| تاریخ ارسال: 1401/6/23 |
دانشجو: آقای محمد یوسفی
استاد راهنما: دکتر فرشاد عشقی
استاد راهنما: دکتر منوچهر کلارستاقی
زمان: چهارشنبه ۳۰ شهریور ماه ۱۴۰۱ ساعت: ۱۳:۰۰
مکان: تهران سالن سمینار دانشکده فنی و مهندسی
استاد راهنما: دکتر فرشاد عشقی
استاد راهنما: دکتر منوچهر کلارستاقی
زمان: چهارشنبه ۳۰ شهریور ماه ۱۴۰۱ ساعت: ۱۳:۰۰
مکان: تهران سالن سمینار دانشکده فنی و مهندسی
Abstract
Wireless sensor networks are a network of spatially distributed sensors deployed to monitor an environment. Sensors are usually in proximity to a sink that is connected to the outside world. Target coverage, energy constraints, sink connectivity, and ad-hoc topology pose a unique challenge for routing in these networks. Maximizing network lifetime and satisfying these constraints has proven to be NP-Complete. Reinforcement Learning (RL) methods are gaining popularity for solving these combinatorial optimization problems. However, due to their stochastic nature, guarantees of constraint satisfaction are hard to prove. Inspired by previous integer programming and network flow approaches, we propose a novel RL method that guarantees full coverage and sink-connectivity while maximizing network lifetime.
Furthermore, the underlying Markov Decision Process can be used for generating feasible solutions to be utilized in the integer programming formulation of the problem.
Wireless sensor networks are a network of spatially distributed sensors deployed to monitor an environment. Sensors are usually in proximity to a sink that is connected to the outside world. Target coverage, energy constraints, sink connectivity, and ad-hoc topology pose a unique challenge for routing in these networks. Maximizing network lifetime and satisfying these constraints has proven to be NP-Complete. Reinforcement Learning (RL) methods are gaining popularity for solving these combinatorial optimization problems. However, due to their stochastic nature, guarantees of constraint satisfaction are hard to prove. Inspired by previous integer programming and network flow approaches, we propose a novel RL method that guarantees full coverage and sink-connectivity while maximizing network lifetime.
Furthermore, the underlying Markov Decision Process can be used for generating feasible solutions to be utilized in the integer programming formulation of the problem.
دفعات مشاهده: 378 بار |
دفعات چاپ: 66 بار |
دفعات ارسال به دیگران: 0 بار |
0 نظر