Two questions: where and how many
Transit agencies need to answer two basic questions about every vehicle. Where is it right now, and how many people are on board? Connected devices answer both, and the answers serve two audiences: riders who want to know when the next bus arrives, and planners who need to decide how often services should run.
Tracking vehicles
Most tracking relies on a satellite positioning receiver installed on the vehicle. It works out the vehicle’s position and sends it, along with the time, to a central system at regular intervals, often every few seconds to every half minute. The link is typically cellular IoT, because vehicles move across a wide area and need coverage everywhere they travel. Trains in tunnels, where satellite signals are weak, usually depend on other methods such as track-based detection or position estimates from known speed and distance.
On its own, a dot on a map is not very useful. The central system compares the position with the schedule and the route, calculates how late or early the vehicle is, and predicts arrival times at upcoming stops. Prediction improves when it uses past patterns, such as how long a certain stretch typically takes at rush hour. Some agencies apply machine learning to those patterns, though a simple average of recent trips for the same time and day can work reasonably well.
Agencies often publish the position data in a shared, standard format so that apps and websites can show live arrivals. The quality of what riders see depends on how clean the data is. Missed readings in tunnels, a vehicle assigned to the wrong route or a clock set incorrectly all cause errors, which is why data quality checks are part of a tracking system.
Counting passengers
Automatic passenger counting systems record boardings and alightings at each stop. Common sensing methods include:
- Infrared beam sensors at door steps, which register an interruption as someone passes and use the order in which two beams are broken to judge direction.
- Pressure mats on steps, which respond to weight.
- Overhead cameras or depth sensors above doors, which analyze the scene to count people entering and leaving. These image sensors can separate people walking side by side better than simple beams.
- Weight sensors on the vehicle’s suspension, which estimate total load rather than individual counts.
Each number is combined with the vehicle’s location and the door-open signal so the count belongs to a particular stop. Over a day, this builds up a record of boardings at each stop and load on each segment of the route.
What agencies do with the data
Counts and locations feed several decisions:
- Scheduling. If buses are consistently full on one stretch at certain hours, more frequent service might be considered. If they run mostly empty, a different schedule may be appropriate.
- Running time. Location history shows how long trips really take, and timetables can be adjusted to match.
- Reliability checks. Agencies can measure how often vehicles arrive on time and how evenly they are spaced. Buses that bunch together are a familiar problem, and live positions let controllers hold or space vehicles.
- Stop planning. Boarding numbers help decide which stops are busy, and which shelters or accessibility improvements should be prioritized, as part of broader urban planning.
- Rider information. Some systems show how full a vehicle is so riders can choose a less crowded trip.
Accuracy and limits
Passenger counts are estimates. In a crowded vehicle, people move in groups, carry bags or stand in the doorway, and sensors can double count or miss someone. Agencies often check automatic counts against manual counts on sample trips and adjust. Sensors also need periodic sensor calibration, and a unit knocked out of alignment can quietly produce poor data.
Location data has its own weak points. Satellite readings can wander near tall buildings, and communication gaps leave holes in the record. Predictions for the next stop are less reliable when traffic is unusual, such as during an accident or a major event.
Privacy considerations
Counting people is not the same as identifying them, and most systems aim only for counts. Still, camera-based counting raises concerns. Good practice is to process images on the device so only the count leaves the vehicle, an example of edge computing, and to avoid storing video unless there is a separate, stated reason such as safety. When ticketing or payment data is linked to journeys, the privacy stakes rise, and agencies are generally expected to limit retention and explain what they collect. Questions of IoT privacy and data retention apply here as in other connected services.
The bigger context
Transit data is one layer of the city’s connected systems, alongside traffic sensing and parking. Agencies increasingly combine it with other sources to understand how people move through a whole area. The value comes less from any one device than from consistent, well-checked data used to make steady, practical improvements to service. Results vary between agencies, because they depend on the local network, how well the equipment is maintained and how the data is used.