Temporal Multiagent Plan Execution: Explaining What Happened

Abstract

The paper addresses the problem of explaining failures that happened during the execution of Temporal Multiagent Plans (TMAPs), namely MAPs that contain both logic and temporal constraints about the actions conditions and effects. We focus particularly on computing explanations that help the user figure out how failures in the execution of one or more actions propagated to later actions. To this end, we define a model that enriches knowledge about the nominal execution of the actions with knowledge about (faulty) execution modes. We present an algorithm for computing diagnoses of TMAPs execution failures, where each diagnosis identifies the actions that executed in a faulty mode, and those that failed instead because of the propagation of other failures. Diagnoses are then integrated with temporal explanations, that detail what happened during the plan execution by specifying temporal relations between the relevant events.

Blame Attribution for Multi-Agent Path Finding Execution Failures

“…Fault type is a notable point of the difference between previous work. Some studies aim to diagnose faulty agents when a fault in the execution of the plan occurs [8,41,7,34,35,50,37]. Others aim to diagnose failures in the cooperation between agents [6,24,27,25,26,23].…”

The Emerging Landscape of Explainable Automated Planning & Decision Making

“…Finally, we look at the role of plans in explanatory dialogue. Works like [52; 67] have explored explanations in the form of a plan that explains a set of observations, while methods like [78] have looked at ways to generate the most likely explanation for why a plan failed. Beyond inferential support in human-AI interaction, the qualitative structure of plans has also been used for plan-reuse and validation [38].…”

Machine Reasoning Explainability

“…[74] • Why fail and what (temporal) repercussions does the first failure have? [201] • Why did the agent take action A (at that time) rather than action B (resp. earlier or later)?…”

“…[193]. Similarly, pointing out actions executed in a faulty node and their propagations (subsequent failed actions and their relationships) can be used to define explanations in temporal multi-agent planning [201]. Such explanations mostly work by attributing counterexamples to solvability in the current system S. However, counterfactual contrastive explanations in terms of excuses -minimal, restrictive changes to the initial state i that would allow to reach the desired goal o -are also possible [91].…”