Cooperative Bayesian Search, Tracking, Localization and Mapping
发布日期：2012-05-20

报告题目

时间

2012-06-25 15:40:00

地点

木兰船建大楼A1008

报告人

Tomonari Furukawa

主办:	工程力学系
联系人:	胡文蓉

The talk will present cooperative Bayesian search, tracking, localization and mapping (STLAM) developed by the speaker as a result of his research into cooperative robotics over 10 years, which allows multiple autonomous vehicles to cooperatively search for, track and localize targets while self-localizing and constructing a map of environments in a unified theoretical framework. The talk will also cover its implementation where the platform- and the hardware-in-the-loop simulator (PHILS), the simulator specifically developed for cooperative control and construction of indoor and outdoor test sites are noticeably introduced in addition to successful experimental validations in the virtual and real environments.

The proposed cooperative STLAM effectively uses the extended Kalman filter (EKF) as the Gaussian estimator for tracking and localization and the element-based method (EM) proposed by the speaker as the non-Gaussian recursive Bayesian estimator for search while the occupancy element mapping (OEM) also proposed by the speaker is used for mapping. The proposed strategy also takes the computational architecture into account and performs EKF on the central processing unit (CPU) whilst the EM and the OEM are performed on the graphic processing unit (GPU). The belief fusion proposed by the speaker further enables the beliefs of targets and vehicles to be updated by multiple autonomous vehicles in real time without loss of information. The reliability and efficiency of autonomous control are significantly improved due to the unified Bayesian implementation in a seamless software architecture.

The PHILS has been developed based on the FlightGear server-client system to share an environment by multiple autonomous vehicles and runs on a PC cluster with Gigabit switch such that a real-time environment, which can even control communication delay, can be created. Construction of permanent, fully monitored indoor and outdoor test sites adjacent to the PHILS equipped with a number of monitors has allowed multiple autonomous robots to be tested and evaluated in various virtual and real environments. The effectiveness of the proposed strategy has been successfully demonstrated on the PHILS and in the test sites.

Cooperative Bayesian Search, Tracking, Localization and Mapping发布日期：2012-05-20

报告题目

时间

地点

报告人

Cooperative Bayesian Search, Tracking, Localization and Mapping
发布日期：2012-05-20