Probabilistic Mapping of Spatial Motion Patterns for Mobile Robots by unknow

Author:unknow
Language: eng
Format: epub
ISBN: 9783030418083
Publisher: Springer International Publishing

3.5.2 Limitations and Future Work

In the previous section, we have listed the advantages of CT-map and T-CT-map. However, these methods also have some limitations and can undergo further development.

CT-map and T-CT-map are local methods, that allows to build partial models from incomplete data, in contrast to global methods where incomplete data may make building a model more difficult. However, it also causes that the global motion patterns (reaching beyond a neighbourhood) are not explicitly modelled. In order to extract them it is necessary use other techniques (i.e. CPP-tree).

Furthermore, CT-map and T-CT-map are parametric models limited by the underlying grid map. That means they are not able to differentiate sub-pixel orientation differences neither sub-pixel velocity changes. T-CT-map also assumes that the durations for entry and exit shift are unimodal and normally distributed. These assumptions can limit the quality of the predictions generated with this model.

CT-map and T-CT-map are limited to what type of dynamics they can represent. These representations are tailored to model motion patterns of discrete dynamic objects. That means, that motion of continuous media or changes of the environment caused by semi-static objects can be modelled with neither of the representations.

In the first part of this section, we have presented limitations of the expressiveness of the representation. Another type of limitations is related to the procedure of parameter estimation of CT-map and T-CT-map.

CT-map and T-CT-map use as an input sequences of binary occupancy grid maps. The process of binarization can introduce additional noise, which can substantially affect both representations. One of the ways to avoid this issue is to use the continuous occupancy grid map as input.

However, binarization is just one of the sources of noise. Therefore, a key research direction is robustification of the transition detection process. This task can be approached in at least to ways. The first approach is to detect the edges between occupied and free areas in the maps using an edge detector [5]. Once the edges in the maps are reliably detected, this information can be used for detecting the shifts of the said edge over the cell borders. The second approach focuses on processing the cells occupancy signals. This can include filtering the signal and application of robust step detection methods.

Further research is not only limited to improving the representations but also relates to the possible applications. The two primary areas of application of CT-map and T-CT-map are motion prediction and motion planning.

In the case of motion prediction CT-map and T-CT-map can be used as an enhancement for Bayesian occupancy filter (BOF) [6]. BOF aims at predicting the future state of occupancy map based on the most recent changes in occupancy. In this case CT-map and T-CT-map can provide information about the motion patterns present in the environment. The information about motion patterns in the environment will help to improve the quality of the prediction and also enable BOF to make long-time predictions.

The application of CT-map and T-CT-map is two fold. First, these representations can be used for estimation of the future shape of the environment in the presence of dynamic objects.

Download

Copyright Disclaimer:
This site does not store any files on its server. We only index and link to content provided by other sites. Please contact the content providers to delete copyright contents if any and email us, we'll remove relevant links or contents immediately.