13.07.2015 Views

Assessment and Future Directions of Nonlinear Model Predictive ...

Assessment and Future Directions of Nonlinear Model Predictive ...

Assessment and Future Directions of Nonlinear Model Predictive ...

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

L i (s i ,u i ):=Numerical Methods for Efficient <strong>and</strong> Fast NMPC 167t∫i+1t iL(x i (t),z i (t),φ i (t, u i )) dt. (6)Note that up to now the multiple shooting parameterization does not involveany discretization <strong>of</strong> differential operators f,g, but is exact.Structured <strong>Nonlinear</strong> Programming ProblemThe parameterization <strong>of</strong> problem (3a)–(3f) using multiple shooting <strong>and</strong> a suitablecontrol representation leads to the following structured nonlinear programming(NLP) problem :N−1∑minu,si=0L i (s i ,u i ) + E(s x N ) (7a)subject to s x 0 = x 0, (7b)0=g(s x i ,s z i ,φ i (t i ,u i )), i =0, 1,...N−1, (7c)s x i+1 = x i(t i+1 ; s i ,u i ), i =0, 1,...N−1, (7d)r(s x N ) ≥ 0,(7e)with initial condition (7b), consistency conditions (7c), continuity conditions(7d), <strong>and</strong> terminal constraint (7e). Additional control <strong>and</strong> path constraints aresupposed to be imposed pointwise for a suitable discretization (at n i points τ ijon each interval, τ ij ∈ [t i ,t i+1 ), j =0,...,n i − 1)h(x i (τ ij ; s i ,u i ),z i (τ ij ; s i ,u i ),u i ) ≥ 0, j =0,...,n i −1, i =0,...N−1. (7f)The NLP (7a)–(7e) can be summarized asP (x 0 ):{bx0 (w)=0min a(w) subject towc(w)≥0,(8)where w contains all the multiple shooting state variables <strong>and</strong> controls:w =(s x 0,s z 0,u 0 ,s x 1,s z 1,u 1 ,...,u N−1 ,s x N ) ∈ R nw .The function a(w) is the objective (7a), the vector valued equation b x0 (w) =0summarizes all equalities from (7b)-(7d), <strong>and</strong> the vector valued c(w) ≥ 0containsthe inequality constraints (7f) <strong>and</strong> (7e).It is important to note that the initial condition (7b) is a linear constraintamong the equality constraints, with the varying parameter x 0 entering linearlyonly in this constraint, so that⎡s x 0 − x ⎤⎡ ⎤0−I nxg(s x 0,s z 0,φ 0 (t 0 ,u 0 ))b x0 (w) =⎢ s⎣1 − x 0(t 1 ; s x 0 ,sz 0 ,u 0) ⎥⎦ = b 00(w)+Lx 0 with L :=⎢ 0 ⎥⎣ ⎦ . (9)..

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!