对于所考虑的车辆模型和参数不确定性分布,在重复算法1 两到三次后,每种类型的最坏情况最可能收敛。图7总结了算法最后一次运行时各参数对轨迹刚度和垂直荷载指标的影响。在每个图中,顶部的子图显示了导致参数值和度量值之间的单调递增关系(Nin/Nin + Nde + Not)、单调递减关系(Nde/Nin + Nde +Not)和非单调关系(Not/Nin + Nde + Not)的机动百分比。每个关系(Rin、Rde和Rot)的度量值的最大偏差范围总结在第二个子图中。最后一个子图通过引入投票度量N·R并将其标准化,从而将上述信息组合在一起,使每个参数的最大投票度量为1。然后使用投票度量值来确定每种类型中最可能出现的最坏情况。对于每个参数,如果单调增加关系的投票度量是最大的,则参数下限值在最可能最坏情况的参数值集合中,而如果单调递减关系的投票度量是最大的,参数上限值在集合中。然而,如果非单调关系具有最大的投票度量值,则需要对结果进行更详细的讨论,如下所示。
[1] Khatib O. Real-time obstacle avoidancefor manipulators and mobile robots. Int J Rob Res.1986;5:90–98.
[2] Ogren P, Leonard NE. A convergentdynamic window approach to obstacle avoidance. IEEE Trans Robot.2005;21:188–195.
[3] Shimoda S, Kuroda Y, Iagnemma K.High-speed navigation of unmanned ground vehicles on uneven terrain usingpotential fields. Robotica. 2007;25:409–424.
[4] Park JM, Kim DW, Yoon YS, et al.Obstacle avoidance of autonomous vehicles based on model predictive control.Proc Inst Mech Eng Part D J Automob Eng. 2009;223:1499–516.
[5] Gao Y, Lin T,Borrelli F, et al. Predictive control of autonomousground vehicles with obstacle avoidance on slippery roads. Proceedings of theASME 2010 Dynamic Systems and Control Conference (DSCC) Vol. 1; 2010 Sept12-15; American Society of Mechanical Engineers, Cambridge, MA; 2010. p.265–272.
[6] Tahirovic A,MagnaniG.General frameworkfor mobile robot navigation using passivity-based MPC. IEEE Trans AutomatControl. 2011;56:184–190.
[7] Gray A, Gao Y, Lin T, et al. Predictivecontrol for agile semi-autonomous ground vehicles using motion primitives.Proceedings of the 2012 American Control Conference (ACC 2012); 2012 Jun 27-29;Institute of Electrical and Electronics Engineers Inc., Montréal, Canada; 2012.p.4239–4244.
[8] Frasch JV,Gray A, Zanon M, et al. An auto-generated nonlinear MPCalgorithm for real-time obstacle avoidance of ground vehicles. Proceedings the2013 European Control Conference (ECC 2013); 2013 Jul 17-19; IEEE ComputerSociety, Zurich, Switzerland; 2013. p. 4136–4141.
[9] Beal CE, Gerdes JC.Model predictivecontrol for vehicle stabilization at the limits of handling. IEEE Trans ControlSyst Technol. 2013;21:1258–1269.
[10] Liu J, Jayakumar P, Stein JL, ET AL. Anonlinear model predictive control formulation for
obstacle avoidance in high-speed autonomousground vehicles in unstructured environments.
Vehicle Syst Dyn. 2018;56:853–882.
[11] Liu J, Jayakumar P, Stein JL, et al.Combined speed and steering control in high speed
autonomous ground vehicles for obstacleavoidance usingmodel predictive control. IEEE Trans Veh Technol.2017;66:8746–8763.
[12] Liu J, Jayakumar P, Stein JL, et al. Astudy on model fidelity for model predictive controlbased obstacle avoidance inhigh-speed autonomous ground vehicles. Vehicle Syst Dyn.2016;54:1629–1650.
[13] Walton CL. The design andimplementation of motion planning problems given parameter uncertainty [PhDdissertation]. Santa Cruz: University of California; 2015.
[14] Bemporad A, Morari M. Robust modelpredictive control: A survey. In: Garulli A, Tesi A, editors. Robustness inidentification and control. Lecture Notes in Control and Information Sciences,vol 245. London: Springer; 1999.
[15] Jalali AA,NadimiV.Asurvey onrobustmodel predictive control from1999–2006. Proceedings of the CIMCA 2006International Conference on Computational Intelligence for Modelling, Controland Automation. Jointly with IAWTIC 2006 International Conference onIntelligent AgentsWeb Technology, Institute of Electrical and ElectronicsEngineering Computer Society; 2008 Dec 10-12; Vienna, Austria; 2007.
[16] Adetola V, Guay M. Robust adaptive MPCfor constrained uncertain nonlinear systems. Int J Adapt Control SignalProcess. 2011;25:155–167.
[17] Carvalho A, Lefevre S, Schildbach G,et al. Automated driving: the role of forecasts and uncertainty–a controlperspective. Eur J Control. 2015;24:14–32.
[18] Ramirez DR, Alamo T, Camacho EF, etal. Min-MaxMPC based on a computationally efficient upper bound of the worstcase cost. J Process Control. 2006;16:511–519.
[19] Carson JM, Acikmese B, Murray RM, etal. A robust model predictive control algorithm augmented with a reactivesafetymode. Automatica. 2013;49:1251–1260.
[20] Gao Y, Gray A, Tseng HE, et al. Atube-based robust nonlinear predictive control approach to semiautonomousground vehicles. Vehicle SystDyn. 2014;52:802–823.
[21] SchildbachG, Fagiano L, Frei C, et al. The scenario approach forstochasticmodel predictive control with bounds on closed-loop constraintviolations. Automatica. 2014;50:3009–3018.
[22] Carvalho A, Gao Y, Gray A, et al.Predictive control of an autonomous ground vehicle using an iterativelinearization approach. Proceedings of the 16th International IEEE Conferenceon Intelligent Transportation Systems (ITSC 2013); 2013 Oct 6-9; The Hague, TheNetherlands; 2013. p. 2335–2340.
[23] Maiworm M, Bathge T, Findeisen R.Scenario-based model predictive control: recursive feasibility and stability.IFAC Proc Vol. 2015;28:50–56.
[24] Xiong F, Xiong Y, Xue B. Trajectoryoptimization under uncertainty based on polynomial chaos expansion. Proceedingsof the AIAA Guidance, Navigation, and Control Conference 2013; 2013 Aug 19-22;American Institute of Aeronautics and Astronautics Inc, AIAA, Boston, MA; 2015.
[25] Xiu D, Em Karniadakis G. TheWiener-Askey polynomial chaos for stochastic differential equations. SIAM J SciComput. 2003;24:619–644.
[26] Shim T, Ghike C. Understanding thelimitations of different vehicle models for roll dynamics studies. User ModelUser-Adapt Interact. 2007;45:191–216.
[27] Pacejka H. Tire and vehicle dynamics.Oxford: Elsevier; 2012.
[28] Rucco A, Notarstefano G, Hauser J.Optimal control based dynamics exploration of a rigid car with longitudinalload transfer. IEEE Trans Control Syst Technol. 2014;22:1070–1077.
[29] Viana FAC. Things you wanted to knowabout the Latin hypercube design and were afraid to ask. Proceedings of the10th World Congress on Structural and Multidisciplinary Optimization; 2013 May5-24; Orlando, Florida; 2013. p. 1–9.
[30] Viana FAC, Venter G, Balabanov V. Analgorithm for fast optimal latin hypercube design of experiments. Int J NumerMethods Eng. 2010;82:135–156.