Bias Estimation of Spatiotemporal Traffic Sensor Data with Physics-Informed Deep Learning Techniques
Advanced Search
Select up to three search categories and corresponding keywords using the fields to the right. Refer to the Help section for more detailed instructions.

Search our Collections & Repository

For very narrow results

When looking for a specific result

Best used for discovery & interchangable words

Recommended to be used in conjunction with other fields



Document Data
Clear All
Clear All

For additional assistance using the Custom Query please check out our Help Page


Bias Estimation of Spatiotemporal Traffic Sensor Data with Physics-Informed Deep Learning Techniques

Filetype[PDF-3.46 MB]



  • Creators:
  • Corporate Creators:
  • Subject/TRT Terms:
  • Resource Type:
  • Geographical Coverage:
  • Corporate Publisher:
  • Abstract:
    Efficient operations of intelligent transportation systems rely on high-quality traffic data. Infrastructure-based traffic sensors, though providing major data sources for ITS, are subject to data quality issues. Existing studies have attended to identifying malfunctioning sensors or recovering missing data. Nevertheless, critical gaps remain to be addressed. Firstly, most studies only attempt to label sensors as either ‘good’ or ‘bad’. In this way, any useful information contained in the partially ‘bad’ sensors is always discarded while the potentially erroneous information given by the partially ‘good’ sensors is always preserved. Secondly, the traffic dynamics attributes have not been effectively exploited when examining the sensor data, which is a missed opportunity for utilizing valuable information. This dissertation will try to fill these research gaps. In the dissertation, we first construct three networked sensor error correction models using transportation domain knowledge and Physics-Informed Deep Learning (PIDL) techniques based on fully-connected Neural Network, Convolutional Neural Network (CNN), and Long Short-Term Memory (LSTM), to extract features from the spatiotemporal sensor data and quantify the traffic flow measurement biases for all the sensors in the network. In addition, narrowing the scope down to a traffic segment with only two sensors, we develop another measurement error correction model using the Physics-Informed LSTM neural network combined with prior knowledge of macroscope traffic models. With minimum data requirements, this LSTM-based PIDL model is able to correct the measurement biases for traffic flux and occupancy simultaneously for both sensors. Overall, experimental results demonstrate the merits of combining machine learning techniques with domain knowledge of the physics of traffic flows in the context of sensor health monitoring and error estimation. The sensitivity analyses demonstrate the reliability and robustness of our results with respect to different testing scenarios.
  • Format:
  • Collection(s):
  • Main Document Checksum:
  • Download URL:
  • File Type:

Supporting Files

  • No Additional Files
More +

You May Also Like

Checkout today's featured content at