Sequentially Teaching Sequential Tasks (ST)2: Teaching Robots Long-Horizon Manipulation Skills
Zlatan Ajanovic, Ravi Prakash, Leandro de Souza Rosa, Jens Kober
AI summary
Problem
Teaching robots long-horizon manipulation tasks via traditional monolithic demonstrations leads to accumulated errors, distributional shifts, and significant human teacher fatigue.
Approach
The authors introduce (ST)2, a framework that allows teachers to segment tasks by inserting key-points, enabling incremental, step-by-step policy learning and real-time corrections.
Key results
- (ST)2 framework for user-controlled sequential segmentation and incremental learning
- FMEA-compatible benchmark task for direct teaching framework comparison
- User study showing sequential teaching outperformed monolithic in 10 of 16 cases
- Insights into human teaching preferences balancing iterative control and simplicity
Why it matters
Enables developers to design less fatiguing, more effective teaching interfaces for complex robotic manipulation in real-world logistics and retail environments.
Abstract
Learning from demonstration has proved itself use- ful for teaching robots complex skills with high sample efficiency. However, teaching long-horizon tasks with multiple skills is challenging as deviations tend to accumulate, the distributional shift becomes more evident, and human teachers become fatigued over time, thereby increasing the likelihood of failure. To address these challenges, we introduce (ST)2, a sequential method for learning long-horizon manipulation tasks that allows users to control the teaching flow by specifying key points, enabling struc- tured and incremental demonstrations. Using this framework, we study how users respond to two teaching paradigms: (i) a traditional monolithic approach, in which users demonstrate the entire task trajectory at once, and (ii) a sequential approach, in which the task is segmented and demonstrated step by step. We conducted an extensive user study on the restocking task with 16 participants in a realistic retail store environment, evaluating the user preferences and effectiveness of the methods. User- level analysis showed superior performance for the sequential approach in most cases (10 users), compared with the monolithic approach (5 users), with one tie. Our subjective results indicate that some teachers prefer sequential teaching—as it allows them to teach complicated tasks iteratively—or others prefer teaching in one go due to its simplicity.