Bus Sensor Networks

1 Overview and operating context

A bus carries two overlapping sensor families. The first is the powertrain and chassis data shared with heavy trucks, broadcast on SAE J1939 CAN at 250 kbit/s (engine, transmission, brakes, tyres, aftertreatment). The second is passenger-service data specific to buses: per-door open state, automatic passenger counting (APC), kneeling and ramp, stop request, HVAC, GNSS for schedule adherence and, on electric buses, the battery management system (BMS) with a 400-800 V high-voltage pack.

Powertrain variants covered: diesel, CNG, hybrid and battery-electric; service types covered: urban transit (stop every 300-500 m) and intercity (stops every 20-100 km). A telematics gateway reads the buses through a read-only FMS (Fleet Management System) interface (Bus FMS 2007; FMS 3.0 of 2012 harmonised bus and truck) and uploads over cellular. Availability of individual FMS signals varies by vehicle model [S].

Fleet-scale framing from the bridge research [E]: a city bus operating 8-16 h per day produces about 2-10 records per second at the gateway, uploaded as one publish per 10-60 s, roughly 5-40 MB per day as JSON or 1-8 MB per day as binary or compressed payloads, or about 150-1000 MB per month.

2 Sensor inventory

Period values are bus broadcast periods ([K]) unless noted. Telematics export cadence is described in section 7. "Source bus" is the network on which the value normally appears.

3 On-vehicle networks and protocols

Bridge layer between the bus and the cloud

Telematics reporting from the research notes: 1-30 s while moving, 1-15 min heartbeat while parked, store-and-forward on cellular loss [K].

4 Duty cycles and operating profiles

• Diesel: idle 600-800 rpm, 1.5-3 L/h, 20-35% of engine hours; tank 250-350 L refuelled nightly; DEF 20-40 L.
• Electric: SOC 95% at 05:00 falling to 20-30% by evening over 200-250 km; optional midday opportunity charge; overnight depot charge 3-6 h at 50-150 kW; battery pre-heat at depot; winter HVAC drains SOC faster.
• Weekend service is 60-70% of weekday. Passenger load follows the peaks, spikes at hubs, and is zero at the end of line.
• GNSS dropouts occur in tunnels, depots and urban canyons.
• Power: key-off sleep with CAN or accelerometer wake; cranking brownout below 6 V reboots 0.1-1 per day on weak-battery vehicles [E].

5 Faults and anomaly signatures

All injection rates are [K] estimates unless noted; bridge-layer rates are [E].

6 Events, alerts and thresholds

Note on door signals: door switches distinguish holding time from dwell, and some fleets drive with doors open, making that signal useless on those fleets [S TCRP].

7 Generator guidance for the simulator

8 How the Sense simulator implements this

8.1 Scenario codes (tenant subvariant)

The Fleet tenant has the variant bus and a subvariant that selects the scenario. The codes are the ones the Controller and the platform use. A tenant with another subvariant (or none) runs no bus; the simulator's own selector (default staff-shuttle) is used only when the tenant's subvariant is empty. Table 7.1 above covers city transit, intercity and electric fleet; the hospital staff shuttle is a fourth scenario that Table 7.1 does not list.

8.2 Protocols emitted and the one-vehicle-per-tenant model

Wialon input bits: 0 ignition, 1 front door, 2 rear door, 3 panic (shuttle), stop request (city, coach) or plugged in (electric). Params are name:type:value (1 integer, 2 decimal). J1939 values are quantised to the J1939-71 resolutions listed in section 7. NMEA and CAN are sent only while the ignition is on; CAN slow groups every 10 s and the VIN every 600 s in all four scenarios.

8.3 Signals actually emitted

"Diesel" means staff-shuttle, city-transit and intercity-coach; "all" includes electric-fleet. Sense makes one reading per J1939 signal and per Wialon param, named as in the table.

DM1 codes sent (amber lamp): diesel SPN 110 FMI 16 (coolant over 102 degC), SPN 1761 FMI 18 (DEF under 10%), SPN 117 FMI 1 (air pressure under 600 kPa); electric SPN 520193 FMI 0 (cell over 50 degC), SPN 520192 FMI 1 (SOC under 10%), SPN 117 FMI 1. Degraded GNSS is sent as 3-4 satellites, HDOP 3-10, 10-50 m position error with jumps over 100 m; a tunnel gives fix quality 0 and RMC status warning.

8.4 What Sense can and cannot read yet

The simulator document lists these as Sense-side gaps; Sense itself was not changed by the simulator work. The state below follows BUS_POINTS.md section 8, the later TRUCK_POINTS.md section 8 and Sense's BUILD.md.

• Read by the working copy of Sense, not by the released build of 12 Oct 2026: J1939 TIRE1 (65268, one point per tyre) and B1 (65274, brake pressures); NMEA GGA HDOP and the RMC time; the Wialon record date and time (kept as the source time); a Wialon unit for the names in Table 8.5 through a name table (unknown names stay without a unit). The setting common.use_source_time is off by default.
• Electric bus battery, motor and charger: no J1939 high-voltage energy-storage group and no OBD-II PID 5B in Sense, so these travel only as Wialon params.
• Not read at all: APC, HVAC and door-control signals (no standard group; carried as Wialon params), the tachograph state beyond speed, string params and the iButton field of Wialon IPS, the native UDP form of Wialon IPS 2.0 (only the session form, login plus data lines in each datagram, is read), Teltonika over UDP, and the J1939 vehicle number (65260). Sense also does not use seq or boot_id, so duplicates and replayed old records count as new readings.
• Not reachable from the simulator: MQTT, HTTP JSON, HL7 and TCP trackers; the simulator sends UDP only.
• Vehicle profiles: the Fleet compatibility check compares what is heard with a profile chosen by the tenant variant and, for a bus, by the subvariant (profiles exist for staff-shuttle, city-transit, intercity-coach and electric-fleet; an empty or unknown subvariant uses the plain Bus profile).

8.5 Faults and the ground-truth file

Each bus carries latent traits (device clock error in ppm, a chronic flag in 1% of buses with 5% frame loss and bus-off 5 a day, frame loss 0.01-0.5%, bus-off 0.05-0.5 a day, a weak battery in 5% with a reboot on 40% of ignition-on events, APC miss rate 1-5%). Fault rates are those of Tables 5.1 and 5.2, scaled by --bus-faults (100 = as in the report). Injected faults include the J1939 not-available block after ignition on, spikes, x10 encoding errors, dropped and duplicated CAN frames, bus-off, VIN transport aborts, stuck-at and drift, GNSS episodes, jumps and cold-start teleport, link outages with store-and-forward replay, reboots (seq restarts, boot_id +1), door stuck, APC miscount and stuck at zero, fuel theft, TPMS leak, engine overtemperature, and for electric buses SOC steps, cell temperature sensors reading -40 or 215 and BMS dropouts.

With --truth FILE the simulator writes one JSON object per line. Every line has t (simulated seconds since start), utc, vehicle and type; the first lines are run (seed, clock, options) and one vehicle per bus (scenario, IMEI, VIN, source, clock error, block hours). Then one line per event, for example stop (true in and out, dwell), apc_error, link_down and link_up (cause dead_zone, tunnel or dropout), replayed, buffer_overflow, frame_dropped, bus_off, tp_abort, j1939_na, spike, encoding_error, stuck_at, drift, gps_jump, reboot, door_stuck_start, fuel_theft, engine_overtemp_start, soc_recalibration_step, bms_comms_dropout and end. The file is the ground truth against which a detector can be scored. Not simulated: multi-code DM1 transport failures, wrong CAN bit rate, firmware or schema mismatch, clock jump, odometer reset, reconnect storms from many buses, and signals absent by tenant.

8.6 Measured differences between the scenarios

Headless runs of 12 Oct 2026 with a fixed seed and simulated clock 60 (staff-shuttle 2 buses for 3 simulated hours, city-transit 4 buses for 4 h, intercity-coach 2 buses for 9 h, electric-fleet 3 buses for 4 h). Every datagram (41,057 / 86,949 / 74,848 / 57,681 / 106,035 over the runs including a night run) was identified by Sense's protocol handler as NMEA 0183, OBD-II, J1939 or Wialon IPS, with none unknown or unreadable. The table is taken from NMEA and the truth file.

Night run: three electric buses charged 4.4-5.5 h at 71-92 kW (266-342 kWh); the diesel buses pulled in and refuelled; no NMEA datagram was heard inside any outage interval of the truth file. The coach covered about 340 km per bus in 9 h. Two tenants at once (staff-shuttle and city-transit on separate Sense instances and port triples) each heard only their own IMEI and address.

9 Uncertainty and assumptions