Abstract
As the widespread of mobile devices in recent years, mobile crowdsensing (MCS) has become a powerful mechanism to produce knowledge by collecting the individual contributed sensor data. In this paper, we aim to solve the target tracking problem through mobile crowdsensing. The traditional tracking method tends to rely on photos or videos provided by pre-deployed monitors, which may consume much power resources. Different from the traditional tracking method, the tracking approach through mobile crowdsensing (TAMC) proposed in this paper utilizes the wireless communication of mobile users to collect and contribute the valuable information about the target's whereabouts. Specifically, whenever the mobile users witness the target person, they will take photos of the target person and report the location and time of witnessing the target to the platform. Due to the fact that mobile users communicate with the platform only when they witness the target, the crowdsensing network composed of mobile users can be seen as a green network. In this way, the visited location history and corresponding time sequence of the target are available through the reports of mobile users. Once a new report is uploaded to the platform, the location history is updated. Then, according to the latest report, we apply a tree-based location prediction model named XGBoost, which is a scalable machine learning system, to predict the next place to be visited by the target. Finally, we conduct extensive experiments on a large-scale real-world dataset, namely, the Gowalla check-in dataset. The experimental results show that compared with the baseline methods, the tracking approach can predict the next places accurately.
Introduction: The Challenge of Ubiquitous Tracking
Traditional target tracking systems are heavily reliant on "pre-deployed monitors," such as fixed CCTV networks and dedicated sensors. While effective in controlled urban environments, these systems face significant hurdles. Coverage is often non-existent in suburban or rural areas, and the continuous operation of high-resolution video sensors is extremely resource-intensive, consuming vast amounts of power and bandwidth. Furthermore, environmental factors like poor weather can severely degrade performance.
The proposed Tracking Approach through Mobile Crowdsensing (TAMC) shifts this paradigm by leveraging the smartphones of everyday citizens. By turning mobile users into a dynamic, distributed sensor network, tracking becomes ubiquitous. However, this introduces a new challenge: witness reports are sporadic and non-consecutive. To track a target effectively, the system must "fill in the blanks" and predict future movements based on incomplete, crowdsourced data.
The TAMC Framework
The TAMC framework operates through a collaborative cycle between mobile users and a central processing platform.
1. Witness Reporting and Data Collection
When a mobile user witnesses the target, they act as a human sensor, uploading a photo, timestamp, and location to the central platform. This "green" network is only active when a sighting occurs, drastically reducing the energy footprint compared to always-on surveillance.
2. Multi-Aspect Feature Extraction and Prediction
Once a report is received, the platform processes the data through a sophisticated architecture. Because crowdsourced reports are sparse, the system extracts features from three distinct domains to bolster prediction accuracy:
- Individual Mobility: Analyzes the target's historical transitions and time-based patterns.
- Global Mobility: Considers the general popularity of locations across the entire user base.
- Contextual Features: Incorporates social "friendship" information and venue attributes (e.g., school vs. park).
The system frames location prediction as a binary classification problem. Using XGBoost, a scalable gradient-boosted decision tree system, the platform ranks candidate locations by their probability of being the target's next destination, generating a "top-k" list of likely places.
Experimental Evaluation
The researchers evaluated TAMC using the Gowalla check-in dataset, containing over 30 million records. To ensure robustness, they tested the model on data from two distinct cities: Honolulu and London.
Key Results
TAMC was compared against standard machine learning baselines, including Logistic Regression (LR), Random Forest (RF), and K-Nearest Neighbor (KNN).
- Superior Accuracy: TAMC’s F1 score outperformed Random Forest by 8% in Honolulu and 13% in London.
- Precision and Ranking: The model achieved an Average Percentile Rank (APR) of 0.99, indicating that the actual next location was almost always at the top of the predicted list.
- Accuracy@N: For a list of the top 9 potential locations, TAMC achieved over 70% accuracy, significantly outperforming KNN in handling sparse data.
Insights and Future Directions
A key insight of this research is the power of social influence on mobility. By including friendship data, the model acknowledges that human movement is socially driven—if a target's associates frequent a location, the target is more likely to appear there.
While promising, the framework faces challenges regarding privacy and real-time latency. Future work will need to address the ethical implications of accessing historical social data and develop incentive mechanisms to maintain an active and honest crowdsensing community.
Summary updated