Advertisement

Model Validation and Analysis

  • Grégory Mermoud
Part of the Springer Tracts in Advanced Robotics book series (STAR, volume 93)

Abstract

Even with a great deal of care in constructing and calibrating a model, a validation of its underlying assumptions is necessary. One cannot prove the correctness of a model in the mathematical acceptation of the term; one can merely test it until it fails, and then modify it accordingly. An immediate corollary of this statement is that a model can only be proved wrong; however, a contradiction often provides deep insights into the system and its model. This observation is one of the strongest justification of multi-level approaches to the modeling of distributed robotic systems. Indeed, the different models of the same hierarchy generally disagree with each other to some extent, thereby pinpointing the assumptions that are not fulfilled and allowing the modeler to explore the various trade-offs between accuracy, scalability, and computational cost. Hereafter, we illustrate this claim by discussing in detail the validity of the models presented in the previous chapters; in particular, we analyze the impact of various assumptions on the models’ predictive accuracy.

Keywords

Reference Distribution High Abstraction Level Submicroscopic Level Baseline Prediction Geometrical Classis 
These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

Preview

Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.

Copyright information

© Springer International Publishing Switzerland 2014

Authors and Affiliations

  1. 1.Cisco SystemsRolleSwitzerland

Personalised recommendations