Learning-Based Risk-Averse Model Predictive Control for Adaptive Cruise Control with Stochastic Driver Models

Schuurmans, Mathijs; Katriniok, Alexander; Tseng, H. Eric; Patrinos, Panagiotis

doi:10.1016/j.ifacol.2020.12.2037

Cited by 13 publications

(13 citation statements)

References 17 publications

(17 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Consequently, if the risk measures employed in the definition of the DR-OCP (14) belong to the broad family of conic risk measures and the dynamics f ( • , • , w), w ∈ W are linear, then, using the reformulations in [8], ( 14) can be cast as a convex conic optimization problem. This is the case for many commonly used coherent risk measures, including the risk measure induced by the ℓ 1 -ambiguity set discussed in Example II.5 (see [24] for a numerical case study). For nonlinear dynamics, the problem is no longer convex but can in practice still be solved effectively with standard NLP solvers.…”

Section: Definition Iii3 (Dr-ocp) Given An Augmented State Z ∈ Z the ...mentioning

confidence: 95%

Learning-Based Distributionally Robust Model Predictive Control of Markovian Switching Systems with Guaranteed Stability and Recursive Feasibility

Schuurmans¹,

Patrinos²

2020

Preprint

Self Cite

View full text Add to dashboard Cite

We present a data-driven model predictive control scheme for chance-constrained Markovian switching systems with unknown switching probabilities. Using samples of the underlying Markov chain, ambiguity sets of transition probabilities are estimated which include the true conditional probability distributions with high probability. These sets are updated online and used to formulate a time-varying, risk-averse optimal control problem. We prove recursive feasibility of the resulting MPC scheme and show that the original chance constraints remain satisfied at every time step. Furthermore, we show that under sufficient decrease of the confidence levels, the resulting MPC scheme renders the closed-loop system mean-square stable with respect to the true-but-unknown distributions, while remaining less conservative than a fully robust approach.

show abstract

Section: Definition Iii3 (Dr-ocp) Given An Augmented State Z ∈ Z the ...mentioning

confidence: 95%

Learning-Based Distributionally Robust Model Predictive Control of Markovian Switching Systems with Guaranteed Stability and Recursive Feasibility

Schuurmans¹,

Patrinos²

2020

Preprint

Self Cite

View full text Add to dashboard Cite

show abstract

“…A. Aggressive Navigation in Known Cluttered Environments 1) Simulation Setup: In this experiment, we use the kinematics model of a differential drive robot presented in [16] for sampling trajectories in the conventional MPPI and propagating sigma points in the proposed U-MPPI technique, as outlined in (9). The model's state includes its position and orientation in the world frame, given by x = [x, y, θ] ⊤ ∈ R 3 .…”

Section: Simulation-based Evaluationmentioning

confidence: 99%

“…To address the challenges posed by uncertainties, Risk-Sensitive MPC (RSMPC) approaches have gained traction in recent years, thanks to their ability to balance the benefits and drawbacks of robust and stochastic MPC methods. By integrating the concept of risk measures or risk metrics into the optimization problem, RSMPC can evaluate the impact of uncertainty and adjust responses accordingly to different levels of uncertainty [7]- [9].…”

Section: Introductionmentioning

confidence: 99%

Autonomous Navigation of AGVs in Unknown Cluttered Environments: Log-MPPI Control Strategy

Mohamed

Yin

Liu

2022

IEEE Robot. Autom. Lett.

View full text Add to dashboard Cite

“…When designing controllers for this class of systems, it is often assumed that the discrete mode is directly measurable [2], [7]- [9]. However, in practice, the discrete mode typically needs to be inferred from measurements of the continuous M. Schuurmans and P. Patrinos are with the Department of Electrical Engineering (ESAT-STADIUS), KU Leuven, Kasteelpark Arenberg 10, 3001 Leuven, Belgium.…”

Section: A Background and Motivationmentioning

confidence: 99%

“…By adopting a distributionally robust framework, we can do so while retaining system theoretic guarantees, such as stability. We will illustrate this by means of a linear controller design, but keeping in mind more advanced applications, involving model predictive approaches [2].…”

Section: A Background and Motivationmentioning

confidence: 99%

Data-driven distributionally robust control of partially observable jump linear systems

Schuurmans¹,

Patrinos²

2021

Preprint

Self Cite

View full text Add to dashboard Cite

We study safe, data-driven control of (Markov) jump linear systems with unknown transition probabilities, where both the discrete mode and the continuous state are to be inferred from output measurements. To this end, we develop a receding horizon estimator which uniquely identifies a sub-sequence of past mode transitions and the corresponding continuous state, allowing for arbitrary switching behavior. Unlike traditional approaches to mode estimation, we do not require an offline exhaustive search over mode sequences to determine the size of the observation window, but rather select it online. If the system is weakly mode observable, the window size will be upper bounded, leading to a finite-memory observer. We integrate the estimation procedure with a simple distributionally robust controller, which hedges against misestimations of the transition probabilities due to finite sample sizes. As additional mode transitions are observed, the used ambiguity sets are updated, resulting in continual improvements of the control performance. The practical applicability of the approach is illustrated on small numerical examples.

show abstract

Learning-Based Risk-Averse Model Predictive Control for Adaptive Cruise Control with Stochastic Driver Models

Cited by 13 publications

References 17 publications

Learning-Based Distributionally Robust Model Predictive Control of Markovian Switching Systems with Guaranteed Stability and Recursive Feasibility

Learning-Based Distributionally Robust Model Predictive Control of Markovian Switching Systems with Guaranteed Stability and Recursive Feasibility

Autonomous Navigation of AGVs in Unknown Cluttered Environments: Log-MPPI Control Strategy

Data-driven distributionally robust control of partially observable jump linear systems

Contact Info

Product

Resources

About