Samenvatting
We propose a learning-based, distributionally robust model predictive control approach towards the design of adaptive cruise control (ACC) systems. We model the preceding vehicle as an autonomous stochastic system, using a hybrid model with continuous dynamics and discrete, Markovian inputs. We estimate the (unknown) transition probabilities of this model empirically using observed mode transitions and simultaneously determine sets of probability vectors (ambiguity sets) around these estimates, that contain the true transition probabilities with high confidence. We then solve a risk-averse optimal control problem that assumes the worst-case distributions in these sets. We furthermore derive a robust terminal constraint set and use it to establish recursive feasibility of the resulting MPC scheme. We validate the theoretical results and demonstrate desirable properties of the scheme through closed-loop simulations.
| Originele taal-2 | Engels |
|---|---|
| Pagina's (van-tot) | 15128-15133 |
| Aantal pagina's | 6 |
| Tijdschrift | IFAC-PapersOnLine |
| Volume | 53 |
| Nummer van het tijdschrift | 2 |
| DOI's | |
| Status | Gepubliceerd - 2020 |
| Extern gepubliceerd | Ja |
| Evenement | 21st World Congress of the International Federation of Aufomatic Control (IFAC 2020 World Congress) - Berlin, Duitsland Duur: 12 jul. 2020 → 17 jul. 2020 Congresnummer: 21 https://www.ifac2020.org/ |
Financiering
Abstract Abstract WAbestproractpose a learning-baseffl, fflistributionally robust mofflel prefflictiΩe control approach towarffls We propose a learning-baseffl, fflistributionally robust mofflel prefflictiΩe control approach towarffls We propose a learning-baseffl, fflistributionally robust mofflel prefflictiΩe control approach towarffls autonomous stochastic system, using a hybriffl mofflel with continuous fflynamics anffl ffliscrete, autonomous stochastic system, using a hybriffl mofflel with continuous fflynamics anffl ffliscrete, Mar∆oΩian inputs. We estimate the (un∆nown) transition probabilities of this mofflel empirically Mar∆oΩian inputs. We estimate the (un∆nown) transition probabilities of this mofflel empirically Mar∆oΩian inputs. We estimate the (un∆nown) transition probabilities of this mofflel empirically (ambiguity sets) arounffl these estimates, that contain the true transition probabilities with (ambiguity sets) arounffl these estimates, that contain the true transition probabilities with (ambiguity sets) arounffl these estimates, that contain the true transition probabilities with case fflistributions in these sets. We furthermore ffleriΩe a robust terminal constraint set anffl use case fflistributions in these sets. We furthermore ffleriΩe a robust terminal constraint set anffl use it to establish recursiΩe feasibility of the resulting MPC scheme. We Ωalifflate the theoretical it to establish recursiΩe feasibility of the resulting MPC scheme. We Ωalifflate the theoretical results anffl fflemonstrate fflesirable properties of the scheme through closeffl-loop simulations. results anffl fflemonstrate fflesirable properties of the scheme through closeffl-loop simulations. Creospuyltrisgahnt ffl©ffl2e0m20o nTshter aAtuethffloerssi.r aTbhlies ipsr aonpeorpteine sacocfetssh earstcichlee munedtehr rtohue CC BY-NC-ND licengh closeffl-loop simuse lations. (Khtetypw:/o/crrdesa:tiLveeacronminmgoanns.ffloragffl/laicpetnasteiosn/byin-nacu-ntodn/4o.m0)ous Ωehicles, Intelligent fflriΩer aiffls, Motion Keywords: Learning anffl afflaptation in autonomous Ωehicles, Intelligent fflriΩer aiffls, Motion Kcoeyntwrolords: Learning anffl afflaptation in autonomous Ωehicles, Intelligent fflriΩer aiffls, Motion control control1. INTRODUCTION 1. INTRODUCTION 1. INTRODUCTION In recent fflecaffles, the usage of afflaptiΩe cruise control In recent fflecaffles, the usage of afflaptiΩe cruise control In recent fflecaffles, the usage of afflaptiΩe cruise control research anffl infflustry, as they haΩe fflemonstrateffl numerous research anffl infflustry, as they haΩe fflemonstrateffl numerous benefits in terms of safety, fuel efficiency, passenger comfort, benefits in terms of safety, fuel efficiency, passenger comfort, etc. The term ACC generally refers to longitufflinal control etc. The term ACC generally refers to longitufflinal control etc. The term ACC generally refers to longitufflinal control Ωelocity, while aΩoiffling collisions with preceffling Ωehicles. Ωelocity, while aΩoiffling collisions with preceffling Ωehicles. Ωelocity, while aΩoiffling collisions with preceffling Ωehicles. The recently proposeffl Responsibility-Sensitive Safety (RSS) The recently proposeffl Responsibility-Sensitive Safety (RSS) framewor∆ (ShaleΩ-Shwartz et al., 2017), prescribes min-framewor∆ (ShaleΩ-Shwartz et al., 2017), prescribes min-framewor∆ (ShaleΩ-Shwartz et al., 2017), prescribes min-Ωehicle ∆inematics, which can guarantee collision aΩoifflance Ωehicle ∆inematics, which can guarantee collision aΩoifflance Ωehicle ∆inematics, which can guarantee collision aΩoifflance tions. Furthermore, the authors fflefine rules that prescribe tions. Furthermore, the authors fflefine rules that prescribe tions. Furthermore, the authors fflefine rules that prescribe of this safety fflistance. Although safe, the prescribeffl rules of this safety fflistance. Although safe, the prescribeffl rules of this safety fflistance. Although safe, the prescribeffl rules maneuΩers, refflucing passenger comfort anffl fuel efficiency. maneuΩers, refflucing passenger comfort anffl fuel efficiency. maneuΩers, refflucing passenger comfort anffl fuel efficiency. By contrast, mofflel prefflictiΩe control (MPC) methoffls By contrast, mofflel prefflictiΩe control (MPC) methoffls optimize a specifieffl performance infflex baseffl on the optimize a specifieffl performance infflex baseffl on the optimize a specifieffl performance infflex baseffl on the ★ This work was supported by the fflord-KU Leuven Research Alliance. This work was supported by the fflord-KU Leuven Research Alliance. G086318N; No. G086518N; fflonds de la Recherche Scientifique – The work of P. Patrinos was supported by: fflWO projects: No. fflNRS, the fflonds Wetenschappelijk Onderzoek – Vlaanderen under fflNRS, the fflonds Wetenschappelijk Onderzoek – Vlaanderen under fflNRS, the fflonds Wetenschappelijk Onderzoek – Vlaanderen under EOS Project No. 30468160 (SeLMA), Research Council KU Leuven EOS Project No. 30468160 (SeLMA), Research Council KU Leuven C1 project No. C14/18/068. C1 project No. C14/18/068. 2405-8963 Copyright © 2020 The Authors. This is an open access article under the CC BY-NC-ND license. Peer review under responsibility of International Federation of Automatic Control. 10.1016/j.ifacol.2020.12.2037 prefflicteffl eΩolution of the controlleffl system in the near prefflicteffl eΩolution of the controlleffl system in the near future, which enfflows the control system with the capability future, which enfflows the control system with the capability to behaΩe proactiΩely, anffl afflapt its actions with respect to to behaΩe proactiΩely, anffl afflapt its actions with respect to potential future eΩents. HoweΩer, fflue to the inΩolΩement potential future eΩents. HoweΩer, fflue to the inΩolΩement potential future eΩents. HoweΩer, fflue to the inΩolΩement in the preffliction of traffic situations. In orffler to explicitly in the preffliction of traffic situations. In orffler to explicitly in the preffliction of traffic situations. In orffler to explicitly particularly popular approach (Bichi et al. (2010); Moser particularly popular approach (Bichi et al. (2010); Moser etpartical. (u2018)larly;pMopulacDonrouapghproetachal.(B(2013)ichi e)t.al. (2010); Moser et al. (2018); McDonough et al. (2013)). etInalan. (2018)attem;ptMcDtoonmoua∆ghe aetccuraal.te(2013)prefflic). tions about the future behaΩior of the leaffl Ωehicle, many fflifferent fflriΩer future behaΩior of the leaffl Ωehicle, many fflifferent fflriΩer mofflels haΩe been proposeffl in the literature (see Wang et al. mofflels haΩe been proposeffl in the literature (see Wang et al. mofflels haΩe been proposeffl in the literature (see Wang et al. continuous physics-baseffl fflynamics with a ffliscrete (anffl continuous physics-baseffl fflynamics with a ffliscrete (anffl potentially stochastic) fflecision mofflel for the fflriΩer (e.g., potentially stochastic) fflecision mofflel for the fflriΩer (e.g., Saffligh et al. (2014); Kienc∆e et al. (1999); Bichi et al. Saffligh et al. (2014); Kienc∆e et al. (1999); Bichi et al. Saffligh et al. (2014); Kienc∆e et al. (1999); Bichi et al. preceffling Ωehicle using fflouble integrator fflynamics, where preceffling Ωehicle using fflouble integrator fflynamics, where the fflriΩer’s inputs are generateffl by a Mar∆oΩ chain. the fflriΩer’s inputs are generateffl by a Mar∆oΩ chain. the fflriΩer’s inputs are generateffl by a Mar∆oΩ chain. their fflepenfflence on accurate ∆nowlefflge of all probability their fflepenfflence on accurate ∆nowlefflge of all probability fflistributions inΩolΩeffl in the stochastic mofflel. Since, fflistributions inΩolΩeffl in the stochastic mofflel. Since, in practice, these are estimateffl baseffl on finitely sizeffl in practice, these are estimateffl baseffl on finitely sizeffl in practice, these are estimateffl baseffl on finitely sizeffl unfflerlying fflistributions — we will refer to this uncertainty unfflerlying fflistributions — we will refer to this uncertainty on probability fflistributions as ambigfiity. Due to this on probability fflistributions as ambigfiity. Due to this on probability fflistributions as ambigfiity. Due to this ambiguity, stochastic controllfflrs may pfflrflorm unrfflliably with rfflspfflct to thffl truffl ffistributions.
Duurzame ontwikkelingsdoelstellingen van de VN
Deze output draagt bij aan de volgende duurzame ontwikkelingsdoelstelling(en)
-
SDG 3 – Goede gezondheid en welzijn
Vingerafdruk
Duik in de onderzoeksthema's van 'Learning-Based Risk-Averse Model Predictive Control for Adaptive Cruise Control with Stochastic Driver Models'. Samen vormen ze een unieke vingerafdruk.Citeer dit
- APA
- Author
- BIBTEX
- Harvard
- Standard
- RIS
- Vancouver