A Brief Survey on the Role of Dimensionality Reduction in Manipulation Learning and Control Strategies

Bio-inspired designs are motivated by efficiency, adaptability and robustness of biological systems dynamic behaviors in complex environment. Despite progress in design, the lack of sensorimotor and learning capabilities is the main drawback of human-like manipulation systems. Dimensionality reduction has demonstrated in recent robotics research to solve problems that affect high degrees of freedom (DoFs) devices. In this paper, a survey on the role of dimensionality reduction in learning and control strategies is provided by discussing different techniques adopted for dimensionality reduction, as well as learning and control strategies built on subspaces of reduced dimension across different fully-actuated and underactuated anthropomorphic designs.