Planning with Learning¶
Integrated Approaches¶
Combine symbolic planning with statistical learning.
Methods¶
Learning Plan Heuristics¶
Learn search heuristics from experience.
Learning Action Models¶
Improve planning accuracy by learning better action models.
Plan Repair with Learning¶
Use learned models to repair failed plans.
Neuro-Symbolic Planning¶
# Hybrid architecture
symbolic_planner = Planner()
neural_perception = CNN()
policy_network = RLAgent()
# Integration
state = neural_perception(observation)
high_level_plan = symbolic_planner.plan(state)
low_level_action = policy_network.select_action(state)
Applications¶
- Robot manipulation
- Autonomous driving
- Game playing
