Optimal Interactive Learning on the Job via Facility Location Planning

2025-05-01

AI Summary

Collaborative robots must adapt to new tasks and user preferences in sustained multi-task settings while minimizing human cognitive and physical burden. Method: This paper proposes the Cost-Optimal Interactive Learning (COIL) framework, which formulates multi-task human–robot interaction planning as an uncapacitated facility location (UFL) problem—integrated with belief-space planning to robustly handle preference uncertainty. COIL jointly schedules three interaction types: skill queries, preference queries, and human assistance requests. Contribution/Results: COIL is the first to cast interactive multi-task adaptation as UFL, enabling polynomial-time computation of bounded-suboptimal solutions. It guarantees high task success rates while drastically reducing human involvement. In simulation and real-world robotic arm experiments, COIL reduces total interaction count by 42% compared to baseline methods, while maintaining a task completion rate of ≥98.7%.

Technology Category

• Intelligent Robots: Learning & Optimization for ROB
• Humans and AI: Human-Aware Planning and Behavior Prediction
• Planning, Routing, and Scheduling: Learning for Planning and Scheduling

Application Category

• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems
• Responsible Web: Machine-in-the-loop, human agency and autonomy
• User Modeling, Personalization and Recommendation: User modeling and simulation for interactive and conversational systems

Abstract

Collaborative robots must continually adapt to novel tasks and user preferences without overburdening the user. While prior interactive robot learning methods aim to reduce human effort, they are typically limited to single-task scenarios and are not well-suited for sustained, multi-task collaboration. We propose COIL (Cost-Optimal Interactive Learning) -- a multi-task interaction planner that minimizes human effort across a sequence of tasks by strategically selecting among three query types (skill, preference, and help). When user preferences are known, we formulate COIL as an uncapacitated facility location (UFL) problem, which enables bounded-suboptimal planning in polynomial time using off-the-shelf approximation algorithms. We extend our formulation to handle uncertainty in user preferences by incorporating one-step belief space planning, which uses these approximation algorithms as subroutines to maintain polynomial-time performance. Simulated and physical experiments on manipulation tasks show that our framework significantly reduces the amount of work allocated to the human while maintaining successful task completion.

Problem

• Adapting collaborative robots to novel tasks and user preferences efficiently
• Minimizing human effort in multi-task interactive learning scenarios
• Handling uncertainty in user preferences during robot task planning

Innovation

• Multi-task interaction planner minimizing human effort
• Formulates learning as uncapacitated facility location problem
• Handles uncertainty with one-step belief space planning