A Review of Paradigm Shifts and Collaborative Evolution in Robot Chassis Design Driven by VLA

Abstract

Vision-Language-Action (VLA) foundation model-driven embodied intelligence is a cutting-edge field at the intersection of artificial intelligence and robotics, endowing robots with powerful semantic understanding and generalization capabilities. However, traditional mechanical body design paradigms are based on assumptions of structured environments and struggle to adapt to the open-task requirements driven by VLA across the dimensions of perception, execution, and interaction. This results in insufficient coordination between intelligent modules and mechanical bodies in open tasks, thereby limiting the overall system performance. This paper systematically reviews the paradigm-shifting challenges posed by the rise of VLA to robotic mechanical body design and explores pathways for their collaborative evolution. First, we dissect the limitations of traditional design in open scenarios; second, we analyze the morphological evolution of the robot body from rigid to compliant and from dedicated to programmable; and third, we elucidate the coordination mechanisms between the intelligent core and the physical shell. This paper aims to provide a theoretical framework for the development of next-generation robot platforms and to advance the transition of general embodied intelligence from theoretical research to practical application.