University of Minnesota Driven to Discover
U of MNUniversity of Minnesota
Center for Transportation Studies

Inductive Loop Detector Signal Analysis

Presentation by Prof. Stanley Burns, Dept. of Electrical and Computer Engineering, U of M Duluth

October 21, 2003

Inductive loop detectors (ILDs) embedded in pavements are among the most widely deployed tools for monitoring traffic on highway networks. They are commonly used to measure traffic flow by registering each time a vehicle passes over them. But researcher Stan Burns of the Northland Advanced Transportation Systems Research Laboratories at the University of Minnesota's Duluth campus thinks that we could get much more information from the existing ILD network.

A typical ILD consists of a solenoid loop of wire buried about two inches below the surface of the pavement with its symmetric axis perpendicular to the road. When a vehicle passes through the detector's magnetic field, its metal structure causes a disturbance in the electrical inductance of the loop. A sensor attached to the loop registers an event each time the inductance change exceeds a threshold value.

However, the actual inductance curve produced by a vehicle is not just a simple spike followed by a return to the baseline value. An automobile's complicated arrangement of metal parts produces a correspondingly complicated set of inductance changes as it passes over the detector. In effect, the traditional method of recording only the crossing over a threshold value discards this potentially valuable information.

Another side effect of the threshold-value system is that large vehicles—particularly commercial trucks—may be recorded as two separate vehicles rather than one large one. This error is produced when the inductance reading from the coil pushes up over the threshold value as the front of the vehicle passes over, then drops below the threshold as the less-metallic central portion passes, then surges over the threshold again when the rear axle passes over the detector. Considering the number of tractor-trailer rigs using the freeway system, this can lead to significant counting errors.

An ILD system capable of capturing the "inductance signature" of different types of vehicles would reduce counting errors due to inaccurate readings of large vehicles. Such a system could also enable traffic managers to calculate more accurate point-to-point travel times by tracking individual vehicles.

The experimental system developed by the NATSRL research team consists of three interdependent software modules:

  • The Sampler module gathers data by controlling the sensing instruments, and saves the inductance profiles for identification. A variety of signal processing algorithms are also available to prepare the data for identification, including normalization, clipping, computing the derivative of the data, fast Fourier transform, and finding a polynomial fit for the inductance curve.
  • The Viewer, as its name implies, allows users to view one or more inductance profiles on a graph after user-selected signal processing algorithms have been applied.
  • The Identifier module classifies vehicles by comparing inductance curve data from the Sampler with inductance profiles stored in a database. The specific procedure employed is to compare the coefficients of the polynomial fit of the sampled curve to stored curves.

Burns and the NATSRL researchers have installed the experimental system at a Mn/DOT test station on Interstate 35W south of Duluth. At this stage, the system is able to distinguish between different vehicle types, although it still has trouble discriminating between cars with similar inductance signatures.

The test results are an indication of the complicated nature of inductance-signature analysis. In addition to vehicles with similar construction (was that a Ford or a Chevy?), a detector system based on inductance curves must also contend with numerous environmental factors that effect each loop's three-dimensional magnetic field. These include interference from vehicles in adjacent lanes and from nearby metallic structures such as road signs or steel reinforcing rods in the roadbed.

Work on inductive loop detectors will continue at NATSRL, including three-dimensional modeling of the loop's magnetic field, exploration of "cross-talk" interference from adjacent loop detectors' fields, improving system accuracy, experimental verification, and the possibility of measuring vehicle speed using a single ILD.