MoveIt Task Constructor (MTC)

The MoveIt Task Constructor (MTC) is a framework designed to define and plan complex actions composed of multiple interdependent subtasks. It leverages MoveIt's planning capabilities to solve individual subproblems within 'black-box' planning stages.

Key features include:

• Interdependent Subtasks: Define sequences or parallel compositions of actions.
• PlanningScene Interface: Uses MoveIt's PlanningScene as a common interface to pass solution hypotheses between different stages.
• Hierarchical Organization: Uses containers to organize basic stages into complex task pipelines, supporting both sequential and parallel execution.

The MoveIt Task Constructor (MTC) framework is a tool for defining and planning complex actions composed of multiple interdependent subtasks. It works by breaking down a high-level task into individual planning stages that solve subproblems using MoveIt's planning capabilities.

Key features include:

• Black-box planning stages: Individual subproblems are solved independently.
• PlanningScene interface: Uses MoveIt's PlanningScene to pass solution hypotheses between stages.
• Hierarchical organization: Uses containers to organize stages, supporting both sequential and parallel compositions.

In complex tasks like 'pick and place', the second part of the task (placing) depends on the state left by the first part (grasping), such as having an object attached to the gripper.

To link these non-linear pipelines, use a MonitoringGenerator. Instead of being completely independent, a MonitoringGenerator monitors the solutions of another stage in the pipeline and 'fast-forwards' them so the dependent stage can continue processing from that specific state.

Planning in the MoveIt Task Constructor (MTC) does not proceed linearly from top to bottom. Instead, it proceeds from the inside out.

When you use Connect stages, the planner computes a motion plan between two previously calculated subordinate solutions. To visualize how these stages connect and how solutions are found, use the MTC RViz panel.

Properties are named attributes of a stage used to configure behavior and control substages. A property maintains two values: a current value and a default value.

To work with properties, you can define them with a name, a description, and an initial value. Before using a property, it is good practice to check if a current value has been defined. You can then safely retrieve the value using the property object.

# Example of defining and using a property
# (Note: Actual implementation details are in the demo scripts)
# property = stage.property("name", "description", value)
# if property.has_value():
#     val = property.get_value()

MoveIt Task Constructor (MTC) decomposes complex motion planning problems into a hierarchical structure of Tasks and Stages.

• Task: The top-level planning problem container.
• Stages: The individual subproblems that compose the task. Stages can be arranged in arbitrary hierarchies and orders, provided their data flow is compatible.

Planning follows a specific lifecycle: processing starts at Generator stages, expands through Propagator stages, and is finalized by Connector stages that bridge the gaps between states.

The merger stage is used to plan and execute multiple task sequences in parallel.

# Example initialization of merger stage
# (Refer to merger.py for full implementation details)
initAndConfigMerger()

The MoveRelative stage uses planners to move the robot along a specified relative offset.

# Example of relative motion
# (Refer to cartesian.py for full implementation details)
initAndConfigMoveTo()
initAndConfigMoveRelative()

Use the Connect functionality to bridge two stages by finding a valid motion plan that transitions from the state of the first stage to the state of the second stage.

# Example of connecting stages
# (Refer to pickplace.py for full implementation details)
initAndConfigConnect()