Any future work on MuID pattern recognition needs to focus on performance, in three major areas: rejection of ghost roads, rejection of duplicate roads, and efficiency. Clearly, all of these effects need to be studied quantitatively.
The number of ghost roads can be reduced by a better treatment of the ambiguities inherent in assigning which horizontal and vertical hits go together. Currently, we choose to attach clusters to a road without checking for a cluster of the opposite orientation within the search window. This might lead to a situation where, for example, a set of horizontal clusters belonging to a single good track may be assigned to multiple roads, some of which have vertical clusters that belong to another track or that do not belong to any track.6 Requiring horizontal and vertical clusters in a plane will probably eliminate some of this behavior, but will also reduce road-finding efficiencies. The question of efficiencies is the reason that this requirement has not been implemented yet. Still, this may be an interesting avenue for future studies.
The problem of duplicate roads has not yet been fully addressed. Primarily, this occurs where a particle strikes a plane in a region where panels overlap. Since each cluster is composed of hits in a single panel, duplicate clusters are typically found, resulting in multiple roads being found for that particle, all of the roads having similar trajectories.
Also, the road-finding efficiencies need to be improved in high-multiplicity events. The key here is to minimize the bias induced by background hits within a cluster, which shifts the cluster centroid and thus shifts the projection of the road to the other planes and to the vertex plane.