Soft robots have gained growing interest in medical and exploration fields due to their inherent flexibility, biocompatibility and environmental adaptability. Nevertheless, conventional dielectric elastomer actuator (DEA)-driven designs suffer from limited deformation, insufficient locomotion efficiency and poor modular reconfigurability. To address these challenges, we propose a car-shaped soft robot based on dielectric elastomer minimum energy structures (DEMES) with unequal biaxial pre-stretching, which improves the initial deflection angle (151°) and blocking force (43 mN) of the actuators. Drawing inspiration from the coordinated muscular movements of ants, the robot features a modular architecture that allows rapid integration of interchangeable ratchet wheels and grippers, endowing it with multifunctional capabilities including fast locomotion (4.6 times body lengths per second), load transportation (3.3 times its body weight), gap traversing, slope climbing (up to 13°) and object grasping (8.6 times its body weight). Additionally, an integrated multilayer perceptron (MLP) model facilitates environment-aware modulation of movement frequency, which provides the foundation for the robot’s environment-aware adaptability. By synergistically combining high-performance actuation, modular reconfigurability and intelligent control, this work provides a versatile and scalable strategy for next-generation soft robots capable of addressing complex tasks in unstructured environments.