Linear-Time Estimation with Tree Assumed Density Filtering and Low-Rank ApproximationReportar como inadecuado


Linear-Time Estimation with Tree Assumed Density Filtering and Low-Rank Approximation


Linear-Time Estimation with Tree Assumed Density Filtering and Low-Rank Approximation - Descarga este documento en PDF. Documentación en PDF para descargar gratis. Disponible también para leer online.

We present two fast and memory-efficient approximate estimation methods, targeting obstacle avoidanceapplications on small robot platforms. Our methods avoida main bottleneck of traditional filtering techniques, whichcreates densely correlated cliques of landmarks, leading toexpensive time and space complexity. We introduce a noveltechnique to avoid the dense cliques by sparsifying them into atree structure and maintain that tree structure efficiently overtime. Unlike other edge removal graph sparsification methods,our methods sparsify the landmark cliques by introducing newvariables to de-correlate them. The first method projects thecurrent density onto a tree rooted at the same variable ateach step. The second method improves upon the first oneby carefully choosing a new low-dimensional root variable ateach step to replace such that the independence and conditionaldensities of the landmarks given the trajectory are optimallypreserved. Our experiments show a significant improvement intime and space complexity of the methods compared to other standard filtering techniques in worst-case scenarios, with smalltrade-offs in accuracy due to low-rank approximation errors.



Computational Perception and Robotics - Computational Perception and Robotics Publications -



Autor: Ta, Duy-Nguyen - Dellaert, Frank - -

Fuente: https://smartech.gatech.edu/







Documentos relacionados