420 likes | 556 Views
M ulticast Utility-Based Scheduling for UWB Networks. Kuang-Hao Liu et al Presented by Xin Che 11/18/09. Introduction . IEEE 802.15.3 For WPANs Piconet Controller Peer-to-Peer mode It is proposed for narrowband wireless communications It is not suitable for UWB
E N D
Multicast Utility-Based Scheduling for UWB Networks Kuang-Hao Liu et al Presented by Xin Che 11/18/09
Introduction • IEEE 802.15.3 • For WPANs • Piconet Controller • Peer-to-Peer mode • It is proposed for narrowband wireless communications • It is not suitable for UWB • Concurrent transmissions: Multiple user interference • Ranging capability
Introduction • UWB-based WPANs formulate the optimal scheduling problem as a utility maximization problem !
Introduction • Pro: • A utility-based scheduling algorithm aiming at multiclass QoS provisioning with fairness consideration • Cons: • an efficient scheduling algorithm requires feedback information from the network to appropriately make scheduling decisions • it is very difficult, if not impossible, for the PNC to acquire instantaneous channel information of each flow.(Due to peer-to-peer communications)
Introduction • To estimate the achievable data rate of a flow • PNC can make use of the ranging capability featured by UWB communications [14], [15]. • But, distance information obtained may be noisy due to multipath fading ! • the utility estimation may be biased, and thus affects the scheduling decisions !
Introduction • Solution in this paper • resort to metaheuristic methods and choose to use the global search algorithm (GSA) [17]. • its convergence to the global optimum can be proved, • the tradeoff between computational complexity and efficiency is tunable. • the exclusive-region-based GSA (ER-GSA) • a desired convergence with reasonable computational complexity for practical implementations
Introduction • Contributions of this paper • The scheduling algorithm for concurrent UWB transmissions maximizes the weighted utility is formulated (NP-Hard) • a utility-based scheduling scheme is proposed to support multiclass traffic with fairness constraint • The assumption of perfect distance information for measuring flow throughput is relaxed by factoring estimation errors into the objective function. • The stochastic optimization problem is solved by the proposed ER-GSA, and its convergence property and computational complexity are studied
System Model • Network Structure
System Model • Simplified channel model • Assume that a UWB receiver can adapt its transmission rate to an arbitrary SINR level • the achievable data rate r_iof flow i is upper bounded by • neglect the multipath fast fading when we estimate the average data rate ri
System Model • Utility Function • Utility is defined as the satisfaction level of a user with respect to the amount of allocated bandwidth. • For heterogeneous traffic, general nondecreasing functions with values within [0, 1] • Traffic types are classified into three classes
System Model • Class 1 • constant bit-rate app. E.g. audio streams • Class 2 • Can adapt to the allocated bandwidth to a certain extent : video stream
System Model • Class 3 • Can adapt to the allocated bandwidth to a certain extent : video stream
Optimal Scheduling • Deriving • very difficult, if not impossible, as U(k) is combinatiorial : dependent on the element in κ. • Use discrete approximation • Let be
Proposed Algorithm • ER-GSA • the optimal flow set κ* can be found by evaluating the utility value for each member in K to locate the maximal member • Simple, but has exponential complexity. • Cannot deal with estimation errors. • the GSA is selected as the base to solve (13) since its convergence to a global optimum can be theoretically proved under certain conditions.
Proposed Algorithm • GSA • relies on a random sequence generated during the algorithm iterations to efficiently find the optimum. • The resulting random sequence is a Markov chain, where each state represents a point in the solution space that has been visited by the algorithm • In each iteration, the transition of the Markov chain is determined by comparing the objective value of the current state and that of a randomly chosen point from the solution space
Proposed Algorithm • The convergence of ER-GSA
Proposed Algorithm • Utility Update
Proposed Algorithm • The scheduling policy has the followoing properties :
Performance Evaluation • Experiment Setting
Performance Evaluation • Experiment Setting • Each superframe contains ten slots. • The size of exclusive region, which is denoted as dER, is set to 2 m, • in Section V-C, we vary the size of exclusive region to study its impact on the aforementioned three performance metrics.
Performance Evaluation • Traffic
Performance Evaluation • Utility-Based Scheduling
Performance Evaluation • Utility-Based Scheduling
Performance Evaluation • Utility Vs. Fariness • Total Utility Vs. Fairness
Performance Evaluation • Utility Vs. Fariness
Performance Evaluation • Minimum Utility
Performance Evaluation • Algorithm Efficiency and Stability
Performance Evaluation • Algorithm Efficiency and Stability
Performance Evaluation • Algorithm Efficiency and Stability • Stability factor
Performance Evaluation • Stability
Conclusion • a utility-based optimal scheduling for concurrent UWB transmissions supporting heterogeneous traffic has been proposed • it is found that the size of the exclusive region in UWB networks is independent of the transceiver distance, which, on the contrary, is a dependent parameter in narrowband wireless systems. • The proposed algorithm can also maintain a good balance between the computation complexity and the robustness against measurement and estimation errors, and thus, it suits UWB network schedulers with limited computation power.