The Power of Smartphones

Feng Xia, Ching-Hsien Hsu, Xiaojing Liu, Fangwei Ding, Wei ZhangMultim. Syst 2013 · DOI · OpenReview

Abstract

While smartphones have evolved into sophisticated mobile sensing platforms, their operational longevity is strictly governed by battery capacity. Wireless communication, particularly the process of discovering and maintaining Wi-Fi connections, represents a significant source of energy depletion. This paper presents SwitchR, a location-assisted Wi-Fi discovery scheme designed to mitigate the high energy costs of network scanning. By utilizing low-power sensors like A-GPS and ZigBee to predict proximity to known access points, SwitchR intelligently manages radio activation, significantly extending battery life during high-bandwidth activities.

The Problem: The High Cost of Discovery

Wireless data transfer presents a paradox: Wi-Fi is significantly more energy-efficient than cellular data (such as GPRS or EDGE) for large file transfers, yet it is remarkably inefficient when idle or searching for a signal. The standard method for locating Wi-Fi networks involves periodic scanning, which can consume up to 1.426W. In environments with sparse coverage, a device may waste substantial energy searching for non-existent networks. The challenge lies in transitioning to Wi-Fi only when it is available and necessary, without the overhead of continuous scanning.

Methodology: The SwitchR System

The authors propose SwitchR, a system comprising three functional modules:

  1. Rate Monitor Module: Tracks the user's current cellular data rate.
  2. Switching Decision Engine: Triggers the switching process when bandwidth demand exceeds a specific threshold (e.g., during video streaming).
  3. Switching Module: The core innovation that replaces active Wi-Fi scanning with location-based prediction.

To achieve efficient switching, the system relies on a hierarchy of sensors that balance power consumption against positioning accuracy.

Table 1: Comparison of power consumption and accuracy for GPS, GSM, and ZigBee sensors.

As shown in the table above, while GSM positioning is low-power, its 400m accuracy is too coarse for effective Wi-Fi triggering. SwitchR utilizes Assisted GPS (A-GPS) for outdoor environments and ZigBee for high-accuracy indoor positioning.

The "Wait" Logic

Instead of activating the Wi-Fi radio as soon as a user enters a general vicinity, SwitchR calculates a specific activation time ($t_{switch}$). By referencing a local log of known Access Points (APs), the system determines the distance ($d$) to the nearest AP. Assuming a constant movement speed ($v$), it sets a timer ($t_{switch} = d/v$). The Wi-Fi radio remains powered down until this timer expires, at which point a single scan is performed to establish a connection.

Key Insights and Results

The study highlights a critical "Wi-Fi Efficiency Paradox": Wi-Fi is the best interface for transferring data but the worst for waiting for it. The necessity of intelligent switching is underscored by the relationship between bandwidth and energy efficiency.

Figure 1: Energy efficiency (KB/J) of EDGE vs. Wi-Fi as bandwidth demand increases.

As illustrated in the graphs, the energy efficiency of Wi-Fi increases exponentially relative to cellular data as bandwidth demand grows. SwitchR captures this efficiency while eliminating the "failed scan" energy waste common in standard scanning methods. In simulations involving heavy downloaders (User Profile 4), SwitchR consistently outperformed both standard scanning and GSM-based switching. The high accuracy of A-GPS (10m) and ZigBee (5m) allowed for much tighter activation windows compared to the 400m error margin of GSM, which often triggered "false positive" scans.

Limitations and Future Directions

While promising, the current SwitchR model relies on a simplistic movement model assuming constant speed and linear paths. Real-world human mobility is often more erratic. Additionally, the indoor positioning model utilizes ZigBee, which is currently less common in consumer hardware than Bluetooth Low Energy (BLE). Future work aims to integrate Quality of Service (QoS) metrics into the decision engine and explore offloading AP-matching computations to cloud platforms to further reduce local CPU load.

Summary updated

Abstract

Smartphones have been shipped with multiple wireless network interfaces in order to meet their diverse communication and networking demands. However, as smartphones increasingly rely on wireless network connections to realize more functions, the demand of energy has been significantly increased, which has become the limit for people to explore smartphones' real power. In this paper, we first review typical smartphone computing systems, energy consumption of smartphone, and state-of-the-art techniques of energy saving for smartphones. Then we propose a location-assisted Wi-Fi discovery scheme, which discovers the nearest Wi-Fi network access points (APs) by using the user's location information. This allows the user to switch to the Wi-Fi interface in an intelligent manner when he/she arrives at the nearest Wi-Fi network AP. Thus we can meet the user's bandwidth needs and provide the best connectivity. Additionally, it avoids the long periods in idle state and greatly reduces the number of unnecessary Wi-Fi scans on the mobile device. Our experiments and simulations demonstrate that our scheme effectively saves energy for smartphones integrated with Wi-Fi and cellular interfaces.
BibTeX
@article{xia2013power,
  author = {Feng Xia and Ching-Hsien Hsu and Xiaojing Liu and Fangwei Ding and Wei Zhang},
  title = {{The Power of Smartphones}},
  journal = {Multim. Syst.},
  year = {2013},
  doi = {10.1007/s00530-013-0337-x},
}
Figure