A Computationally Efficient Nonparametric Approach for Robot Imitation Learning
Yijin Wang, Shaokang Wu, Chen Liu, Chuankai Zhang, João Silvério, Yanlong Huang
AI summary
Problem
Standard nonparametric imitation learning models like KMP scale cubically with demonstration length, making real-time adaptation and high-dimensional skill generalization computationally prohibitive.
Approach
The authors introduce instant KMP (iKMP), which uses block matrix inversion to avoid full kernel matrix recomputation during updates, and MiKMP, which partitions the reference trajectory into local segments and fuses their Gaussian predictions to accelerate adaptation.
Key results
- Reduces update time complexity from O(D³N³) to O(D³N²)
- MiKMP further accelerates computation by fusing local model predictions via Gaussian product
- Matches standard KMP and parametric baselines in trajectory accuracy and adaptation performance
- Enables instant real-time adaptations across 2-D, 3-D, and 7-D robotic tasks
Why it matters
Enables scalable, real-time nonparametric skill learning for complex robotic applications without sacrificing adaptation fidelity.
Abstract
Transferring human skills to robots through learn- ing from demonstrations has been an important topic in the robotics community, and many models have been developed for learning and adapting such skills. Among them, nonparametric representations are an appealing choice, since nonparametric solutions alleviate the explicit definition of basis functions, require fewer hyperparameters, and facilitate straightforward generalization for tasks involving high-dimensional inputs (e.g., human–robot collaboration and dual-arm manipulation). How- ever, a commonly raised concern for nonparametric models is their computational complexity. In this paper, we propose a computationally efficient solution for nonparametric skill learning, whose computation time grows quadratically with the length of demonstrations, as opposed to the cubic growth in a standard nonparametric model. The solution is further im- proved by exploiting local models and fusing their predictions. We evaluate our approach in a 2-D writing task with time input, a 3-D human–guided obstacle avoidance task, and a dual- arm transportation task associated with 7-D input. The results show that our solution achieves comparable performance to the parametric method and enables instant adaptations in tasks associated with time or multi-dimensional inputs.