Sentiance Insights-Control-Tower Legacy Docs

Crash Detection

The Vehicle Crash Detection feature must be enabled for your app by Sentiance, before you can start using it. Please reach out to our support team to enable the feature.

App Integration

Follow the below steps to set up vehicle crash detection in you app.

1. Integrate the Sentiance SDK

If you haven't already integrated the Sentiance SDK, follow the steps in our Getting Started guide to complete the integration. Once you have completed those steps, continue along with the steps documented below.

2. [Android Only] Add the Crash Detection Dependency

To support crash detection, add a dependency to the crash detection library artifact in your app. This will make the CrashDetectionApi class available for you to subscribe for crash events.

app/build.gradle
dependencies {
    implementation(platform('com.sentiance:sdk-bom:<sentiance-version>'))
    implementation('com.sentiance:sdk-crash-detection')
    ...
}

3. Check if Vehicle Crash Detection Is Supported

There are several reasons why vehicle crash detection may not be supported on the device. The two most common of these are:

  • the feature is not enabled for your app;

  • the device lacks the necessary sensors (e.g. accelerometer).

You can check to see whether vehicle crash detection is support on the device, as follows.

CrashDetectionApi.getInstance(context).isVehicleCrashDetectionSupported

4. Subscribe for Vehicle Crash Events

In order to be notified of vehicle crash events, you must set a listener that the SDK can invoke whenever it detects vehicle crashes.

CrashDetectionApi.getInstance(context).setVehicleCrashListener(new VehicleCrashListener() {
    @Override
    public void onVehicleCrash(VehicleCrashEvent crashEvent) {
        // Handle the vehicle crash event

        long epochTimeMs = crashEvent.getTime();
        Location location = crashEvent.getLocation();
        Float speedAtImpact = crashEvent.getSpeedAtImpact();
        Float magnitude = crashEvent.getMagnitude();
        Float deltaV = crashEvent.getDeltaV();
        Integer confidence = crashEvent.getConfidence();
    }
});

The crash event contains the time and location of the detected crash, in addition to a number of metrics to estimate the severity of the crash:

Speed at impact

The estimated speed of the vehicle before the impact, in m/s.

Magnitude

The magnitude of the impact, in m/s².

Delta-V

The estimated change in velocity at impact, in m/s.

Confidence

The level of confidence that the accelerometer signal reflects a true crash pattern (range 0 - 100). It is recommended to filter out events below the confidence of 50.

5. Test Your Integration

The final step is checking your integration, to make sure that your vehicle crash listener is properly set up to handle crash events. Add the following method call in your app to trigger a dummy crash event.

CrashDetectionApi.getInstance(context).invokeDummyVehicleCrash();

This should invoke a dummy crash event, passing it to the listener that you previously set. You can then test how your app handles the event at runtime.

It's also possible to subscribe for additional diagnostic information that the SDK outputs at runtime, and receive a human-readable form of the state of the crash detector, which can be used to facilitate testing, when simulating a crash.

CrashDetectionApi.getInstance(context).setVehicleCrashDiagnosticListener { diagnostic ->
    val state = diagnostic.crashDetectionState
    val description = diagnostic.crashDetectionStateDescription
}

The result will reflect one of the following detection states:

  • crash candidate detected

  • crash candidate discarded - impact is too weak

  • crash candidate discarded - transport mode is not a vehicle

  • crash candidate discarded - pre-impact signal contains too much noise

  • crash candidate discarded - speed before impact is too low

  • crash candidate discarded - post-impact signal contains too much noise

  • crash candidate discarded - speed after impact is too high

Crash Report

A crash report contains contextual information pertaining to a crash event. This includes information specific to the crash i.e., time, location, speed before impact, a severity indicator and sensor data. Also, it contains user timeline information before, during, and after the crash event, as well as weather and traffic data. A single crash event produces one or more aggregative crash reports. Each new crash report is generated asynchronously as soon as new information is available on the Sentiance platform or on external data sources. So, we recommend to look for the latest file for a given crash event.

Crash reports are securely stored on Amazon S3. Reports will be generated in client-specific root folders with the following directory structure:

s3://sentiance-u1-crash-reports/<CLIENT_APP_ID>/<UTC_DATE>/<REPORT_ID>-<UTC_TIMESTAMP>.json.gz

  • CLIENT_APP_ID: The client app ID (e.g. 0000xxxxxxxxxxxxxxxxxxx).

  • UTC_DATE: The date of the crash event in UTC timezone in YYYY-MM-DD format.

  • REPORT_ID: A unique identifier for a crash report, belonging to a single crash event.

  • UTC_TIMESTAMP: Unix timestamp (UTC) in milliseconds of when the report was generated.

The procedure to access the crash report files in Amazon S3 is the same as that of offload files. Please refer to here.

A single crash report file may consist of multiple report segments. Each report segment contains different information related to the crash event.

  • CRASH EVENT: This segment contains the most basic information about a crash event and is usually generated first and as soon as possible after a crash event occurs (though this may change in the subsequent reports due to the asynchronous nature of the Sentiance platform).

  • TRANSPORTS: This segment contains information about one or more transports directly related to the crash event. Transports may include information like total distance traveled, trip waypoints, road types traveled, driving scores, and driving events.

  • TIMELINE: This segment contains information about the user’s timeline before, during, and after a crash event. This may include either stationaries (venues/locations where the user remained static) or other transports.

  • WEATHER: This segment contains information about the weather at the time and location of the crash event.

  • TRAFFIC: This segment contains information about traffic conditions (traffic incidents and the traffic flow) at the time and location of the crash event.

Here you can find sample data for each of those segments:

{
    "reportId": "xxxx556a-XXXX-41cd-xxxx-3930936a875f",
    "reportSegments": [
        "INITIAL"
    ],
    "reportTier": 3,
    "userId": "5f3256cab1xxxxxxxxxxxxxxx",
    "appId": "000000000xxxxxxxxxxxxxxx",
    "correlationId": "c7a1xxxx-xxxx-xxxx-xxxx-91xxxxb144aa",
    "sessionId": "D43DXXXX-xxxx-xxxx-xxxx-E54FCXXXX2EA",
    "externalUserId": "abc123",
    "processingTime": 1600439157581,
    "crashEvent": {
        "timestamp": 1599564413382,
        "waypoint": {
            "location": {
                "latitude": 53.08163,
                "longitude": 9.98515,
                "horizontalAccuracy": 65,
                "verticalAccuracy": 2,
                "elevation": 4,
                "provider": "GPS"
            },
            "timestamp": 1599564468547,
            "base": 0,
            "speed": -1.0,
            "direction": -1
        },
        "sensorData": {
            "type": "ACCELEROMETER",
            "baseTimestamp": 1599564399613,
            "baseOffsets": [
                0,
                40,
                80,
                ...
            ],
            "measurements": [
                [
                    -0.028,
                    -0.063,
                    0.036,
                    ...
                ],
                [
                    -0.294,
                    -0.337,
                    -0.373,
                    ...
                ],
                [
                    -0.812,
                    -0.83,
                    -1.083,
                    ...
                ]
            ]
        },
        "timezoneOffset": 120,
        "maxMagnitude": 100,
        "confidence": 75,
        "severity": null,
        "numberImpacts": null,
        "speedAtImpact": 25.2,
        "deltaV": 6.408,
        "crashEventOrigin": "AUTOMATICALLY_DETECTED",
        "roadType": "service",
        "speedLimit": 15
    },
    "timelineEvents": [],
    "weatherData": [],
    "trafficIncidentData": [],
    "trafficFlow": [],
    "isSensorDataInStore": true,
    "sensorDataReference": "8e1648ae-b866-4a34-8882-2a3ddfddf248",
    "transportEventId": "52d0f5cxxxxxxxxxxx78ee4a6cedcc73bcad3xxxxxxxxxxxd8f63c",
    "significantTransportMode": null,
    "significantStationaryIds": [],
    "atomId": null,
    "speedBeforeImpact": null,
    "speedAfterImpact": null,
    "speedBeforeImpactTimestamp": null,
    "speedAfterImpactTimestamp": null,
    "speeding": null,
    "speedingBeforeImpact": null,
    "phoneHandling": null,
    "phoneHandlingBeforeImpact": null,
    "harshEvents": null,
    "harshEventsBeforeImpact": null,
    "speedingDistance": null,
    "phoneHandlingSeconds": null,
    "hardAccelerationTimes": null,
    "hardBrakingTimes": null
}

The data model of the fields in the above sample segments of the crash report are explained in the attached file.

PreviousPrivacy Report & DashboardNextOn-Device Crash Detection Testing

Last updated