An Optimal Solution to the Linear Search Problem for a Robot with Dynamics

In general, the ability to find a target quickly can involve great improvements and be a step forward for many different real applications. This project implements a new control policy that optimizes the process of searching a target that is distributed uniformly over the unit interval and its location is unknown by the searcher until it passes over it. This algorithm minimizes the total time it takes a point mass with bounded acceleration to find an object, starting from the origin at rest, and then return to its position. The implementation was made on proof-of-concept hardware with a two-link planar robot arm that searched for metal coins using an inductive proximity sensor that did not respond to nonmetallic objects. The dynamics of the manipulator were taken into account in order to compensate for them and therefore be able to perform a smoother movement and have a more accurate control. Particularly, as an extension of this work, it is demonstrated here that the optimal solution approach can also be extended to optimal search along arbitrary curves such as raster-scan patterns, meaning that the linear search problem is not only confined to straight lines. This can be of interest to other applications like robot search-and-rescue.unpublishednot peer reviewedU of I OnlyUndergraduate senior thesis not recommended for open acces