Optimal Detection of Faulty Traffic Sensors Used in Route Planning

Amin Ghafouri, Aron Laszka, Abhishek Dubey, Xenofon D. Koutsoukos

Abstract

In a smart city, real-time traffic sensors may be deployed for various applications, such as route planning. Unfortunately, sensors are prone to failures, which result in erroneous traffic data. Erroneous data can adversely affect applications such as route planning, and can cause increased travel time. To minimize the impact of sensor failures, we must detect them promptly and accurately. However, typical detection algorithms may lead to a large number of false positives (i.e., false alarms) and false negatives (i.e., missed detections), which can result in suboptimal route planning. In this paper, we devise an effective detector for identifying faulty traffic sensors using a prediction model based on Gaussian Processes. Further, we present an approach for computing the optimal parameters of the detector which minimize losses due to false-positive and false-negative errors. We also characterize critical sensors, whose failure can have high impact on the route planning application. Finally, we implement our method and evaluate it numerically using a real-world dataset and the route planning platform OpenTripPlanner.

Introduction

In the ecosystem of a "smart city," real-time traffic sensors—such as inductive loops or cameras—form the backbone of modern navigation applications. However, these sensors are notoriously prone to hardware failures and miscalibrations. Standard anomaly detection usually treats all errors equally, aiming to balance False Positives (FP) and False Negatives (FN). In the context of route planning, however, not all errors are equal. A false alarm on a minor side street might have negligible impact, while a missed detection of a failure on a major highway could lead thousands of drivers into gridlock. This paper addresses the gap between generic statistical anomaly detection and the specific performance requirements of traffic routing applications by introducing an application-aware framework.

Methodology

The proposed detection framework utilizes a three-stage pipeline designed to move beyond simple error-rate minimization toward minimizing actual travel time losses.

  1. Traffic Prediction (Gaussian Processes): To identify a faulty sensor, the system must first establish a baseline of expected traffic. The authors use Gaussian Process (GP) regression with an Automatic Relevance Determination (ARD) kernel. GPs are ideal for this application because they provide both a mean prediction and a measure of uncertainty (variance), which is essential for statistical testing.
  2. Fault Detection (CUSUM): The system employs the Cumulative Sum (CUSUM) algorithm. Unlike point-in-time thresholds, CUSUM integrates deviations over time, making it highly sensitive to small, persistent shifts in sensor data—such as a 5% overcount—that might otherwise go unnoticed.
  3. Threshold Optimization: This is the core innovation. The authors define a loss function based on the difference in travel time between the route chosen using sensor data versus the route that should have been chosen. A random-restart hill-climbing algorithm is used to find the specific detection threshold ($\eta$) that minimizes this travel-time-based loss.
System architecture for the proposed optimal fault detection pipeline.

Datasets and Experimental Setup

The study utilizes the Caltrans Performance Measurement System (PeMS) database, focusing on 40 sensors in Downtown Los Angeles. Data was collected at 5-minute intervals over two weeks in September 2016. The first week was used to train the GP models, while the second week served for testing and evaluation. Since real-world ground truth for every failure is rarely available, the researchers simulated common fault types: Constant Relative Overcount (3-7%) and Conditional Undercount (7-13%).

Key Insights: Criticality and Application Awareness

A major takeaway of this research is that the optimal threshold for a detector should depend on the topology of the road network. Sensors located on "bottleneck" links require different tuning than those on redundant paths. The paper formalizes this through the concept of $\delta$-critical sensors—those where a failure results in an average travel time loss greater than a threshold $\delta$. This allows city planners to prioritize maintenance for the most influential parts of the network.

Furthermore, the authors explicitly quantify the cost of a False Positive. If a detector wrongly flags a healthy sensor, the routing engine switches to a GP-based prediction. If that prediction is less accurate than the (actually correct) sensor reading, the resulting route becomes suboptimal, leading to measurable travel time loss.

Results and Evaluation

The method was evaluated using OpenTripPlanner (OTP), an open-source multi-modal routing engine. The results showed that optimized dynamic thresholds significantly outperformed static baselines. For example, for sensor VDS 774685, the average loss with the optimized threshold was 16.2, compared to 31.2 using a standard static baseline—a nearly 50% reduction in travel time impact.

Performance trade-off curves showing False Positive vs. False Negative probabilities for different simulated fault types.

As shown in the performance curves, the system effectively distinguishes between different fault types, such as Conditional Undercounts (often caused by sensor saturation) and Constant Relative Overcounts (often caused by sensitivity issues), allowing for tailored optimal thresholds for each scenario.

Conclusion

By shifting the focus from statistical error minimization to application-specific performance, this framework provides a robust solution for managing faulty traffic sensors. The identification of critical sensors and the use of optimized thresholds ensure that urban navigation systems remain reliable and efficient. Future work will explore scaling this approach to city-wide networks and handling simultaneous sensor failures in localized areas.

Summary updated

Abstract

In a smart city, real-time traffic sensors may be deployed for various applications, such as route planning. Unfortunately, sensors are prone to failures, which result in erroneous traffic data. Erroneous data can adversely affect applications such as route planning, and can cause increased travel time. To minimize the impact of sensor failures, we must detect them promptly and accurately. However, typical detection algorithms may lead to a large number of false positives (i.e., false alarms) and false negatives (i.e., missed detections), which can result in suboptimal route planning. In this paper, we devise an effective detector for identifying faulty traffic sensors using a prediction model based on Gaussian Processes. Further, we present an approach for computing the optimal parameters of the detector which minimize losses due to false-positive and false-negative errors. We also characterize critical sensors, whose failure can have high impact on the route planning application. Finally, we implement our method and evaluate it numerically using a real-world dataset and the route planning platform OpenTripPlanner.
BibTeX
@inproceedings{ghafouri2017optimal,
  author = {Amin Ghafouri and Aron Laszka and Abhishek Dubey and Xenofon D. Koutsoukos},
  title = {{Optimal Detection of Faulty Traffic Sensors Used in Route Planning}},
  booktitle = {SCOPE@CPSWeek},
  year = {2017},
  url = {https://openreview.net/forum?id=mJZUYQKI8_T},
}
Figure