Object dexterous manipulation in hand based on Finite State Machine

We propose a simple but efficient control strategy to in-hand manipulate objects of unknown shape, weight, and friction properties. With this strategy, the object can be manipulated in hand in a large scale regardless there is rolling or sliding motion between the fingertips and object. We define several finger/fingers manipulation primitives and propose the hierarchical plan and control structure to facilitate the performing of the complex object manipulation task. The low level plan-local manipulation plan, is defined in continuous object configuration space, and fingers motion are planned in joints space according to the desired object motion and current perception feedback. The high level plan-global manipulation plan, is defined in finger gaits discrete space. We employ FSM (Finite State Machine) to modify fingers gaits to a new configuration in which new cycle low level plan will start again. In this way, we can solve the problem of robot hand workspace limitation. At last we design a four fingers manipulation in hand physics simulation experiment to prove the strategy feasibility. Simulation result shows the object manipulation result in ideal and simulated artificial noise cases.