Flexible Thermal Management Method for High-Power Chips in the Humanoid Robot

Xiangbin Du,Yuxin Ye,Binbin Jiao,Yanmei Kong,Lihang Yu,Ruiwen Liu,Shichang Yun,Dichen Lu,Jingping Qiao,Ziyu Liu,Ronggui Yang
DOI: https://doi.org/10.2139/ssrn.4606561
2024-01-01
Device
Abstract:High-performance artificial intelligence facilitates cognitive capabilities in humanoid robots, while the implementation of bionic skin restores human physiology. The convergence of these two cutting-edge technologies enables the development of fully autonomous artificial life forms. However, the flexible skin of humanoid robots hinders heat dissipation; this leads to a thermal build-up when computational demands increase, necessitating the development of thermal regulation systems compatible with flexible environments. Drawing inspiration from the human blood circulation system, this study proposes a flexible recirculating cooling microflow system for multiple chips in the humanoid robot covered with flexible skin. The system employs a distributed heat transfer and centralized heat dissipation model, comprising three components: flexible microfluidic heat exchange modules (FMHEMs) that are dispersed throughout the robot body, a fluid circulation module for coolant flow, and an external centralized cooling module that concentrates and dissipates heat into the external environment. In the module-level study, the FMHEM exhibited a low junction-to-ambient thermal resistance (0.089 cm2K/W), and the stress induced on the chip owing to deformation was only 5% of that in conventional packaging methods. Moreover, application verification on the robotic arm and the robot demonstrated that the system maintained stable heat dissipation during body movement (with a maximum temperature difference of 3K), while reducing the power usage effectiveness (PUE) by 8% compared to the traditional water-cooling methods. This system optimizes energy management in humanoid robots, and provides a viable solution to the problem of blocked chip heat dissipation paths caused by high skin coverage during the development of extremely intelligent humanoid robots.
What problem does this paper attempt to address?