Chapter 18 - Trips, Stops, Routes, and Activity Analytics

Trips and stops are interpretations of observations. A delivery vehicle, runner, pedestrian, and asset beacon require different thresholds. The platform should provide configurable profiles and preserve the raw evidence behind each derived result.

Segmentation

A basic segmentation pipeline considers:

• session boundaries;
• assignment changes;
• long reporting gaps;
• movement-state transitions;
• ignition or activity signals when available;
• maximum plausible distance jumps;
• explicit task start and completion.

A trip may begin after sustained movement and end after a dwell threshold. A stop is not simply a point with speed zero; GPS jitter and missing updates make that unreliable.

Profile-specific thresholds

Example profiles:

These values are illustrative. Field data and product expectations determine the real defaults.

Trip model

trips
  id
  organization_id
  subject_id
  session_id
  started_at
  ended_at
  start_location_id
  end_location_id
  distance_m
  moving_duration_s
  elapsed_duration_s
  max_speed_mps
  average_speed_mps
  algorithm_name
  algorithm_version
  generation
  quality_summary

Stops record center, radius, arrival, departure, dwell, and confidence. Use a robust centroid or representative point rather than averaging latitude and longitude blindly over large areas.

Distance

Distance calculation should ignore obviously invalid segments and account for accuracy. For each segment:

1. ensure timestamps increase;
2. compute geodesic distance;
3. estimate implied speed;
4. apply activity and accuracy policy;
5. classify accepted or excluded distance;
6. accumulate both totals for diagnostics.

Store the algorithm and thresholds used. Map matching to a road network can improve vehicle distance and route display, but it adds external data, computational cost, and failure modes. It should be optional and never overwrite raw measurements.

Elevation and pace

For sport activities, altitude from phones can be noisy. Elevation gain should use smoothing and thresholds. Document whether altitude comes from GPS, barometer, or a later terrain model.

Pace is inverse speed over meaningful intervals. Avoid presenting instantaneous pace from one noisy sample. Use a rolling window and show data quality.

Route representation

A route can have several representations:

• ordered raw point IDs;
• a detailed polyline for close playback;
• simplified polylines for zoom levels;
• bounding box and start/end points;
• aggregate statistics.

Generate representations asynchronously after a trip is stable. A live route can be incremental and later reconciled.

Playback

The playback API returns points or a polyline with timestamps in bounded chunks. The client interpolates between observations but indicates gaps. Do not animate smoothly across a twenty-minute network outage as if the exact path were known.

Provide controls for:

• speed multiplier;
• pause and seek;
• event markers;
• quality flags;
• task stops;
• geofence transitions;
• battery and network timeline.

Late data and generations

A late batch can change a completed trip. Use a reconciliation window, such as several hours or a day, depending on device behavior. Recompute affected intervals into a new generation. The current generation is selected for normal views; older generations remain available for audit when required.

After the reconciliation window closes, very late points may be stored but not automatically change business reports without an explicit correction job.

Activity summaries

Daily and session summaries reduce expensive repeated scans:

activity_summaries
  subject_id
  period_start
  period_end
  activity_type
  distance_m
  moving_duration_s
  elapsed_duration_s
  trip_count
  stop_count
  max_speed_mps
  elevation_gain_m
  algorithm_version

Use jobs to update summaries idempotently. Reports read summaries for long periods and raw history only for drill-down.

Chapter checklist

Derived mobility analytics should:

• use profile-specific, versioned thresholds;
• split on sessions, gaps, and provenance changes;
• account for accuracy and implausible segments;
• preserve raw observations;
• provide multiple route representations;
• expose gaps honestly in playback;
• reconcile bounded late data through generations;
• use summaries for long-range reporting.