Motivation
Wireless networks are designed to be flexible, supporting applications with different performance requirements. For example, home networks can provide service for a time-sensitive application such as voice, at the same time they support a time-insensitive but reliable file transfer. This flexibility is valuable to users in a range of scenarios, but it comes with the cost of performance.
Consider a cyber-physical system such as a modern car, in which sensors are sampling information from physical phenomena (e.g. pressure of the tire, quantity of fuel, engine rotational speed and proximity to obstacles) and sending this information to the central computer of the car through an unreliable wireless channel. In this scenario, guaranteeing high communication performance to critical devices such as the proximity sensor that can avoid collisions is primary, while flexibility is (at best) secondary. Similar scenarios can be found in industrial plants, smart grids, oil platforms, military operations, et al. For these networks, algorithm design must be custom-made and centered around the performance requirements of the specific application.
In our work, we create models for wireless networks transporting data with stringent performance requirements and use tools from Stochastic Control and Mathematical Optimization to develop networking control algorithms with superior performance, tailored to the application's requirements. These tools often allows us to provide performance guarantees for our algorithms. Recent work focuses on single-hop wireless networks with unreliable channels and addresses traffic that requires high throughput, low latency and/or information freshness.
Scheduling for Freshest Information
Age of Information (AoI) has been receiving increasing attention in the literature, particularly for applications that generate time-sensitive data such as position, command and control, or sensor data. An interesting feature of this performance metric is that it captures the freshness of the information from the perspective of the destination, in contrast to the long-established packet delay, that represents the freshness of the information with respect to individual packets. In particular, AoI measures the time elapsed since the generation of the packet that was most recently delivered to the destination, while packet delay measures the time elapsed from the generation of a packet to its delivery.
We consider a single-hop wireless network with a number of clients sending time-sensitive information to a base station through unreliable channels and formulate a discrete-time decision problem to find a scheduling policy that minimizes the expected weighted sum AoI of the clients in the network.
The results are divided in two parts. First, we show that a Greedy policy, which transmits the packet with highest current age, is optimal for the case of symmetric networks. Then, for the case of general networks, we develop three additional low-complexity scheduling policies: a Randomized policy, a Max-Weight policy and a Whittle’s Index policy. By comparing the performance of each policy against the optimal AoI, we derive their performance guarantees. To the best of our knowledge, this is the first project to derive performance guarantees for transmission scheduling policies that attempt to minimize AoI in wireless networks with unreliable channels. Numerical results show that both Max-Weight and Whittle’s Index policies outperform Randomized and Greedy in every configuration simulated, and achieve near optimal performance.
Publication: Allerton'16, INFOCOM'18, ISIT'18 and others being prepared.
Scheduling of Periodic Real-Time Traffic
We consider a single-hop wireless network composed of a base station and a number of clients, with the goal of scheduling real-time traffic when the network provides delayed feedback. Delayed feedback is of increasing importance in systems where the round trip delay is much greater than the packet transmission time, and it has a significant effect on the scheduling decisions and network performance.
Previous work in the literature considered the problem of scheduling real-time traffic with instantaneous feedback and without feedback. In this project, we address the general case of delayed feedback and use Dynamic Programming to characterize the optimal scheduling policy. An optimal algorithm that fulfills any feasible minimum delivery ratio requirements is proposed. Moreover, we develop a low-complexity suboptimal heuristic algorithm which is suitable for platforms with low computational power. Both algorithms are evaluated using simulations.
Publications: SMThesis'16 and Allerton'15
Optimization of IEEE 802.11e
Every IEEE 802.11e station separates packets into four queues, one for each QoS category: Voice, Video, Best Effort and Background. Each queue transmits packets according to its own Binary Exponential Backoff (BEB) mechanism. The BEB tunes the rate of transmission by controlling the Contention Window (CW). Basically, BEB increases CW when it senses that the network has too much traffic and reduces CW when traffic is light.
We develop a simple and standard-compliant mechanism that estimates the total number of active queues of each QoS category in the network and uses this estimate to find the CWs that maximize throughput. Then, instead of using the standard BEB to tune CW, we propose that the optimal CWs are periodically broadcast by the Access Point using Beacon Frames and employed by all stations in the network. Simulation results demonstrate the superior performance of the mechanism in saturated and non-saturated traffic conditions.
Publication: Computer Communications'14
