Application of Automatic Differentiation to Race Car Performance Optimisation
by
Daniele Casanova
Cranfield University
Coauthors: Robin S. Sharp (Cranfield University), Mark Final (University of Hertfordshire), Bruce Christianson (University of Hertfordshire), Pat Symonds (Benetton Formula Ltd.)

A formal method for the evaluation of the minimum time vehicle manoeuvre is described. The problem is treated as one of Optimal Control and is solved using a direct transcription method. The resulting Non Linear Programming problem is solved using the Sequential Quadratic Programming algorithm SNOPT for constrained optimisation. The Automatic Differentiation software tool ADopt is used for the evaluation of the first-order derivatives of objective and constraint functions with respect to the control variables. The implementation of Automatic Differentiation is shown to achieve a considerable time saving and greater robustness compared to the use of a finite difference determination of the Jacobian. The paper also focuses on aspects of the implementation of Automatic Differentiation relating to the complexity of the mechanical system to be optimised.

http://www.cranfield.ac.uk/sme/ace/auto/publications.htm#ad2000

Date received: December 17, 1999