Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Human-Centric Goal Reasoning with Ripple-Down Rules

Kenji Brameld, Germán Castro, Claude Sammut, Mark Roberts, David W. Aha

TL;DR

Human-Centric Goal Reasoning with Ripple-Down Rules presents learning-by-demonstration for goal selection in ActorSim by integrating Ripple-Down Rules ($RDR$) within $FrameScript$ and linking to a $PDDL$ planner via $POPF2$. The method enables scalable, explainable rule construction through teacher-guided exception and generalisation, extending ActorSim to handle an order-of-magnitude more goals in RoboCup Rescue. Key contributions include the first application of $RDR$ to goal reasoning, integration with a $PDDL$ planner, and a demonstrable ability to learn from demonstrations while maintaining explainability. This approach supports human-in-the-loop autonomy and points toward future hierarchical command structures and improved explainable-rule generation.

Abstract

ActorSim is a goal reasoning framework developed at the Naval Research Laboratory. Originally, all goal reasoning rules were hand-crafted. This work extends ActorSim with the capability of learning by demonstration, that is, when a human trainer disagrees with a decision made by the system, the trainer can take over and show the system the correct decision. The learning component uses Ripple-Down Rules (RDR) to build new decision rules to correctly handle similar cases in the future. The system is demonstrated using the RoboCup Rescue Agent Simulation, which simulates a city-wide disaster, requiring emergency services, including fire, ambulance and police, to be dispatched to different sites to evacuate civilians from dangerous situations. The RDRs are implemented in a scripting language, FrameScript, which is used to mediate between ActorSim and the agent simulator. Using Ripple-Down Rules, ActorSim can scale to an order of magnitude more goals than the previous version.

Human-Centric Goal Reasoning with Ripple-Down Rules

TL;DR

Human-Centric Goal Reasoning with Ripple-Down Rules presents learning-by-demonstration for goal selection in ActorSim by integrating Ripple-Down Rules () within and linking to a planner via . The method enables scalable, explainable rule construction through teacher-guided exception and generalisation, extending ActorSim to handle an order-of-magnitude more goals in RoboCup Rescue. Key contributions include the first application of to goal reasoning, integration with a planner, and a demonstrable ability to learn from demonstrations while maintaining explainability. This approach supports human-in-the-loop autonomy and points toward future hierarchical command structures and improved explainable-rule generation.

Abstract

ActorSim is a goal reasoning framework developed at the Naval Research Laboratory. Originally, all goal reasoning rules were hand-crafted. This work extends ActorSim with the capability of learning by demonstration, that is, when a human trainer disagrees with a decision made by the system, the trainer can take over and show the system the correct decision. The learning component uses Ripple-Down Rules (RDR) to build new decision rules to correctly handle similar cases in the future. The system is demonstrated using the RoboCup Rescue Agent Simulation, which simulates a city-wide disaster, requiring emergency services, including fire, ambulance and police, to be dispatched to different sites to evacuate civilians from dangerous situations. The RDRs are implemented in a scripting language, FrameScript, which is used to mediate between ActorSim and the agent simulator. Using Ripple-Down Rules, ActorSim can scale to an order of magnitude more goals than the previous version.
Paper Structure (7 sections, 7 figures, 2 tables)

This paper contains 7 sections, 7 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Goal Lifecycle
  3. Software Architecture
  4. Updating Ripple-Down Rules
  5. Operation of the System
  6. Evaluation
  7. Conclusion

Figures (7)

  • Figure 1: ActorSim goal life cycle from roberts2014iterative
  • Figure 2: A typical RoboCup Rescue Map. Civilians are shown as green dots, with darker hues indicating poorer health. A black dot means the victim has died. Fire trucks are red, ambulances, white and police yellow. Coloured buildings are on fire or in danger of collapse. Black 'X' denotes blockages in a road.
  • Figure 3: Interaction between centralised goal planning and agent actions planning.
  • Figure 4: ExpertGUI displaying the centre's knowledge of the world
  • Figure 5: Controls to start, pause the simulation and replay world states
  • ...and 2 more figures