Adaptive Manipulation using Behavior Trees
Jacques Cloete, Wolfgang Merkt, Ioannis Havoutis
TL;DR
This work addresses manipulation tasks that depend on non-visual, interaction-derived information by introducing an adaptive behavior tree (adaptive BT) that couples a data-driven strategy selector with a reactive condition monitor. The adaptive BT selects among multiple manipulation strategies to optimize task performance while preempting safety-critical failures, and it learns from past attempts to improve prediction of the optimal strategy. The authors formalize the strategy selection as a constrained optimization that minimizes estimated task duration $t_{est}$ over feasible strategies $a \in \mathcal{A}_{safe}(s)$ and demonstrate safety, robustness, and efficiency improvements across diverse industrial and domestic manipulation tasks, including needle valve tightening/loosening, globe valve handling, e-stop button manipulation, and groceries packing. The approach is modular, easily integrated into existing BT architectures, and supports future work on automatic strategy synthesis and richer conditioning to generalize to new tasks with minimal changes to the system.
Abstract
Many manipulation tasks pose a challenge since they depend on non-visual environmental information that can only be determined after sustained physical interaction has already begun. This is particularly relevant for effort-sensitive, dynamics-dependent tasks such as tightening a valve. To perform these tasks safely and reliably, robots must be able to quickly adapt in response to unexpected changes during task execution, and should also learn from past experience to better inform future decisions. Humans can intuitively respond and adapt their manipulation strategy to suit such problems, but representing and implementing such behaviors for robots remains a challenge. In this work we show how this can be achieved within the framework of behavior trees. We present the adaptive behavior tree, a scalable and generalizable behavior tree design that enables a robot to quickly adapt to and learn from both visual and non-visual observations during task execution, preempting task failure or switching to a different manipulation strategy. The adaptive behavior tree selects the manipulation strategy that is predicted to optimize task performance, and learns from past experience to improve these predictions for future attempts. We test our approach on a variety of tasks commonly found in industry; the adaptive behavior tree demonstrates safety, robustness (100% success rate) and efficiency in task completion (up to 36% task speedup from the baseline).