Deliberative Acting

Planning-Based Control

Deliberative systems use planning to select actions.

Model

1. World model: Current and predicted states
2. Plan library: Available actions
3. Goal specification: Desired outcomes

Acting Loop

while not goal_achieved:
    state = perceive()
    if plan_invalid(plan, state):
        plan = replan(state, goal)
    action = next_action(plan)
    execute(action)
    update_state()

Anytime Algorithms

Algorithms that improve solution quality over time:

• Hierarchical Task Networks (HTN)
• A* with pathmax
• RA* (Randomized A*)

Integration with Execution

• Plan monitoring: Detect failures
• Plan repair: Modify plan on failure
• Plan execution: Carefully sequence actions

---

← Back to Overview