Development of a Multi-robotic Exploration System for Power Plants

  • Sihai An

Student thesis: Master of Philosophy


On-site exploration is an important procedure in scheduled Operation and Maintenance (O&M) in power plants. Currently, the efficiency of O&M work is challenging due to human reliability and the limitations of the deployed robots. Thus, the goal of this project is to develop an enhanced robotic exploration system for on-site data collection of power plants. Specifically, this project focuses on developing an efficient coordination method for a multi-robotic exploration system, with the aim of maximising utilisation of the limited onboard energy of exploration robots to accomplish on-site inspection tasks with high efficiency. In the exploration scenarios, this project considered using a limited number of robots to conduct continuous exploration of multiple targets. Two modes were specifically considered for on-site exploration at a power plant: (1) temporary exploration, whereby the robots are required to conduct the exploration as quickly as possible to diagnose faults zone-by-zone, and (2) long-term exploration, whereby the robots are required to use their limited energy resource to maximise the number of inspected targets explored. In the exploration system’s development, this project considered two factors for optimal exploration-efficiency: (1) scheduling of an optimal exploration plan, and (2) appropriate charging controls. Consequently, three multi-robotic exploration approaches were developed in this study: (1) the Greedy algorithm and general charging method for temporary exploration, (2) the Genetic Algorithm (GA) and general charging method for long-term exploration, and (3) the GA and predicted charging method for exploration system improvement. A comparison of these three approaches showed that the developed Greedy based method was suitable for temporary exploration tasks to diagnosis faults zone-by-zone. However, the developed GA based method had more advantages in long-term exploration. Finally, it was found that the predicted charging method could save energy and increase inspection efficiency of the exploration system. In application, these developed exploration approaches can be used for different scenarios inside a power plant, and can be applied to other similar domains, such as cooperate rescue, farming or cleaning.
Date of Award31 Dec 2020
Original languageEnglish
Awarding Institution
  • The University of Manchester
SupervisorSimon Watson (Supervisor) & Farshad Arvin (Supervisor)

Cite this