A Dynamic Power Scheduling and Real-Time Monitoring System and Its Application to Robotic System
           

- 指導教授 黃漢邦 博士 研究生 陳癸宏

- Advisor :Dr.Han-Pang Huang Student :Quey-Hung Chen

Lab. of Robotics., Department of Mechanical Engineering National Taiwan University Taiwan

Abstract:

In view of enhancement of environmental awareness in recent years, the global energy conservation has become a topic of common concern. Especially in the field of power conservation, how to use the electricity more efficiently on robots without redundancy and also meet the fundamental requirements is an important issue. The purpose of this thesis is to develop a dynamic power scheduling and real-time monitoring system to make power management more structural and convenient so that the robots need not worry about the problem of excessive power consumption, and then complete the assigned tasks. First of all, in the aspect of dynamic power scheduling, a neighborhood search method which can search for different states to satisfy all constraints of the optimal scheduling will be developed, such as simulated annealing, genetic algorithm and Tabu search, etc. All the methodologies mentioned can be applied to handle most of the scheduling problems. Secondly, for real-time monitoring, in addition to the general real-time monitoring of the battery power status, we propose a Multi-Phase Principal Component Analysis (PCA) based Statistical Process Control (SPC) methods to monitor the process of optimal scheduling. The abnormal behaviors can be observed by a few of control charts, a main function of real-time monitoring system which will notify the remote monitoring personnel to take immediate action through the constructed GMPP system. Finally, in addition to promptly notify the remote monitoring personnel, the proposed method of dynamic power scheduling will also take corresponding measures to do rescheduling automatically to minimize the damage if the real-time monitoring system detect abnormal behaviors in the monitoring process. In this thesis, the cooperation of the dynamic power scheduling and real-time monitoring system which enable the power supply on robots to reach high reliability and good stability will be implemented practically in the future.





中文摘要:

有鑒於近幾年來環保意識的提升,全球性節約能源也成了大家關注的話題。其中在電力運用方面,若應用於機器人上,在達成目標的基本條件下,如何讓用電用的更 有效率,不必浪費多餘的電力也是個非常重要的議題。本論文的目的便是發展一套動態電力排程與即時監控系統來讓使用者對電源的管理更具架構、便利,以期能幫 助各式各樣的機器人不必擔心耗電過量的問題,進而完成被指派的任務。


首先,在動態電力排程方面,將開發出使用優化後的搜索演算法來找出在不同狀態下,滿足各限制條件的最佳化排程,像是模擬退火法、基因演算法和禁忌搜尋法等,都能用來解決大部份的排程問題。


其 次,在即時監控方面,除了一般的即時監控電池耗電狀態外,我們還提出一個以主成分分析為基之統計製程控管方法作為在最佳化排程執行的過程中是否有異常行為 的偵測。異常行為可藉由觀察少數的管制圖來找出變異,透過即時監控系統發現緊急情況,進而藉由自行開發之GMPP系統通知遠端監控人員採取立即的處置。


最後,即時監控系統在監控的過程中若發現有異常行為,除了及時通知遠端監控人員外,提出的動態電力排程方法也會自行採取相對應的重新排程策略,將傷害降至最低。


在本論文中,動態電力排程與即時監控系統的配合,可使機器人在電源管理上達到可靠度高、穩定性佳的供電效果,而該系統也會在未來實際運用於機器人上。