a circular step: it weakens the generality of the 100% claim but does not reduce the derivation to its own inputs. No fitted parameter is renamed as a prediction, and no uniqueness theorem or ansatz is imported via self-citation to force the claimed result.

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

The paper's results depend on a set of hand-selected cost weights and safety factors that are not reported, and on traffic behavior assumptions (lane keeping, no lane changes, constant-speed prediction) that are acknowledged as limitations. These do not invalidate the empirical demonstration but limit generalization.

free parameters (5)
  • Cost weights w_i (Eq. 6) = not reported; hand-tuned
    All trajectory selection depends on the weighted sum J; the paper states the same weights are used for all scenarios but gives no values or tuning procedure.
  • Safety factors alpha_i in desired speed heuristic (Eq. 20) = not reported
    Determine the target speed relative to lead and rear vehicles; central to the claimed success-rate improvement.
  • Merge cost parameters c and mmerge (Eq. 14) = not reported
    Control how aggressively the planner leaves the merge lane; the paper only says c must be high enough and mmerge must be at least m.
  • Collision cost weights alpha1 and alpha2 (Eqs. 16, 18) = not reported
    Balance safety terms against progress; values are never given.
  • Longitudinal safety constants t_reaction, a_max_dec, a_max_acc (Eqs. 17, 19) = not reported
    Set safe-following distances and predictive speed adjustments; treated as design constants but values are absent.
assumptions (5)
  • domain assumption Kinematic bicycle model (Eq. 1) adequately represents ego vehicle motion.
    Standard for planning but ignores tire slip, load transfer, and actuator dynamics.
  • domain assumption Traffic vehicles follow their lane and do not change lane (Section II-D).
    Used for collision prediction and desired speed; authors acknowledge this as future work.
  • domain assumption Traffic future motion is predicted with a constant-speed model (Eq. 19).
    Underlies time-to-collision, collision positions, and desired speed; no uncertainty or prediction error is modeled.
  • ad hoc to paper One fixed set of cost weights generalizes across all tested on-ramp geometries, densities, and traffic behaviors (Section II-C).
    No evidence is given that a single weight vector is sufficient; overfitting risk is unquantified.
  • domain assumption CarMaker traffic models (car-following and MR-IDM) reproduce real merge reactions (Section III-A).
    The paper notes these models exclude emergency braking and sudden lane changes, limiting transfer.

how reviews work

0 comments
Cite this review

Pith. "Pith review of SMART-Merge Planner: A Safe Merging and Real-Time Motion Planner for Autonomous Highway On-Ramp Merging." pith.science (2026). https://pith.science/paper/UE2RARZC

@misc{pith2026250710968,
  author       = {Pith},
  title        = {Pith review of: SMART-Merge Planner: A Safe Merging and Real-Time Motion Planner for Autonomous Highway On-Ramp Merging},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/UE2RARZC}},
  note         = {Machine review of arXiv:2507.10968}
}
read the original abstract

Merging onto a highway is a complex driving task that requires identifying a safe gap, adjusting speed, often interactions to create a merging gap, and completing the merge maneuver within a limited time window while maintaining safety and driving comfort. In this paper, we introduce a Safe Merging and Real-Time Merge (SMART-Merge) planner, a lattice-based motion planner designed to facilitate safe and comfortable forced merging. By deliberately adapting cost terms to the unique challenges of forced merging and introducing a desired speed heuristic, SMART-Merge planner enables the ego vehicle to merge successfully while minimizing the merge time. We verify the efficiency and effectiveness of the proposed merge planner through high-fidelity CarMaker simulations on hundreds of highway merge scenarios. Our proposed planner achieves the success rate of 100% as well as completes the merge maneuver in the shortest amount of time compared with the baselines, demonstrating our planner's capability to handle complex forced merge tasks and provide a reliable and robust solution for autonomous highway merge. The simulation result videos are available at https://sites.google.com/view/smart-merge-planner/home.

Figures

Figures reproduced from arXiv: 2507.10968 by the authors.

Figure 1
Figure 1. Conformal state lattice for autonomous highway [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Lane centering/Merge cost function Ccenter (s). where sf is the path length and a is the acceleration. To ensure that the ego vehicle stays within the desired speed vd, we penalize the deviation from the desired speed Jvelocity = Z sf 0 (v (s) − vd) 2 ds. (10) The inconsistency between successive motion plans can result in controller instability. Therefore, to ensure the consis￾tency of the planned trajectory, we ad… view at source ↗
Figure 3
Figure 3. Illustration of the dynamic obstacle collision cost components in the merge state for yielding and overtaking cases. [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: Illustration of the traffic vehicles of high interest on the road for the autonomous highway merge scenario. The traffic [PITH_FULL_IMAGE:figures/full_fig_p005_4.png]
Figure 6
Figure 6. Figure 6: Snapshots of on-vehicle validation of the SMART [PITH_FULL_IMAGE:figures/full_fig_p006_6.png]
Figure 7
Figure 7. Figure 7: A snapshot of a test case with a dense traffic in [PITH_FULL_IMAGE:figures/full_fig_p007_7.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

29 extracted references · 28 canonical work pages

  1. [1]

    On social interactions of merging behaviors at highway on-ramps in congested traffic,

    H. Wang, W. Wang, S. Yuan, X. Li, and L. Sun, “On social interactions of merging behaviors at highway on-ramps in congested traffic,” IEEE Transactions on Intelligent Transportation Systems, vol. 23, no. 8, pp. 11 237–11 248, 2021

    work page 2021

  2. [2]

    Should I stay or should I go? Evidence accumulation drives decision making in human drivers,

    A. Zgonnikov, D. Abbink, and G. Markkula, “Should I stay or should I go? Evidence accumulation drives decision making in human drivers,” PsyArXiv, 2020

    work page 2020

  3. [3]

    Interaction aware trajectory planning for merge scenarios in congested traffic situations,

    N. Evestedt, E. Ward, J. Folkesson, and D. Axehill, “Interaction aware trajectory planning for merge scenarios in congested traffic situations,” in 2016 IEEE 19th International Conference on Intelligent Transportation Systems (ITSC) . IEEE, 2016, pp. 465–472

    work page 2016

  4. [4]

    Mptree: A sampling- based vehicle motion planner for real-time obstacle avoidance,

    M. Piazza, M. Piccinini, S. Taddei, and F. Biral, “Mptree: A sampling- based vehicle motion planner for real-time obstacle avoidance,” IFAC- PapersOnLine, vol. 58, no. 10, pp. 146–153, 2024, 17th IFAC Sym- posium on Control of Transportation Systems CTS 2024

    work page 2024

  5. [5]

    Sampling-based optimal trajectory generation for autonomous vehicles using reachable sets,

    G. W ¨ursching and M. Althoff, “Sampling-based optimal trajectory generation for autonomous vehicles using reachable sets,” in 2021 IEEE International Intelligent Transportation Systems Conference (ITSC). IEEE, 2021, pp. 828–835

    work page 2021

  6. [6]

    Reinforcement learning with iterative reasoning for merging in dense traffic,

    M. Bouton, A. Nakhaei, D. Isele, K. Fujimura, and M. J. Kochenderfer, “Reinforcement learning with iterative reasoning for merging in dense traffic,” in 2020 IEEE 23rd International Conference on Intelligent Transportation Systems (ITSC) . IEEE, 2020, pp. 1–6

    work page 2020

  7. [7]

    Dq-gat: Towards safe and efficient autonomous driving with deep q-learning and graph attention networks,

    P. Cai, H. Wang, Y . Sun, and M. Liu, “Dq-gat: Towards safe and efficient autonomous driving with deep q-learning and graph attention networks,” IEEE Transactions on Intelligent Transportation Systems , vol. 23, no. 11, pp. 21 102–21 112, 2022

    work page 2022

  8. [8]

    Developing a merge lane change decision policy for autonomous vehicles by deep reinforce- ment learning,

    B. Fan, Y . Zhou, and H. Mahmassani, “Developing a merge lane change decision policy for autonomous vehicles by deep reinforce- ment learning,” in 2021 IEEE International Intelligent Transportation Systems Conference (ITSC) . IEEE, 2021, pp. 963–968

    work page 2021

Show all 29 references
  1. [9]

    A deep reinforcement learning-based approach for autonomous driving in highway on-ramp merge,

    H. Wang, S. Yuan, M. Guo, X. Li, and W. Lan, “A deep reinforcement learning-based approach for autonomous driving in highway on-ramp merge,” Proceedings of the Institution of Mechanical Engineers, Part D: Journal of Automobile Engineering, vol. 235, pp. 2726–2739, 2021

    2021

  2. [10]

    Interaction-aware trajectory prediction and planning for autonomous vehicles in forced merge scenarios,

    K. Liu, N. Li, H. E. Tseng, I. Kolmanovsky, and A. Girard, “Interaction-aware trajectory prediction and planning for autonomous vehicles in forced merge scenarios,” IEEE Transactions on Intelligent Transportation Systems, vol. 24, no. 1, pp. 474–488, 2022

    2022

  3. [11]

    Model pre- dictive control for lane merging automation with recursive feasibility guarantees,

    M. Geurts, A. Katriniok, E. Silvas, and W. Heemels, “Model pre- dictive control for lane merging automation with recursive feasibility guarantees,” IFAC-PapersOnLine, vol. 56, no. 2, pp. 4858–4864, 2023

    2023

  4. [12]

    Active learning with dual model predictive path- integral control for interaction-aware autonomous highway on-ramp merging,

    J. Knaup, J. D’sa, B. Chalaki, T. Naes, H. N. Mahjoub, E. Moradi- Pari, and P. Tsiotras, “Active learning with dual model predictive path- integral control for interaction-aware autonomous highway on-ramp merging,” arXiv preprint arXiv:2310.07840 , 2023

    arXiv

  5. [13]

    Safe trajectory genera- tion for complex urban environments using spatio-temporal semantic corridor,

    W. Ding, L. Zhang, J. Chen, and S. Shen, “Safe trajectory genera- tion for complex urban environments using spatio-temporal semantic corridor,” IEEE Robotics and Automation Letters , vol. 4, no. 3, pp. 2997–3004, 2019

    2019

  6. [14]

    Multi-profile quadratic programming (mpqp) for optimal gap selection and speed planning of autonomous driving,

    A. M. Anon, S. Bae, M. Saroya, and D. Isele, “Multi-profile quadratic programming (mpqp) for optimal gap selection and speed planning of autonomous driving,” arXiv preprint arXiv:2401.06305 , 2024

    2024 arXiv

  7. [15]

    A review of motion planning for highway autonomous driving,

    L. Claussmann, M. Revilloud, D. Gruyer, and S. Glaser, “A review of motion planning for highway autonomous driving,” IEEE Transactions on Intelligent Transportation Systems , vol. 21, pp. 1826–1848, 2019

    2019

  8. [16]

    A survey of motion planning and control techniques for self-driving urban vehicles,

    B. Paden, M. ˇC´ap, S. Z. Yong, D. Yershov, and E. Frazzoli, “A survey of motion planning and control techniques for self-driving urban vehicles,” IEEE Transactions on intelligent vehicles , vol. 1, no. 1, pp. 33–55, 2016

    2016

  9. [17]

    Autonomous driving in urban environments: Boss and the urban challenge,

    C. Urmson, J. Anhalt, D. Bagnell, C. Baker, R. Bittner, M. Clark, J. Dolan, D. Duggins, T. Galatali, C. Geyer, et al. , “Autonomous driving in urban environments: Boss and the urban challenge,” Journal of field Robotics , vol. 25, no. 8, pp. 425–466, 2008

    2008

  10. [18]

    Motion planning for autonomous driving with a conformal spatiotemporal lattice,

    M. McNaughton, C. Urmson, J. M. Dolan, and J.-W. Lee, “Motion planning for autonomous driving with a conformal spatiotemporal lattice,” in 2011 IEEE International Conference on Robotics and Automation. IEEE, 2011, pp. 4889–4895

    2011

  11. [19]

    Motion planning under uncertainty for on-road autonomous driving,

    W. Xu, J. Pan, J. Wei, and J. M. Dolan, “Motion planning under uncertainty for on-road autonomous driving,” in IEEE International Conference on Robotics and Automation , 2014, pp. 2507–2512

    2014

  12. [20]

    Fast, continuous state path smoothing to improve navigation accuracy,

    H. Andreasson, J. Saarinen, M. Cirillo, T. Stoyanov, and A. J. Lilien- thal, “Fast, continuous state path smoothing to improve navigation accuracy,” in 2015 IEEE International Conference on Robotics and Automation (ICRA). IEEE, 2015, pp. 662–669

    2015

  13. [21]

    Learning a lattice planner control set for autonomous vehicles,

    R. De Iaco, S. L. Smith, and K. Czarnecki, “Learning a lattice planner control set for autonomous vehicles,” in 2019 IEEE Intelligent Vehicles Symposium (IV). IEEE, 2019, pp. 549–556

    2019

  14. [22]

    Merit-based motion planning for autonomous vehicles in urban scenarios,

    J. Medina-Lee, A. Artu ˜nedo, J. Godoy, and J. Villagra, “Merit-based motion planning for autonomous vehicles in urban scenarios,” Sensors, vol. 21, no. 11, p. 3755, 2021

    2021

  15. [23]

    Motion planning for autonomous vehicles in urban scenar- ios: a sequential optimization approach,

    W. Xu, “Motion planning for autonomous vehicles in urban scenar- ios: a sequential optimization approach,” Ph.D. dissertation, Carnegie Mellon University, 2021

    2021

  16. [24]

    Modeling the lane-change reactions to merging vehicles for highway on-ramp simulations,

    D. Holley, J. D’sa, H. N. Mahjoub, G. Ali, T. Naes, E. Moradi-Pari, and P. S. Kallepalli, “Modeling the lane-change reactions to merging vehicles for highway on-ramp simulations,” in 2024 IEEE Intelligent Vehicles Symposium (IV), 2024, pp. 651–658

    2024

  17. [25]

    Development of a new integrated local trajectory planning and tracking control framework for autonomous ground vehicles,

    X. Li, Z. Sun, D. Cao, D. Liu, and H. He, “Development of a new integrated local trajectory planning and tracking control framework for autonomous ground vehicles,” Mechanical Systems and Signal Processing, vol. 87, pp. 118–137, 2017

    2017

  18. [26]

    Reactive nonholonomic trajectory generation via parametric optimal control,

    A. Kelly and B. Nagy, “Reactive nonholonomic trajectory generation via parametric optimal control,” The International Journal of Robotics Research, 2003

    2003

  19. [27]

    Trajectory planning with comfort and safety in dynamic traffic scenarios for autonomous driving,

    J. Zhang, Z. Jian, J. Fu, Z. Nan, J. Xin, and N. Zheng, “Trajectory planning with comfort and safety in dynamic traffic scenarios for autonomous driving,” in 2021 IEEE Intelligent Vehicles Symposium Workshops (IV Workshops), 2021, pp. 342–349

    2021

  20. [28]

    Reference manual version 9.0. 1,

    C. IPG, “Reference manual version 9.0. 1,” IPG Automotive GmbH: Karlsruhe, Germany, 2021

    2021

  21. [29]

    Mr-idm - merge reactive intelligent driver model: Towards en- hancing laterally aware car-following models,

    D. Holley, J. D’sa, H. N. Mahjoub, G. Ali, B. Chalaki, and E. Moradi- Pari, “Mr-idm - merge reactive intelligent driver model: Towards en- hancing laterally aware car-following models,” in 2023 IEEE 26th In- ternational Conference on Intelligent Transportation Systems (ITSC) ,...

    2023

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.