pith. sign in

arxiv: 2402.01370 · v4 · submitted 2024-02-02 · 💻 cs.RO · cs.SY· eess.SY

CC-VPSTO: Chance-Constrained Via-Point-Based Stochastic Trajectory Optimisation for Online Robot Motion Planning under Uncertainty

Lara Bruderm\"uller , Guillaume Berger , Julius Jankowski , Raunak Bhattacharyya , Rapha\"el Jungers , Nick Hawes This is my paper

Pith reviewed 2026-05-24 03:59 UTC · model grok-4.3

classification 💻 cs.RO cs.SYeess.SY
keywords chance-constrained optimizationstochastic trajectory optimizationrobot motion planninguncertaintyMonte Carlo samplingmodel predictive controlonline planningvia-point optimization
0
0 comments X

The pith

CC-VPSTO turns chance-constrained robot trajectory planning into a Monte Carlo-sampled deterministic problem solvable online without assumptions on uncertainty distributions.

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

The paper presents CC-VPSTO as a way for robots to plan motions that respect inequality constraints with a user-specified probability even when uncertainty from sensors or models is present. It converts the generally intractable probabilistic problem into a deterministic surrogate via Monte Carlo samples that a gradient-free optimizer can solve quickly enough for receding-horizon use. The framework keeps solutions valid by quantifying sampling error and adding padding, while remaining general across uncertainty types, dynamics, costs, and constraint forms. Integration into model-predictive control yields reactive behavior that avoids both unsafe and excessively cautious trajectories. Experiments on a simulated and physical Franka Emika arm illustrate the balance between constraint satisfaction and task efficiency.

Core claim

CC-VPSTO formulates stochastic control as a chance-constrained optimisation problem, approximates it with a deterministic Monte Carlo surrogate solved by gradient-free optimisation, and embeds the result in a receding-horizon MPC loop so that trajectories satisfy constraints with high probability while remaining task-efficient and suitable for online execution.

What carries the argument

The Monte Carlo sampling surrogate for the chance constraints together with padding strategies that control approximation error and guarantee validity of the surrogate solution.

If this is right

  • Reactive task-efficient control is possible in receding-horizon MPC under uncertainty.
  • The method applies without requiring specific forms for uncertainty, dynamics, cost, or inequality constraints.
  • Sample-efficient constraint approximation supports real-time execution.
  • Conditions for surrogate validity can be maintained during online optimisation.
  • The approach demonstrates validity and efficiency both in simulation and on physical hardware.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The sampling-based surrogate could be paired with learned uncertainty models to further reduce the number of required samples.
  • The same padding analysis might apply to other gradient-free solvers or to distributed multi-agent settings.
  • Hardware tests with injected sensor noise distributions could directly measure how padding margins trade off against task speed.

Load-bearing premise

Monte Carlo sampling combined with padding strategies produces a deterministic surrogate whose solutions remain feasible for the original probabilistic constraints when used in fast online replanning.

What would settle it

A set of closed-loop trials that counts the actual fraction of constraint violations and checks whether the observed rate stays at or below the allowed risk level for the chosen sample count and padding margin.

Figures

Figures reproduced from arXiv: 2402.01370 by Guillaume Berger, Julius Jankowski, Lara Bruderm\"uller, Nick Hawes, Rapha\"el Jungers, Raunak Bhattacharyya.

Figure 1
Figure 1. Figure 1: Real-robot experiment. The robot is tasked to move its ball-shaped end effector from a start point on one side to a goal point on the other side of a conveyor belt (indicated by the yellow balls in the right simulation view). Meanwhile, the ball end effector has to avoid the box obstacle on a moving conveyor belt which is controlled according to a stochastic policy. Depending on the anticipated box movemen… view at source ↗
Figure 2
Figure 2. Figure 2: CDF of the binomial distribution for different values of [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Graph of the penalty function used in CC-VPSTO. When observing [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figure 5
Figure 5. Figure 5: Offline Planning Experiment. We show Neval = 104 red circles for the uncertain obstacle position and the mean trajectory for the two local optima from CC-VPSTO, which used N = 100 particles in the optimisation, for varying values of η across Nexp = 105 experiments. The blue circle shows the robot’s radius and starting position. the Appendix. We also visualize the mean trajectories for the two local optima … view at source ↗
Figure 4
Figure 4. Figure 4: Offline Planning Experiment. We evaluate the heuristic ηbinom for different values of η and numbers of particles N (100, 1000) by running CC-VPSTO Nexp = 105 times and evaluating the solutions on a set of Neval = 104 new unseen samples. We compare the heuristic ηbinom to the Rademacher-complexity bound ηrad. We also show the mean collision probability ηˆavg and the (1 − β)-percentile of the collision proba… view at source ↗
Figure 6
Figure 6. Figure 6: Overview of the environments used for the MPC experiments. In each column, one per environment configuration, we show three examples of how the obstacle rollouts can look like, given the same environment configuration. The initial obstacle positions and their radii are shown by the blue circles. The smaller circles along the trajectories indicate the ground truth MPC updates, whilst the opaque trajectories… view at source ↗
Figure 7
Figure 7. Figure 7: Simulation: MPC experiments. Evaluating motion duration, success rate, collision rate and the minimum distance to obstacles across 1000 experiments on 3 different environments. One experiment corresponds to running online-CC-VPSTO until reaching the goal or until a maximum number of 100 MPC steps is reached. Goal and start location remain fixed across all experiments and environments, whilst the obstacle t… view at source ↗
Figure 8
Figure 8. Figure 8: Robot experiment setup. The robot task is to move from one side to the other side of the conveyor belt while assuring that the probability of colliding with the box obstacle is below a user-defined threshold η. The motion of the box obstacle is stochastic, as the conveyor belt is actuated with constant velocity, but the rate of direction change follows an exponential distribution. robot is not allowed to s… view at source ↗
Figure 9
Figure 9. Figure 9: We do not compare our approach to “ML-VPSTO” as for the given stochastic model, the mean over the exponential distribution is not meaningful. However, we include η = 0.0 as a baseline, which corresponds to a VPSTO approach with 6Note we ignore measurement noise in this setup [PITH_FULL_IMAGE:figures/full_fig_p012_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: Analysis of the binomial distribution with [PITH_FULL_IMAGE:figures/full_fig_p015_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: Illustration of the constraint h1(x) ≥ h2(δ) to avoid collision between a stochastic obstacle and the robot. In the figure, p¯ is an arbitrarily chosen nominal point; h2(δ) is the maximal distance between p¯ and any point of the obstacle in scenario δ; and h1(x) is the minimal distance between p¯ and any point of the robot in configuration x. The chance constraint in the optimisation of x is that Pδ∼∆[h1(… view at source ↗
read the original abstract

Reliable robot autonomy hinges on decision-making systems that account for uncertainty without imposing overly conservative restrictions on the robot's action space. We introduce Chance-Constrained Via-Point-Based Stochastic Trajectory Optimisation (CC-VPSTO), a real-time capable framework for generating task-efficient robot trajectories that satisfy constraints with high probability by formulating stochastic control as a chance-constrained optimisation problem. Since such problems are generally intractable, we propose a deterministic surrogate formulation based on Monte Carlo sampling, solved efficiently with gradient-free optimisation. To address bias in na\"ive sampling approaches, we quantify approximation error and introduce padding strategies to improve reliability. We focus on three challenges: (i) sample-efficient constraint approximation, (ii) conditions for surrogate solution validity, and (iii) online optimisation. Integrated into a receding-horizon MPC framework, CC-VPSTO enables reactive, task-efficient control under uncertainty, balancing constraint satisfaction and performance in a principled manner. The strengths of our approach lie in its generality, i.e. no assumptions on the underlying uncertainty distribution, system dynamics, cost function, or the form of inequality constraints; and its applicability to online robot motion planning. We demonstrate the validity and efficiency of our approach in both simulation and on a Franka Emika robot.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 0 minor

Summary. The manuscript introduces CC-VPSTO, a real-time framework for chance-constrained via-point-based stochastic trajectory optimization in online robot motion planning under uncertainty. It formulates stochastic control as a chance-constrained problem, derives a deterministic surrogate via Monte Carlo sampling solved by gradient-free optimization, quantifies approximation error with padding strategies for reliability, and embeds the method in a receding-horizon MPC loop. The approach claims generality with no assumptions on uncertainty distributions, dynamics, costs, or inequality constraints, and reports validation in simulation plus hardware experiments on a Franka Emika robot.

Significance. If the central claims on surrogate validity and online performance hold, the work would provide a distribution-free, non-conservative method for balancing task efficiency and probabilistic constraint satisfaction in uncertain robotic settings. The explicit focus on conditions for surrogate validity and integration with MPC could support broader adoption in real-time autonomy applications.

major comments (2)
  1. [Abstract] Abstract: the description of the Monte Carlo surrogate, error quantification, and padding strategies is high-level only, with no explicit derivation, error bounds, or conditions for validity provided; this is load-bearing for the central claim that the deterministic proxy reliably solves the original chance-constrained problem.
  2. [Abstract] The manuscript states that approximation error is quantified and padding improves reliability, yet provides no concrete bounds, sample-size analysis, or counter-example checks that would confirm the surrogate remains valid under the stated generality (no assumptions on distributions or dynamics).

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their detailed feedback on our manuscript. We address each major comment below, focusing on the concerns regarding the level of detail in the abstract and the supporting analysis for the Monte Carlo surrogate.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the description of the Monte Carlo surrogate, error quantification, and padding strategies is high-level only, with no explicit derivation, error bounds, or conditions for validity provided; this is load-bearing for the central claim that the deterministic proxy reliably solves the original chance-constrained problem.

    Authors: The abstract is intentionally concise to provide an overview within typical length constraints. The explicit derivation of the Monte Carlo surrogate, error quantification through padding, and conditions for surrogate validity (addressing challenge (ii) in the abstract) are developed in detail in Section III of the manuscript. This includes the mathematical formulation showing how the deterministic proxy approximates the chance-constrained problem while maintaining the claimed generality. We can revise the abstract to include a brief reference to these conditions and the relevant section if the editor prefers. revision: partial

  2. Referee: [Abstract] The manuscript states that approximation error is quantified and padding improves reliability, yet provides no concrete bounds, sample-size analysis, or counter-example checks that would confirm the surrogate remains valid under the stated generality (no assumptions on distributions or dynamics).

    Authors: The manuscript quantifies the approximation error and introduces padding in the main body (Section III), with the analysis designed to hold without assumptions on distributions or dynamics via Monte Carlo sampling. However, we agree that additional concrete sample-size analysis and counter-example validation would better substantiate the claims. We will incorporate these elements, such as guidelines on required sample sizes for desired reliability levels and further simulation checks across varied uncertainty types, in a revised version. revision: yes

Circularity Check

0 steps flagged

No significant circularity

full rationale

The paper presents CC-VPSTO as a Monte Carlo-based deterministic surrogate for chance-constrained trajectory optimization, integrated into receding-horizon MPC. The abstract and description explicitly address approximation error quantification and padding strategies as external safeguards for surrogate validity, with no equations or steps shown that reduce a claimed prediction or uniqueness result to a fitted input or self-citation by construction. The generality claim rests on distribution-free sampling properties, which are standard and externally grounded rather than internally defined. No load-bearing derivation collapses to its own inputs; the approach is self-contained against external benchmarks in optimization and sampling methods.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract provides no information on specific parameters, axioms or entities; insufficient detail for ledger.

pith-pipeline@v0.9.0 · 5788 in / 1062 out tokens · 23524 ms · 2026-05-24T03:59:05.638770+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

40 extracted references · 40 canonical work pages · 2 internal anchors

  1. [1]

    Pr ´ekopa, Stochastic programming

    A. Pr ´ekopa, Stochastic programming. Springer Science & Business Media, 2013, vol. 324

    work page 2013

  2. [2]

    Chance constrained motion planning for high-dimensional robots,

    S. Dai, S. Schaffert, A. Jasour, A. Hofmann, and B. Williams, “Chance constrained motion planning for high-dimensional robots,” in 2019 International Conference on Robotics and Automation (ICRA) . IEEE, 2019, pp. 8805–8811

    work page 2019

  3. [3]

    On the road between robust optimization and the scenario approach for chance constrained optimiza- tion problems,

    K. Margellos, P. Goulart, and J. Lygeros, “On the road between robust optimization and the scenario approach for chance constrained optimiza- tion problems,” IEEE Transactions on Automatic Control, vol. 59, no. 8, pp. 2258–2263, 2014

    work page 2014

  4. [4]

    The scenario approach for stochastic model predictive control with bounds on closed- loop constraint violations,

    G. Schildbach, L. Fagiano, C. Frei, and M. Morari, “The scenario approach for stochastic model predictive control with bounds on closed- loop constraint violations,” Automatica, vol. 50, no. 12, pp. 3009–3018, 2014

    work page 2014

  5. [5]

    A proba- bilistic particle-control approximation of chance-constrained stochastic predictive control,

    L. Blackmore, M. Ono, A. Bektassov, and B. C. Williams, “A proba- bilistic particle-control approximation of chance-constrained stochastic predictive control,” IEEE Transactions on Robotics , vol. 26, no. 3, pp. 502–517, 2010

    work page 2010

  6. [6]

    VP-STO: Via-point-based stochastic trajectory optimization for reactive robot behavior,

    J. Jankowski, L. Bruderm ¨uller, N. Hawes, and S. Calinon, “VP-STO: Via-point-based stochastic trajectory optimization for reactive robot behavior,” in 2023 IEEE International Conference on Robotics and Automation (ICRA). IEEE, 2023, pp. 10 125–10 131

    work page 2023

  7. [7]

    Stochastic model predictive control—how does it work?

    T. A. N. Heirung, J. A. Paulson, J. O’Leary, and A. Mesbah, “Stochastic model predictive control—how does it work?” Computers & Chemical Engineering, vol. 114, pp. 158–170, 2018

    work page 2018

  8. [8]

    Stochastic model predictive control: An overview and perspectives for future research,

    A. Mesbah, “Stochastic model predictive control: An overview and perspectives for future research,” IEEE Control Systems Magazine , vol. 36, no. 6, pp. 30–44, 2016

    work page 2016

  9. [9]

    Scenario- based motion planning with bounded probability of collision,

    O. de Groot, L. Ferranti, D. Gavrila, and J. Alonso-Mora, “Scenario- based motion planning with bounded probability of collision,” arXiv preprint arXiv:2307.01070, 2023

    work page arXiv 2023

  10. [10]

    Evaluating Trajectory Collision Probability through Adaptive Importance Sampling for Safe Motion Planning

    E. Schmerling and M. Pavone, “Evaluating trajectory collision proba- bility through adaptive importance sampling for safe motion planning,” arXiv preprint arXiv:1609.05399 , 2016

    work page internal anchor Pith review Pith/arXiv arXiv 2016

  11. [11]

    A probabilistic particle control approach to optimal, robust predictive control,

    L. Blackmore, “A probabilistic particle control approach to optimal, robust predictive control,” in AIAA Guidance, Navigation, and Control Conference and Exhibit , 2006, p. 6240

    work page 2006

  12. [12]

    The scenario approach to robust control design,

    G. C. Calafiore and M. C. Campi, “The scenario approach to robust control design,” IEEE Transactions on automatic control, vol. 51, no. 5, pp. 742–753, 2006

    work page 2006

  13. [13]

    Random convex programs,

    G. C. Calafiore, “Random convex programs,” SIAM Journal on Opti- mization, vol. 20, no. 6, pp. 3427–3464, 2010

    work page 2010

  14. [14]

    Chance-constrained quasi- convex optimization with application to data-driven switched systems control,

    G. O. Berger, R. M. Jungers, and Z. Wang, “Chance-constrained quasi- convex optimization with application to data-driven switched systems control,” in Learning for Dynamics and Control . PMLR, 2021, pp. 571–583

    work page 2021

  15. [15]

    Moment state dynamical sys- tems for nonlinear chance-constrained motion planning,

    A. Wang, A. Jasour, and B. Williams, “Moment state dynamical sys- tems for nonlinear chance-constrained motion planning,” arXiv preprint arXiv:2003.10379, 2020

    work page arXiv 2003

  16. [16]

    Chance constrained stochastic optimal control based on sample statistics with almost surely probabilistic guarantees,

    S. Priore and M. Oishi, “Chance constrained stochastic optimal control based on sample statistics with almost surely probabilistic guarantees,” arXiv preprint arXiv:2303.16981 , 2023

    work page arXiv 2023

  17. [17]

    A probabilistic approach to optimal robust path planning with obstacles,

    L. Blackmore, H. Li, and B. Williams, “A probabilistic approach to optimal robust path planning with obstacles,” in 2006 American Control Conference. IEEE, 2006, pp. 7–pp

    work page 2006

  18. [18]

    Stochastic model predictive control with discounted probabilistic constraints,

    S. Yan, P. Goulart, and M. Cannon, “Stochastic model predictive control with discounted probabilistic constraints,” in 2018 European Control Conference (ECC). IEEE, 2018, pp. 1003–1008

    work page 2018

  19. [19]

    Convex approximations of chance constrained programs,

    A. Nemirovski and A. Shapiro, “Convex approximations of chance constrained programs,” SIAM Journal on Optimization , vol. 17, no. 4, pp. 969–996, 2007

    work page 2007

  20. [20]

    Semidefinite programming for chance constrained optimization over semialgebraic sets,

    A. M. Jasour, N. S. Aybat, and C. M. Lagoa, “Semidefinite programming for chance constrained optimization over semialgebraic sets,” SIAM Journal on Optimization , vol. 25, no. 3, pp. 1411–1440, 2015

    work page 2015

  21. [21]

    Differential dynamic programming with non- linear safety constraints under system uncertainties,

    G. Alcan and V . Kyrki, “Differential dynamic programming with non- linear safety constraints under system uncertainties,” IEEE Robotics and Automation Letters, vol. 7, no. 2, pp. 1760–1767, 2022

    work page 2022

  22. [22]

    Iterative risk allocation: A new approach to robust model predictive control with a joint chance constraint,

    M. Ono and B. C. Williams, “Iterative risk allocation: A new approach to robust model predictive control with a joint chance constraint,” in 2008 47th IEEE Conference on Decision and Control . IEEE, 2008, pp. 3427–3432

    work page 2008

  23. [23]

    Computationally efficient robust mpc using optimized constraint tightening,

    A. Parsi, P. Anagnostaras, A. Iannelli, and R. S. Smith, “Computationally efficient robust mpc using optimized constraint tightening,” in 2022 IEEE 61st Conference on Decision and Control (CDC) . IEEE, 2022, pp. 1770–1775

    work page 2022

  24. [24]

    Safe motion planning for imprecise robotic manipulators by minimizing probability of collision,

    W. Sun, L. G. Torres, J. Van Den Berg, and R. Alterovitz, “Safe motion planning for imprecise robotic manipulators by minimizing probability of collision,” in Robotics Research: The 16th International Symposium ISRR. Springer, 2016, pp. 685–701

    work page 2016

  25. [25]

    Lqg-mp: Optimized path planning for robots with motion uncertainty and imperfect state information,

    J. Van Den Berg, P. Abbeel, and K. Goldberg, “Lqg-mp: Optimized path planning for robots with motion uncertainty and imperfect state information,” The International Journal of Robotics Research , vol. 30, no. 7, pp. 895–913, 2011

    work page 2011

  26. [26]

    Risk-aware model predictive path integral control using conditional value-at-risk,

    J. Yin, Z. Zhang, and P. Tsiotras, “Risk-aware model predictive path integral control using conditional value-at-risk,” in 2023 IEEE Interna- tional Conference on Robotics and Automation (ICRA) . IEEE, 2023, pp. 7937–7943

    work page 2023

  27. [27]

    The CMA Evolution Strategy: A Tutorial

    N. Hansen, “The CMA evolution strategy: A tutorial,” arXiv preprint arXiv:1604.00772, 2016

    work page internal anchor Pith review Pith/arXiv arXiv 2016

  28. [28]

    Lectures on probability theory and mathematical statistics,

    M. Taboga, “Lectures on probability theory and mathematical statistics,” (No Title), 2017

    work page 2017

  29. [29]

    A sampling-and-discarding approach to chance-constrained optimization: feasibility and optimality,

    M. C. Campi and S. Garatti, “A sampling-and-discarding approach to chance-constrained optimization: feasibility and optimality,” Journal of optimization theory and applications, vol. 148, no. 2, pp. 257–280, 2011

    work page 2011

  30. [30]

    From Key Positions to Optimal Basis Functions for Probabilistic Adaptive Control,

    J. Jankowski, M. Racca, and S. Calinon, “From Key Positions to Optimal Basis Functions for Probabilistic Adaptive Control,” IEEE Robotics and Automation Letters, vol. 7, no. 2, pp. 3242–3249, 2022

    work page 2022

  31. [31]

    Mohri, A

    M. Mohri, A. Rostamizadeh, and A. Talwalkar, Foundations of machine learning. MIT press, 2018

    work page 2018

  32. [32]

    Motiondiffuser: Controllable multi-agent motion prediction using diffusion,

    C. Jiang, A. Cornman, C. Park, B. Sapp, Y . Zhou, D. Anguelov, et al., “Motiondiffuser: Controllable multi-agent motion prediction using diffusion,” in Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , 2023, pp. 9644–9653

    work page 2023

  33. [33]

    Wasserstein distributionally robust motion control for collision avoidance using conditional value-at-risk,

    A. Hakobyan and I. Yang, “Wasserstein distributionally robust motion control for collision avoidance using conditional value-at-risk,” IEEE Transactions on Robotics , vol. 38, no. 2, pp. 939–957, 2021

    work page 2021

  34. [34]

    Data-driven chance constrained control using kernel distribution embeddings,

    A. Thorpe, T. Lew, M. Oishi, and M. Pavone, “Data-driven chance constrained control using kernel distribution embeddings,” in Learning for Dynamics and Control Conference . PMLR, 2022, pp. 790–802

    work page 2022

  35. [35]

    Splines and linear control theory,

    Z. Zhang, J. Tomlinson, and C. Martin, “Splines and linear control theory,” Acta Math. Appl , vol. 49, pp. 1–34, 1997

    work page 1997

  36. [36]

    Computational geometry lecture notes hs 2013,

    B. G ¨artner and M. Hoffmann, “Computational geometry lecture notes hs 2013,” Dept. of Computer Science, ETH, Z ¨urich, Switzerland, 2013. APPENDIX APPENDIX A TRAJECTORY REPRESENTATION The way we represent trajectories is based on previous work showing that the closed-form solution to the following optimisation problem min Z 1 0 q′′(s) ⊤ q′′(s)ds s.t. q(s...

    work page 2013

  37. [37]

    Update the Probability of Direction Change: pk+1 = pk · (1 − α) (15) where α is the rate at which the probability of a direction change increases over time

    work page

  38. [38]

    • If r < p k+1 or if the projected position xk + ˙xk∆t is outside the boundaries of the conveyor belt, a direction change occurs

    Determine the Direction Change: • Sample a random number r from a uniform distri- bution between 0 and 1. • If r < p k+1 or if the projected position xk + ˙xk∆t is outside the boundaries of the conveyor belt, a direction change occurs

    work page

  39. [39]

    Update State based on Direction Change: ˙xk+1 = ( − ˙x if direction change occurs ˙x otherwise (16) pk+1 = ( α if direction change occurs pk+1 otherwise (17)

    work page

  40. [40]

    Update Position: xk+1 = x + ˙xk+1∆t (18) Therefore, the updated state vector after each time step is: sk+1 = [xk+1, ˙xk+1, p k+1] (19) In summary, the above models the probabilistic dynamics of one box particle on the conveyor belt, where the direction of motion can change randomly influenced by the parameter α and the physical constraints of the system

    work page