This paper presents a pedestrian motion model that includes both low level trajectory patterns, and high level discrete transitions. The inclusion of both levels creates a more general predictive model, allowing for more meaningful prediction and reasoning about pedestrian trajectories, as compared to the current state of the art. The model uses an iterative clustering algorithm with (1) Dirichlet Process Gaussian Processes to cluster trajectories into continuous motion patterns and (2) hypothesis testing to identify discrete transitions in the data called transition points. The model iteratively splits full trajectories into sub-trajectory clusters based on transition points, where pedestrians make discrete decisions. State transition probabilities are then learned over the transition points and trajectory clusters. The model is for online prediction of motions, and detection of anomalous trajectories. The proposed model is validated on the Duke MTMC dataset to demonstrate identification of low level trajectory clusters and high level transitions, and the ability to predict pedestrian motion and detect anomalies online with high accuracy.
This paper presents an implementation of place cells for a robot navigation using the K-iterations Fast Learning Artificial Neural Networks (KFLANN) clustering algorithm. The KFLANN possesses several desirable properties suitable for place cell robot navigation tasks. The technique proposed is able to autonomously adjust the resolution of cells according to the complexity of the environment. This is achieved through two parameters known as the tolerance and the vigilance of the network. In addition, a navigation system consisting of a topological map building and a place cell path planning strategy is presented. A physical implementation of the system was developed on an autonomous platform and actual results were obtained. The experimental results obtained indicate that the system was able to navigate successfully through the experimental space and also tolerate unexpected discrepancies arising from motor and sensor errors present in a real environment. Furthermore, despite abrupt changes in an environment due to the deliberate introduction of obstacles, the system was still able to cope without changes to the program. The experiment was also extended to include a kidnapped robot scenario and the results were favorable, indicating a positive use of allothetic cue recognition capabilities.
A terrain mapping model is proposed using a generalized Markov random field (MRF) representation. Unlike previous work, the proposed MRF can fully represent uncertainties due to sensor pose and measurement errors, as well as data association errors in a single model. Additionally, neither homoscedasticity nor a predefined shape of the likelihood distribution is assumed. The flexibility of an MRF model allows spatial height correlations to be incorporated. The ability to include spatial correlations not only improves the accuracy through the benefits of Bayesian prior modeling, but also serves as a basis for terrain property characterization. Maximum likelihood solutions of terrain roughness are derived. Benefits of the proposed model are demonstrated experimentally on indoor and outdoor datasets. Results show that the MRF model leads to lower height estimation errors. In addition, the capability of estimating non-Gaussian height distributions allows the information about individual terrain features to be preserved. Finally, the model is able to accurately estimate the roughness of the terrain, which is beneficial for edge detection of obstacles and nontraversible terrain regions.Index Terms-Mapping, Markov random field (MRF), range sensing, sensor fusion.
The study of a semi-supervised clustering has recently attracted great interest from the data clustering community. Work in semi-supervised clustering systems has been done focusing on specific types of auxiliary information, i.e. a partial labeling or pairwise constraints. However, in some applications the clustering characteristics desired may not be restricted to only these predefined types of constraints. Furthermore, the user may not always be able to formulate an explicit clustering specification. In such cases, only a weak good or bad evaluation feedback is obtainable as semi-supervision information. This research proposes a novel form of semi-supervised clustering using Reinforcement K-Iteration Fast Learning Artificial Neural Network (R-KFLANN) architecture that utilizes generic reward or punishment feedback, enabling it to address different types of high-level clustering requirements provided at run-time. To illustrate this concept, the R-KFLANN was tested in two different application domains of data classification and robot mapping. The results indicate that the system was able to adapt to the online reinforcement presented and eventually improve the output in serving the covert specifications of both tasks. It could significantly improve the clusters using only overall failure rate information loosely coupled with the hidden class labels in the classification problem. This was also evident in the navigation problem when the clustering specification could not even be explicitly formulated; R-KFLANN was still able to incorporate the high-level task's characteristics into the cluster representation, yielding a better map efficiency. Additionally, it could fulfill the navigation task requirements which would not be achievable unless the system was tuned manually using a small tolerance. These findings suggest the usefulness of R-KFLANN semi-supervised clustering in serving clustering objectives imposed online by the high-level tasks without being restricted to the traditional semi-supervised constraints.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.