Optimal Smooth Paths Based on Clothoids for Car-like Vehicles in the Presence of Obstacles

Abstract

Automated Guided Vehicles are increasingly used for material transfer in factory and warehouse environments amongst humans and human operated vehicles. Safe and efficient operation is challenging when there is a mix of human and automated traffic as fixed guide paths can become blocked more frequently. In this work we aim to show smooth and efficient paths based on clothoid curves can be used to automatically plan diversions which can be traversed at high speed by automated fork-lift vehicles, which are car-like in the sense they have a limited turning radius and angular acceleration. The approach, based on numerical optimisation within convex region constraints is described in detail, and numerical results for curvature and sharpness are compared to a cubic spline on a small number of simulated environments. The clothoid spline is less affected, in terms of its objective function, by a shift in the obstacle boundaries than a cubic spline, for obstacle shifts below 0.5m. The clothoid spline takes longer to converge for but the output path has attractive qualities like lower peak sharpness, enabling high speed operation. The method is therefore most useful for applications where path quality is important and updates are required less frequently. Changing the objective function by increasing weighting parameter b allowed the path shape to be tuned to reduce the peak sharpness, at the cost of increasing the total length. With b > 100, convergence was poor because parts of the spline were pushed outside the assigned region, an artefact arising from the constraints only being enforced at the start and end of each segment. The analytical Jacobian of the constraints was effective at reducing the number of function evaluations to reach convergence.