Tango 开发指南 — 运动追踪 API（Java 描述）

生命周期

配置

获取 pose 数据
回调方式

时间戳方式

以下是原文

Lifecycle

Configuration

Getting pose data
Callback-based

Polling-based

原文链接：https://developers.google.com/tango/apis/java/java-motion-tracking

本文演示如何通过 Java 来使用运动追踪技术。

生命周期

运动追踪系统一般情况下包含三个状态：TangoPoseData.POSE_INITIALIZING, TangoPoseData.POSE_VALID 以及 TangoPoseData.POSE_INVALID。POSE_INITIALIZING 表示系统还未就绪，无 pose 数据可用。POSE_VALID 表示系统运转正常。POSE_INVALID 表示系统异常，需要重新初始化。还有一个状态，POSE_UNKNOWN，表示出上述状态之外的其他状态。

如果 pose 数据变成 POSE_INVALID 状态，运动追踪系统可以通过两种方法来重新初始化。如果系统配置项 config_enable_auto_recovery = true，系统将立即进入 POSE_INITIALIZING 状态。恢复之后，系统使用最近一次 POSE_VALID 状态的 pose 作为追踪的起始点。如果 TangoConfig.KEY_BOOLEAN_AUTORECOVERY = false，系统将停止工作，并始终返回 POSE_INVALID 状态的 pose 数据，直到 Tango.resetMotionTracking() 方法被调用。与自动恢复不同的是，系统恢复后该方法将重置起始点为故障前的原始点。

更多信息，请参考设备 Pose 原理页面。

配置

要使用运动追踪功能，TangoConfig 中的 KEY_BOOLEAN_MOTIONTRACKING 必须设置为 true。默认的 TangoConfig 中 KEY_BOOLEAN_MOTIONTRACKING = true。

除此之外还得设置 KEY_BOOLEAN_AUTORECOVERY。该项默认为 true。该项表示的意义请查看生命周期章节。

获取 pose 数据

对于基本的运动追踪功能，有两种坐标帧对可供使用：设备帧&服务起始帧，设备帧&前设备帧。对于第一种，设备 pose 数据是相对于运动追踪服务开启时的位置，该模式下 pose 数据可以通过回调和时间戳参数获取。后一种，获取的设备 pose 数据是相对于设备之前某个位置。该模式下只能通过回调获取。

回调方式

在回调方式中，你必须定义对应的坐标帧对，并实现 Tango.OnTangoUpdateListener 接口中的三个方法：onPoseAvailable()，onXyzIjAvailable()，onTangoEvent()。

private void setTangoListeners() {
final ArrayList<TangoCoordinateFramePair> framePairs = new ArrayList<TangoCoordinateFramePair>();
framePairs.add(new TangoCoordinateFramePair(
TangoPoseData.COORDINATE_FRAME_START_OF_SERVICE,
TangoPoseData.COORDINATE_FRAME_DEVICE));

// Listen for new Tango data
mTango.connectListener(framePairs, new OnTangoUpdateListener() {

@Override
public void onPoseAvailable(final TangoPoseData pose) {
// Process pose data from device with respect to start of service
}

@Override
public void onXyzIjAvailable(TangoXyzIjData arg0) {
// We need this callback even if we don't use it
}

@Override
public void onTangoEvent(final TangoEvent event) {
// This callback also has to be here
}
});
}

时间戳方式

这种方式中，你必须先设置合适的帧对。对于简单的运动追踪，使用 TangoPoseData.COORDINATE_FRAME_DEVICE & TangoPoseData.COORDINATE_FRAME_START_OF_SERVICE 就行。

// Define what motion is requested.
TangoCoordinateFramePair frames_of_reference;
frames_of_reference.baseFrame = TangoPoseData.COORDINATE_FRAME_START_OF_SERVICE;
frames_of_reference.targetFrame = TangoPoseData.COORDINATE_FRAME_DEVICE;

然后调用

Tango.getPoseAtTime()

。下面的例子中，时间戳参数是 0.0 来获取最近的一次的 pose 数据。如果设置成其他值，将获得对应时刻的封装好的 pose 数据。这里的时间戳是以设备启动时刻为起始点的。

new Thread(new Runnable() {
final int pollingUpdatePeriodMilliseconds = 66;

@Override
public void run() {
while (true) {
try {
Thread.sleep(pollingUpdatePeriodMilliseconds);
} catch (InterruptedException e) {
e.printStackTrace();
}
try {
final TangoPoseData queryPoseStartDevice =
mTango.getPoseAtTime(0.0, frames_of_reference);
} catch (TangoErrorException e) {
e.printStackTrace();
}
}
}
}).start();

以下是原文

This page describes how the Java API handles motion tracking.

Lifecycle

The normal motion tracking system lifecycle consists of three states: TangoPoseData.POSE_INITIALIZING, TangoPoseData.POSE_VALID, and TangoPoseData.POSE_INVALID. In the POSE_INITIALIZING state, the system is not yet ready and pose data is not available. In the POSE_VALID state, the system is functioning normally. In the POSE_INVALID state, the system believes its estimate was invalid and needs to be reinitialized. A fourth state, POSE_UNKNOWN, is used for all other cases.

Should the pose data become POSE_INVALID, the motion tracking system can be reinitialized in two ways. If config_enable_auto_recovery was set to true, the system will immediately enter the POSE_INITIALIZING state. It will use the last valid pose as the starting point after recovery. If TangoConfig.KEY_BOOLEAN_AUTORECOVERY was set to false, the system will essentially pause and always return poses as POSE_INVALID until Tango.resetMotionTracking() is called. Unlike auto recovery, this will also reset the starting point after recovery back to the origin.

The lifecycle state is recorded in the TangoPoseData object’s statusCode.

For more information, although the page is based around the C API, please see our Device Pose concepts page.

Configuration

In order to use motion tracking, your TangoConfig must have KEY_BOOLEAN_MOTIONTRACKING set to true. If you are using the default TangoConfig as your starting point, it is already set to true.

You also have the option to set KEY_BOOLEAN_AUTORECOVERY. In the default TangoConfig, this is set to true. See the Lifecycle section for the behavior of this parameter.

Getting pose data

There are two coordinate frame pair options for basic motion tracking: device with respect to start of service, and device with respect to the previous device pose. With start of service, the device’s pose is relative to the position where the motion tracking system initialized. You can receive pose data in both the callback and polling forms. With previous device pose, the device’s pose is relative to its last position. Pose data is only available as a callback.

Callback-based

If you are using the callback-based approach, you must define the coordinate frame pairs you are interested in and construct your onPoseAvailable() callback.

You also have to implement onXyzIjAvailable and onTangoEvent, because the Tango.OnTangoUpdateListener interface requires an implementation for all three of those methods.

private void setTangoListeners() {
final ArrayList<TangoCoordinateFramePair> framePairs = new ArrayList<TangoCoordinateFramePair>();
framePairs.add(new TangoCoordinateFramePair(
TangoPoseData.COORDINATE_FRAME_START_OF_SERVICE,
TangoPoseData.COORDINATE_FRAME_DEVICE));

// Listen for new Tango data
mTango.connectListener(framePairs, new OnTangoUpdateListener() {

@Override
public void onPoseAvailable(final TangoPoseData pose) {
// Process pose data from device with respect to start of service
}

@Override
public void onXyzIjAvailable(TangoXyzIjData arg0) {
// We need this callback even if we don't use it
}

@Override
public void onTangoEvent(final TangoEvent event) {
// This callback also has to be here
}
});
}

Polling-based

In the polling-based approach, you must first specify the coordinate frame pair you are interested in. For simple motion tracking, this will always be TangoPoseData.COORDINATE_FRAME_DEVICE with respect to TangoPoseData.COORDINATE_FRAME_START_OF_SERVICE.

// Define what motion is requested.
TangoCoordinateFramePair frames_of_reference;
frames_of_reference.baseFrame = TangoPoseData.COORDINATE_FRAME_START_OF_SERVICE;
frames_of_reference.targetFrame = TangoPoseData.COORDINATE_FRAME_DEVICE;

Then call `Tango.getPoseAtTime() as desired. In the example below, the timestamp is set to 0.0 to get the latest pose. If you specify a specific timestamp, the system will return an interpolated pose at that exact time. Timestamps are relative to the device boot.

new Thread(new Runnable() {
final int pollingUpdatePeriodMilliseconds = 66;

@Override
public void run() {
while (true) {
try {
Thread.sleep(pollingUpdatePeriodMilliseconds);
} catch (InterruptedException e) {
e.printStackTrace();
}
try {
final TangoPoseData queryPoseStartDevice =
mTango.getPoseAtTime(0.0, frames_of_reference);
} catch (TangoErrorException e) {
e.printStackTrace();
}
}
}
}).start();