A Learning Rule-Based Robotics Hand Optimal Force Closure

This article presents an intelligent fuzzy rule-based approach for computing optimal set of joints torques, for manipulating a grasped object by a dexterous multi-fingered robotics hand. The intelligent approached followed here, is to let a learning fuzzy system to approximate a nonlinear force formulation for optimal contact forces. This has been achieved via following two major steps: The first was to formulate the optimal fingertips force distribution as a quadratic force optimization problem, hence to generate a large set of data. The second step was to involve a learning fuzzy system (Neuro- Fuzzy System) to learn the nonlinear relations governing fingertips forces (ℝ∈12 × 1) to hand joint torques (ℝ∈12 × 1). Simulation results show that the proposed Neuro-Fuzzy network do achieve optimal grasping force in real time.