The development of soft sensors for integration into untethered miniature robots is significant for improving their environmental perception in physically challenging scenarios, such as collapsed buildings after an earthquake. The primary objective is to design and manufacture reliable soft sensors that serve as structural and sensing elements within the robots, eliminating the need for post-processing methods like data-driven learning and optimization. The soft sensors employ resistive sensing, similar to strain gauges, and are implemented on a Wheatstone bridge to convert resistive changes into voltage changes under me-chanical actuation or deformation. The study explores two categories of soft sensor designs: sheet-type and 3D shaped sensors. Sheet-type sensors are embedded in the C-legs of a soft quadruped robot (SQuad), enabling gait control, while 3D shaped sensors are structurally integrated into the robots to enhance environmental perception. Manufacturing of the soft sensors is made accessible and efficient through 3D printing technology, using conductive Thermoplastic Polyurethane (cTPU) as the printing material. Challenges arise in integrating the soft sensors into the robots while preserving their soft nature, locomotion, and agility. The thesis addresses these challenges by implementing the soft sensor concept in various robots and their parts, including the C-leg of SQuad, Modular Soft Quadruped (M-SQuad), Suspensionized Soft Quadruped (S-SQuad), Sensorized Collision Resilient Robot (SCoReR), and a tail for Reconfigurable Miniature Modular Robot (ReMBot). The soft sensors enable different functionalities to these robots, such as gait control feedback, obstacle detection, inclination detection, and collision detection, enhancing the adaptability of the robots in physically challenging environments. The thesis highlights the potential of soft 3D printed strain gauges. The ease of manufacturing and cost-efficiency of these sensors make them promising for applications in wearable robots and human-computer interfaces. Future directions are highlighted, emphasizing the need for detailed sensor characterization experiments and the development of detection algorithms to improve reliability. Additionally, a dynamic model of the coil-shaped sensors is proposed to simulate resistance changes, streamlining the design process without repetitive manufacturing iterations. As a result, this thesis presents a reliable soft sensor design, manufacturing, and integration into untethered miniature robots. The outcome of this work demonstrates the effectiveness of soft sensors in enhancing environmental perception, paving the way for innovative solutions in force measurement applications and human-computer interactions.