好奇号火星车13年延寿工程揭秘
值得看指数 45.0 NO. 014 · 2026.06.11
发布2026/06/10Score91Comments7
为什么值得看
NASA JPL 通过软件重构和硬件 workaround 让设计寿命仅2年的好奇号火星车持续运行13年并继续产出科学数据。这对需要长周期维护关键系统的 AI 基础设施团队有直接参考价值。
编辑判断
真正值得注意的不是硬件有多耐用,而是 JPL 的「诊断-降级-重构」运维范式:当轮子轴承磨损后,团队重写了驱动控制算法来补偿偏航;当钻机马达故障后,他们用机械臂其他关节的协同运动替代原有功能。这种「用软件弹性覆盖硬件衰减」的思路,和当前 AI 推理集群用动态调度应对 GPU 故障的做法本质同源,但 JPL 的约束更极端——单次通信延迟 20 分钟、无法现场维修、代码上传带宽以 kbps 计。做边缘 AI 或无人值守系统的团队可以研究他们的变更管理流程,其回滚机制和影子验证策略比多数互联网公司的灰度发布更保守也更可靠。
社区反馈
正面 7 条评论
核心争论:无人探测任务性价比远高于载人航天,应重新分配预算
Curiosity is a teenager now? Damn, I didn't need to feel this much older today..
The total cost of Curiosity to date is well under 5% of the cost of the recent trip humans took around the moon (something like $3B vs. $90B, or $20 vs. $600 per US taxpayer). Imagine the amount of science that could get done if we gave even half the budget of crewed spaceflight to rover / prob
Maybe we would get a microphone on mars. Just kidding i know air pressure is vastly different, but still it would be cool to listen to ambient sound from there