Reconfigurable radios: Part 2 - Energy-aware radio management

Part 1 looks at different hardware architectures, including a multicore solution, as well as reconfigurable analog front ends.

Energy-aware cross-layer radio management exploits flexibility at run-time
In this section, we list state-of-the-art energy saving techniques for wireless communication, emphasizing the need for a system-level approach. Next, we propose a framework for energy-aware cross-layer radio management, able to exploit the flexibility for saving energy introduced in part 1 of the series. The key idea is that significant energy savings are possible by continuously adapting the energy-scalable reconfigurable radio to the actual environment conditions and performance constraints at run-time, exploiting cross-layer synergies.

State-of-the art energy management
Improving the energy efficiency of wireless devices has already resulted in much research, from circuit to communication theory and networking protocols. In general, the energy management problem consists of dynamically controlling a system to minimize the average energy consumption under a performance constraint.

The state-of-the-art research can be listed into 2 categories.

• Top-down approaches. Approaches that are intrinsically utilization- and hardware-aware but communication-unaware are categorized as top-down. The communicating device is treated as any electronics circuit and general-purpose techniques like dynamic power management and energy-aware design are applied. The first technique is defined as dynamically reconfiguring an electronic system to provide the requested performance levels with a minimum number of active components and the minimum loads on those components [15]. The second technique can be defined as designing systems that presents a desirable energy – performance behavior for energy management [16].

• Bottom-up approaches. Approaches that are intrinsically communication-aware but hardware-unaware are categorized as bottom-up. They rely on the fundamentals of information and communication theory to derive energy-aware transmission techniques and communication algorithms. We find here for instance the transmission scaling techniques, which exploit the fundamental trade-off that exists between transmission rate/power and energy [17]. Network power management techniques also fall in this category, targeting the minimization of the transmission power under QoS constraints.

Top-down and bottom-up approaches can contradict when they are applied independently. An example is the conflict between transmission scaling at the PHY layer (bottom-up) and sleeping schemes at the MAC layer (top-down). Scaling tends to minimize transmission energy consumption by transmitting with the lowest power over the longest feasible duration, whereas sleeping tend to minimize the duty cycle of the radio circuitry by transmitting as fast as possible (Figure 8). Clearly, the two techniques are contradictory when it comes to defining the optimal transmit rate and power allocation.

The above reveals the need for a holistic cross-layer approach to carry out system-level energy management, jointly considering techniques to optimize the energy consumption in the different layers. This conclusion is strengthened by the simple observation that user-relevant performance metrics can only be measured on top of the whole protocol stack, close to the user, while energy consumption is mainly conditioned by the setting and policies within the lowest protocol layers, close to the hardware. The topic of cross-layer optimization has recently gained a lot of interest (see [26]). Existing approaches, however, miss the holistic view, hampering real-life deployment and sometimes leading to undesired side effects.

Figure 8. Top-down and bottom-up approaches can be contradictory.

System-level cross-layer energy optimization
For an energy-aware cross-layer management framework for wireless systems, the proposed approach handles cross-layer interactions to minimize energy consumption under performance constraint(s). A key element is the Pareto-based design-time/run-time principle [27], which was recently applied to the context of wireless communications [18]-[19]. The general idea is to shift most of the complexity to the design through a systematic exploration leading to control solutions that result in minimum complexity at run-time. The proposed approach can be summarized as follows.

• Scalable architecture definition (see part 1). A first step is to enable run-time controllable parameters (or knobs) that significantly impact performance and energy consumption at system-level. The control dimension settings in real implementations are discrete, inter-dependent and can have a non-linear influence.

• Discretization of run-time dynamics. The second important design-time step is the discretization of the system run-time dynamics (e.g., environment conditions, application requirements) through the definition of relevant external variables that can be tracked at run-time

• Problem modeling and partitioning. The system energy and performance behavior is modeled in function of these configuration parameters and external variables. A hierarchy of power-management specific abstraction layers is then introduced, to partition the global optimization problem into sub-problems that can be solved locally. A valid abstraction layer can be characterized by intermediate metrics that can be optimized locally without significantly hampering the optimality of the global optimization. It is hence possible to derive the optimal points along these local metrics (according to the Pareto multi-objective optimality criterion), and systematically prune away the non-optimal configurations. This enables a drastic complexity reduction of the next optimization steps.

• Control policy definition. The three first steps result in sufficient system characterization to conduct the global optimization. Given the Pareto-optimal configuration points for each abstraction layer in the considered partitioning, the fourth step then consists in deriving the policy that will be used at run-time to determine the optimal configuration for the whole system. This requires determining (a) how to communicate the local information and (b) how to combine that information to obtain a global optimum.

• Run-time operation. The run-time control algorithm tracks the external variables, select current scenario and runs the optimization policy to select the optimal configuration points of each layer in order to provide the just-required performance at minimum energy consumption. This results in low complexity overhead thanks to the design-time/run-time partitioning, and especially the availability of a set of potential operating points that is fully characterized up-front and that is organized in monotonous Pareto curves.

Page 2: Cognitive radios allow energy-efficient spectrum sharing and seamless connectivity
Page 1 | 2

COURSE
1. Protecting High-Speed I/Os from ESD

TECH PAPER
2. Book Excerpt: "Electrical Engineering 101" (Part 3 of 5)

WEBINAR
3. Detecting Five Distinct Motions with MEMS Inertial Sensors

WEBINAR
4. Achieve greater productivity and ease of use with Targeted Design Platforms enabled by Virtex-6 and Spartan-6 FPGAs