The project advances the state-of-the-art of various robotics fields, including home health care robotics, robot mechanisms and kinematics, visual servoing, modeling and analysis, estimation and control, search and tracking, and localization and mapping. In addition, the integration of these technologies and its demonstration provides an opportunity to verify the capabilities and limitations of the state-of-the-art technologies, discovering new subsequent robotics problems to tackle.
This project addresses the fundamental aspects relevant to many scientific and engineering applications and includes experimental demonstrations, and broadly impacts teaching, training and education. Further, the international collaboration provides graduate students who reside in the rural Southside Virginia opportunities to observe scientific, cultural and political similarities and differences between countries and encourages them to continue their efforts toward research, and to establish an international community of home health care robotics. The project contributes to the modernization of the engineering curriculum in general and the teaching of mobile robotics in particular.
The primary research activities at Virginia Tech have been the developments of (1) a simultaneous localization and mapping (SLAM) technique that most robustly and accurately maps any complicated home environments and (2) a Bayesian tracking technique that allows the robot to reliably localize an MIP under uncertainties and track the MIP and provide assistance. The Virginia Tech group has worked with two groups from Chiba University, one led by Prof. Nonami specializing in the robot design and control and the other led by Prof. Yu engaged in the monitoring of MIPs. The aim of the first year was to develop fundamental technologies for the development of the home healthcare robot. Each group has developed technologies independently and exchanged information. Several Ph.D students participated in the Virginia Tech group as part of educational activities. They developed the major part of the hardware and software systems of the project under the supervision of Prof. Furukawa (PI). They hosted two visits of students from Chiba University. Five students of Prof. Nonami’s group participated in the first visit (2/16/12-2/19/12) whereas Prof. Yu and his four students joined in the second visit (2/27/12-3/5/12). Two Ph.D students of the Virginia Tech group also visited together with Profs. Furukawa and Taheri (12/4/12-12/11/12). The first key technology that has been theoretically formulated and practically implemented at Virginia Tech is the grid-based scan-to-map matching SLAM technique. Whilst a large number of SLAM techniques have been developed to date, none of the existing SLAM techniques have been able to accurately and efficiently develop a map and navigate a robot accordingly in highly complicated environments including the home environments of concern. The existing SLAM techniques can be classified as feature-based SLAM techniques and scan-based SLAM techniques. The major drawback of the feature-based SLAM techniques is the feature recognition. Since no algorithms can detect features reliably, accuracy is significantly limited by the feature recognition. The scan-based SLAM techniques does not have the problem, but almost all the techniques match the raw scan points to corresponding points or other correspondences and thus yields the correspondence error. Further, these conventional techniques require a significant amount of computation as both the feature extraction and the point-to-point matching necessary for the techniques are time consuming. The grid-based scan-to-map matching SLAM technique, achieving normal distribution (ND) to ND matching in a grid space, removed the correspondence error and further enabled efficient computation. Experiments have confirmed that the grid-based scan-to-map matching SLAM technique achieves SLAM most accurately and efficiently. The proposed technique is suited for home healthcare since high-speed operation is necessary for searching for and tracking humans without sacrificing accuracy. The second key technology that the Virginia Tech group has developed is the grid-based Bayesian search, tracking, localization and mapping (STLAM) which enabled not only the SLAM but also search and tracking in a unified grid-based probabilistic framework. Academically, this is the most generalized framework for autonomous estimation and control. It is further the useful technique for the project since the tracking of the MIP and the SLAM can be conducted in the same framework. As outreach activities, the latest results of the project have been uploaded onto the webpage of the Virginia Tech group (Computational Multiphyics Systems Laboratory): www.cmsvt.org/ The PI honorably gave a plenary talk at two different international conferences and gave huge impression about the need of the technology to large audience: 7th International Conference on Intelligent Unmanned Systems (ICIUS 2011), Chiba, November, 2011 2012 IEEE International Conference on Multisensor Fusion and Information Integration (MFI 2012), Hamburg, September, 2012 The PI also organized a panel discussion at ICIUS 2011 on the dual use of robotics technologies so that robotics technologies developed for different purposes can be integrated and improve ability in each application. The project has additionally produced three journal papers and four conference papers.
