Strategies, controllers, and coordination: Bi-manual snap assembly automation

In robotic manipulation, higher levels of dexterity are sought in multiple domains. More unstructured environments demand that robots are able to dexterously and flexibly resolve their tasks. Dual-arm humanoid robots are seen as advantageous for such tasks. Much work has been performed in motion planning, control, and cooperation across multi-independent manipulators, but recent emphasis has been given to dual-armed torsos and bi-manual cooperation and coordination. This work contributed a snap cantilever automation approach by considering and unifying three domains: intuitive assembly strategies, modular control approaches, and bi-manual control policies. In particular, the approach sought to create an approach that can generalize to assemblies of increasing geometric complexity as well as establishing the ground work for more dexterous and flexible bi-manual cantilever snap assemblies. The used strategy systematically discretizes the assembly into intuitive automata states. The latter is run in concert with scalar controllers that flexibly optimize different control objectives. These two are subject to a bimanual coordination policy. The effectiveness of our approach was demonstrated by a bi-manual humanoid robot assembling two 4-snap cantilever parts smoothly and efficiently.