The STRANDS Project: Long-Term Autonomy in Everyday Environments (1604.04384v2)

Abstract: Thanks to the efforts of the robotics and autonomous systems community, robots are becoming ever more capable. There is also an increasing demand from end-users for autonomous service robots that can operate in real environments for extended periods. In the STRANDS project we are tackling this demand head-on by integrating state-of-the-art artificial intelligence and robotics research into mobile service robots, and deploying these systems for long-term installations in security and care environments. Over four deployments, our robots have been operational for a combined duration of 104 days autonomously performing end-user defined tasks, covering 116km in the process. In this article we describe the approach we have used to enable long-term autonomous operation in everyday environments, and how our robots are able to use their long run times to improve their own performance.

• The paper introduces the STRANDS Core System, a novel architecture for robust, long-term autonomous operation in everyday indoor settings.
• It employs adaptive navigation, efficient task scheduling, and Frequency Map Enhancement to predict and respond to dynamic environmental changes.
• Extensive real-world deployments achieved 104 cumulative days of operation with 8,444 tasks completed, validating the system's endurance and effectiveness.

The STRANDS Project: Long-Term Autonomy in Everyday Environments

The paper "The STRANDS Project: Long-Term Autonomy in Everyday Environments," authored by Hawes et al., presents significant advancements in robotic long-term autonomy (LTA) for indoor service robots. The STRANDS project integrates cutting-edge artificial intelligence and robotic technologies into mobile service robots, specifically designed for use in real-world environments. The project's novelty lies in its focus on enabling robots to perform multiple tasks autonomously over extended periods, improving their performance through long-term observation and experience in dynamic indoor settings.

Core Contributions and System Architecture

The primary contribution of the paper is the introduction of the STRANDS Core System, an architecture designed for long-term deployment in everyday environments. This system is robust, featuring both design-time and run-time strategies to manage failures and optimize performance over prolonged autonomous operations. The architecture includes components for navigation, task scheduling, data storage, and learning, using ROS and open-source software to ensure flexibility and replicability.

Key elements of the STRANDS Core System include:

1. Robust Navigation: Utilizes a topological map and adaptive Monte Carlo localization to navigate dynamic environments predictably, enhancing reliability through monitored navigation and recovery behaviors.
2. Task Management: Employs an executive framework to schedule and manage tasks, supporting a predictable behavior while maximizing autonomy through efficient resource management.
3. Data-Driven Learning: Collects environmental data using MongoDB and applies Frequency Map Enhancement (FreMEn) to model and predict environment dynamics, facilitating adaptive behavior and improving task performance.
4. Predictive Models: Utilizes Markov Decision Processes (MDPs) and FreMEn models to adapt navigation strategies based on long-term experience, reducing failure rates, and optimizing performance.
5. User Interaction: In care environments, the system adapts its behavioral schedule based on human-robot interaction predictions, optimizing user engagement through continuous learning and adaptation.

Deployment and Evaluation

STRANDS systems have been evaluated through real-world deployments in security and care scenarios, achieving significant operational milestones. The robots operated autonomously for a cumulative duration of 104 days, completing 8,444 tasks over 116 kilometers. Two notable accomplishments include a 28-day continuous operation without human intervention and substantial improvements in navigation and task execution robustness.

Two critical metrics were used for evaluation: Total System Lifetime (TSL) and Autonomy Percentage (A\%). TSL measures the robot's operational continuity, while A\% reflects the duration the system actively performs tasks relative to available operation time. The project successfully demonstrated improved LTA capabilities, with enhancements derived from learning and adapting to environmental patterns and user interactions.

Implications and Future Directions

The STRANDS project has meaningful implications for the development of LTA systems, demonstrating that robots can extend their autonomous operational periods by continuously learning from their environment. This progress contributes towards solving unique challenges in deploying robots in highly dynamic and interactive settings like hospitals and offices.

Future work could extend the capabilities of the STRANDS system to cope with catastrophic failures and unforeseen changes by employing redundancy and more comprehensive failure detection strategies. Enhanced understanding of human activities and autonomously closing knowledge gaps remain potential areas for further investigation to optimize LTA systems for complex, unstructured environments.

In conclusion, the STRANDS project represents substantial progress in deploying long-term autonomous robots in everyday settings. The robust, adaptable architecture with learning capabilities creates a solid foundation for future advancements in service robotics.