CIRCUIT · Self-Healing Circuits for Resilient Reconfigurable Hardware

Field Programmable Gate Arrays (FPGAs) are widely deployed in space missions and critical computing applications for their flexibility and energy efficiency. Yet, they remain highly vulnerable to radiation-induced faults, where a single bit-flip can compromise functionality. Traditional countermeasures, such as hardware redundancy or manual reprogramming, are costly, energy-intensive, and too slow to guarantee mission reliability in harsh environments. This project explores a novel approach to fault tolerance in FPGAs by applying the PI’s recent breakthrough in Neural Developmental Programs (NDPs), a generative AI method inspired by biological growth and adaptation. Instead of relying on costly redundancy or slow reprogramming, our method enables on-the-fly reconfiguration through localized message-passing directly within the FPGA’s memory architecture. This self-healing mechanism promises rapid, energy-efficient recovery from radiation-induced faults, a critical challenge for space missions where reliability and efficiency are paramount. Beyond space, this approach has wide potential in domains where resilience and efficiency are crucial, such as neuromorphic computing, quantum accelerators, autonomous systems, and critical infrastructure. Ultimately, this high-risk/high-gain project has the potential to redefine hardware design by introducing a generative paradigm that mirrors nature’s ability to produce robust and efficient systems.