报告地点：腾讯会议ID: 271 449 678
报告人： Somaiyeh MahmoudZadeh
Dr. Somaiyeh MahmoudZadeh completed her PhD in Computer Science (AI-Robotics and Autonomous Systems) at Flinders University of South Australia in 2017. She is a lecturer of Artificial Intelligence and Machine Learning at the School of Information Technology, Deakin University. She has a demonstrated track record of research focused on autonomous systems, mission management, and motion planning. Somaiyeh was employed as a Research Fellow at the Monash University with a team of experts on a large international project aimed at using Bayesian network technology in argument analysis within the CREATE program run by US from 2017 to 2018. Her research expertise includes artificial intelligence and machine learning, focusing on autonomous systems including AUVs, USVs, UAVs and their practical applications in oceanography, environmental monitoring and protection, smart agriculture, and mining. Synthesizing and analyzing cooperative and non-cooperative robotics’ problems, including deliberative and reactive mission planning, mission risk assessment, designing control architecture for decision making based on proper situational awareness, contribute to the core part of her research. This also included autonomous motion planning, navigation, and formation planning considering vehicular and environmental constraints. Her research also extends to the data-driven fields of Machine Learning, including causal argumentation analysis using Bayesian Networks, optimization modeling, and decision analysis using CI and Soft computing methods such as (Fuzzy Cognitive Maps, ANN).
Situational and cognitive load awareness allows multi-agent systems to identify when cooperation is needed to share available resources or entire mission reformation when one or more agents fail to continue the intended operation. One of the significant hurdles in multi-agent mission planning is to ensure no overlap occurs in the fleet operations. Establishing synchronized cooperation in multi-agent mission formations guarantees operation stability in terms of time management and mission endurance, thus improving the robustness of the system coverage.
In this presentation, a cooperative mission planner system is designed to accommodate parallel operation, collective resource sharing, and adaptive re-planning for uninterrupted operation in fault occurrence. The proposed system is established upon a hierarchal architecture including three phases to facilitate different stages of a mission plan for a team of unmanned vehicles. The architecture includes three main sub-modules: i) The first module aims at optimal segregating the field and determining the operational zone for each agent; ii) The motion planning module to provide agents with feasible routes in the allocated zone, where the main objective is to optimize the coverage, minimize the travel length, and manage the operation time to increase overall mission productivity; and iii) Adaptive mission re-planning that enables cooperation and fault handling to continue the operation in situations where any vehicle fails. The cooperation and fault-tolerant property of the proposed system guarantees the maximum coverage and retains the system performance at a near-optimal level by preventing interruption when an unforeseen failure occurs, leading an agent to abort its mission. Thus, the proposed system has the capacity of situational awareness to perform adaptive re-planning with respect to changes in the surrounding environment or the internal status of the vehicles.