Dynamic and Self-adaptive Multi-agent Network for Optimal Operation of Engineering Processes

点击翻译按钮获取中文摘要

This proposal concerns dynamic and networked engineering processes with sharable resources. Difficult and complex engineering tasks in hazardous and partially-known environments may require cooperative, automated, and wireless operation of autonomous, dynamic, and heterogeneous agents such as robots (mobile or stationary), unmanned aerial vehicles (UAVs), devices that propel in water, and mobile sensor nodes. Multiple sensors, actuators, and other devices may be used, they may be mobile, and may have to be shared among tasks. Examples of engineering applications in this class, which the applicant is involved in are: 1. Spatiotemporal quality assessment (using mobile sensor nodes) of natural sources of water; 2. Multi-robot cooperation for homecare, human rescue, and industrial production; 3. Inspection and repair of pipeline networks that extract and distribute oil bitumen. The main objective of the proposed work is to develop a scalable system framework that will: 1. Accommodate more than one engineering application; 2. Select proper agents for cooperative optimal execution of a specified task, subjected to constraints (e.g., power consumption, system complexity, cost); 3. Adapt the network structure (e.g., add/drop agents; change: sensor location/orientation and activation choice, sampling rate, agent location and pose, connectivity and device parameters) for performance improvement, optimized with respect to multiple objectives. The work involves analytical research, computer simulation, technology development, implementation, and evaluation. The research activities will pertain to adaptive sensing, estimation using sensory data, and multi-sensor data fusion; multi-agent cooperation; and multi-objective and parameter/structure optimization. Specific research outcomes will include new or enhanced: 1. Methodologies for formulating agent models and cost functions (for performance capability, error, etc.), which may employ sensory estimation, device localization and navigation techniques, and self-awareness modeling for intelligent agents; 2. Sensor fusion methodologies, which may incorporate improved and/or hybrid forms of such techniques as the Bayesian approach, Dempster-Shafer evidence theory, nonlinear variations of Kalman filter, and intelligent/soft computing, which have relative advantages and disadvantages; 3. Optimal cooperation and decision making techniques, which may incorporate enhanced or hybridized biology-inspired methods (e.g., swarm intelligence, artificial immune systems, evolutionary computing with mechatronic design quotient—MDQ and quality of service as objective functions), Markov decision process—MDP, game theory, soft computing, and Pareto-optimal sets. The developed methodologies will be implemented and tested at an industrial site, combining water quality monitoring and pipeline inspection.