Soft robotics with high flexibility & adaptability gain attention for precision tasks in unstructured environments. However, non-contact robotics’ reliance on thermal/optical fields limits stability & stimulus uniformity, resulting in low accuracy, poor scalability, & limited adaptability. This paper introduces a new soft robotic system (PM10B2C robotic) incorporated Internet of Things (IoT) technology and leveraged the Joule thermal effect. The robotic platform is constructed by self-prepared boron nitride nanosheets (BNNS) that exhibit a higher planar area-to-thickness ratio. When the actuation voltage is 8 V, the soft wireless robotic can complete κ = 1.19 cm−1 deformation with deformation efficiency and recovery efficiency of 0.11 cm−1·s−1 and 0.17 cm−1·s−1, and exhibits excellent actuation stability under 500 cycles of testing. Based on the Timoshenko bimetallic constant temperature device model and heat conduction theory, a κmax-U analytical model is constructed to achieve effective prediction of the optimal curvature-low voltage D.C. relationship of PM10B2C robotics. The smart-controlled soft robotic operates continuously and stably within the harsh internal environments and complex spatial configurations of thermal power plant in cooling towers. This design offers novel prospects for developing cleaning robotics tailored for extreme environments and complex spaces in such facilities.