Speaking to engineering teams across the country, we’ve found that 50% of signal operations problems are a result of faulty detection. Detector health should be one of the first questions to answer when a complaint comes in and the Flow Labs detector health suite provides engineers with those answers quickly, easily, and proactively.

A Faulty Detector isn’t always a Failed Detector

Traditional ATMS reporting treats a detector as working or failing - but can’t pick up situations where a detector is counting 52 vehicles where 100, or situations where a detector might be pointing at the wrong lane. These are the types of issues that cause persistent problems for a signal system. On average, a signal issue costs road-users $1,000 per day when left unchecked and the more silent these issues are, the more those numbers can build up across an entire signal network.

Using Integrated Data for Detection Accuracy

Luminus Plus merges probe data with high-resolution controller data and detection data, using algorithms to identify discrepancies between them and identify accuracy issues. Detection gets compared against an independent measure of what moved through the intersection, in real time.

That comparison produces four measures:

• Count accuracy shows the accuracy of vehicle counts.
• Presence accuracy shows how often a vehicle passing through a detection zone is picked up by the detector (and conversely how many vehicles get missed).
• Calibration accuracy shows whether a detector’s arrival pattern is consistent with what it should be, helping to catch a detector channel that has been reassigned without documentation, or a detector that has drifted onto the wrong lane.
• Consistency index shows the consistency of a detector - a major issue is that detection often stops working on one day, comes back the next, leaving technician chasing their tails.

There are a number of additional metrics available in the Flow Platform, and we combine these into a detector health score, a single number per detector that allows engineers to assess the quality of their detection.

View Detector Health Alongside SPMs

Detector health is now visible in Abacus alongside other signal performance measures, so when you pull up an approach with a split failure problem you can see whether the detector connected to that phase is worth trusting before you change the split.

Create a Priority Lists for your Technicians

Cerberus scans every network, intersection and detector and prioritizes what it finds. Sort by detector health and you have a maintenance list ordered by detection health rather than by which one somebody complained about loudest. Export it, email it and hand it over at the start of the month, then check the same list four weeks later to see how much of it got fixed.

Proactively Monitor your Detection

Insights is a standard feature in Luminus and Luminus Plus. Engineers can create and customize alerts based on pre-configured rules and get issues into your inbox as they happen.

We’ve now added detection health metrics into the options for Insights so that engineers can create alerts that combine them with signal performance measures. This makes it easy to surface only the detection issues that are actually degrading signal performance.

This helps to eliminate asset issues that aren’t causing an operational problem. For example, a presence detector might be unhealthy but it is at a coordinated phase which is on max recall so it won’t impact performance. Insights can now remove these from the alert list automatically.

Scan your Detection Today

Detector health is already available as part of Luminus Plus, our integrated signal performance measures platform (ISPMs).

‍The below chart shows a histogram of detection health for a current state agency customer. You can quickly see the distribution of detection quality across their network. Approximately 5.1% of detectors are unhealthy (detector health score less than 50), 74.5% are healthy (detector health score greater than 75), and the rest are borderline. With the detector health suite agencies can see the extent of their detecton problems and pinpoint their locations exactly.

‍

To run detector health on your own network, email our team at contact@flowlabs.ai for a demo.