Components of Navigation
To get started with autonomous navigation, you’ll need a grasp of the following concepts:
Maps are topological graphs consisting of waypoints and edges. Maps are recorded either by operators using the tablet controller or by client applications using RPC calls to the GraphNavRecordingService.
Waypoints define the route a robot follows on map. Each waypoint stores a pointer to a snapshot of data captured during the recording process. During the recording process, robots place waypoints at approximately 2m intervals and otherwise as needed. A waypoint may be created explicitly via the API, either from the tablet or a client application.
Edges connect waypoints. They contain both a transform describing where one waypoint is with respect to another waypoint, and annotations that control how the robot should move between them. For example, annotations can indicate stairs mode or a speed restriction.
Localization is a process used by the robot to determine its location relative to a single waypoint in the graph. The robot updates its location continuously by comparing its real-time sensory data with the sensory data recorded in map waypoints. An initial localization is required, commonly aided by a fiducial.
When a map is initially loaded, GraphNav is unaware of the robot’s location relative to it and the user must help initialize the robot’s localization. If using fiducials, you can elect to initialize to a specific fiducial or a waypoint. If not using a fiducial, you must initialize to a particular waypoint, and may elect to provide an initial guess or search area.
Missions are expressed as behavior trees, which are composed of nodes arranged in a hierarchy. Behavior trees are executed inside the mission loop. Each iteration of a mission loop is referred to as a tick. For any given tick, mission execution starts at the root of the tree and proceeds down particular branches based on the type of node and the state of each node: failed, running, or success (see the NodeState Result enum defined in the
mission.proto file). A failed node does not necessarily result in a failed mission. For example, a node could check for low battery and ‘fails’ when the robot is fully charged.
The mission is complete once the root node has changed from running to failed or success.
See Missions service for details.
Advanced – Area Callbacks
Edges can be annotated with Area Callback regions, which specify that GraphNav should call out to a particular callback service in order to safely enter, cross, and exit that region. These services allow users to integrate new capabilities into GraphNav, such as checking whether it is safe to cross a crosswalk, opening doors, or flashing lights or playing sounds in a particular area.
See Area Callbacks for more information.