Quantifying Demonstration Quality in Learning from Demonstration

Learning from Demonstration empowers novice users to teach robots daily life tasks without writing sophisticated code, thereby promoting the democratization of robotics. However, novice users often provide sub-optimal demonstrations, which can potentially impact the robot’s ability to efficiently learn and execute the tasks. Prior research has assessed the quality of demonstrations by evaluating the robot’s task performance; however, the approach remains insufficient to qualify individual demonstrations, leaving the reason for classifying demonstrations as high- or low-quality unknown. Therefore, this simulation-based study aims to quantify the quality of individual demonstration at each step by incorporating motion-related quality features such as manipulability and joint-space jerk.

The given datasets have been evaluated based on the quality features, including manipulability and joint-space jerk, and ranked them from high to low quality. At next step, we trained an LfD model for each group, where N denotes the total number of groups, followed by rollouts the learned LfD model in order to compute the task performance and assess the quality of task execution based on the quality features.

Our results illustrate a strong correlation between ranked demonstrations and the quality of task execution. Interestingly, we observed that the quality features have a significant impact on task performance, particularly when the provided demonstrations exhibit diversity in terms of quality. Overall, this analysis enables quantifying the quality of individual demonstrations based on motion-related quality features, thus improving learning from demonstration.

Key Findings

This study makes three primary contributions:

  • Utilisation of motion-based features to quantify the quality of individual demonstrations before learning process.
  • Transferability of quality features from provided demonstrations to the corresponding LfD model generated motions.
  • Quality features can serve as an indicator of task performance when used in a diverse set of demonstrations.

Resources

PDF

Citation

Muhammad Bilal, Nir Lipovetzky, Denny Oetomo, and Wafa Johal. Beyond Success: Quantifying Demonstration Quality in Learning from Demonstration. In Proceedings of the IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), pp. 5120–5127, 2024. DOI:10.1109/IROS58592.2024.10802187

 

October 22, 2024 By Learning from Demonstration