Optimization of a Double Wishbone Suspension Geometry for Off-road Vehicles using Genetic Algorithm and Machine Learning

This paper explores methods to predict wheel alignment angles and optimize double wishbone suspension geometry for off-road vehicles. Suspension three dimensional points are evaluated using a Genetic Algorithm for desirable values of toe, camber, kingpin and castor angles. A regression methodology is employed to accurately predict these angles at all values of the bump angle. RMS error on testing dataset for camber, toe, caster and kingpin are found to be 0.0098, 0.0257, 0.0145, and 0.013 respectively, making the values acceptable for use in optimization. For prime vehicle performance of an ATV in bumps and droops, all four angles must lie between their specific ranges. Genetic Algorithm incorporates these constraints to determine chromosome fitness, and generate better off springs. Selection, crossover, and mutation are operated on each generation of the populations to converge towards most ideal coordinates. Results are validated using the Lotus Suspension Analysis software, leading to an optimal wheel alignment.