Cyclists could get the ‘green wave’ at traffic signals with mobile app

The latest report to come out of this multi-project research effort, led by Dr Stephen Fickas of the University of Oregon (UO), introduces machine-learning algorithms to work with their mobile app FastTrack. 

Developed and tested in earlier phases of the project, the app allowed cyclists in Eugene, Oregon, to communicate with traffic signals along a busy bike corridor. The researchers hope to make their app available in other cities.

“Our overall goal is to give cyclists a safer and more efficient use of a city’s signalled intersections,” Fickas explained. “The current project attempts to use two deep-learning algorithms, LSTM and 1D CNN, to tackle time-series forecasting. The goal is to predict the next phase of an upcoming, actuated traffic signal given a history of its prior phases in a time-series format. We’re encouraged by the results.”

Their latest work builds on two prior projects in which Fickas and his team successfully built and deployed a hardware and software product called ‘Bike Connect’ which allowed people on bikes to give hands-free advance information to an upcoming traffic signal, using their speed and direction of travel to increase the likelihood the signal would be green upon arrival.

For the latest study, the researchers explored two separate machine-learning algorithms. Both have a good track record with time-series forecasting: One-Dimensional Convolutional Neural Nets (1D CNN for short) and Long Short-Term Memory models (LSTM for short). 

To measure the effectiveness of each algorithm, they used three metrics. First, precision is concerned with 'When the model predicts the rider will arrive at a green light, how often is it correct?' Second, recall asks, 'For all the actual green lights the rider encountered, how many did the model get correct?' Thirdly, accuracy refers to the number of correct predictions.

The researchers found that LSTM and 1D CNN scored nearly identical results on all three metrics. The researchers could also predict the next phase with roughly 85 per cent accuracy, for each of the time-series forecasting algorithms. 

“We believe we are in the ballpark of being acceptable to add a prediction component to our existing FastTrack app,” Fickas said. This would open up ‘green wave’ capability for non-fixed-time intersections. 

The researchers next plan to gain access to a dataset with a larger range of days, perhaps an entire season. The team has its eyes on 'Better Naito' Parkway in Portland, Oregon - a bike-friendly corridor that contains multiple actuated intersections from which to draw data. Typically, more data leads to stronger results when looking at machine-learning algorithms. They also intend to move to a multivariate dataset that includes date and time and perhaps weather as well. This would not be a vast change to data preparation and may allow a single model that covers all four seasons.

According to the team, the FastTrack app requires a real-time feed from upcoming traffic signals on the cyclist’s path. Cities with older equipment or with older Traffic Management Systems (TMS) may not provide this feed. Fickas and his team remain optimistic: as cities replace older equipment and bring on a modern TMS, they will be fully capable of using a FastTrack app that is effective with both fixed and actuated intersections, giving their biking community green wave opportunities.