I am building a semi-autonomus outdoor robot that needs to drive to user defined point. It doesn't need a state machine. It's for competition.

The problem: I will be provided a point cloud scan of the area few days before the competition and a few hours to drive around the area. I used rtab-map so far but there isn't a way to import a point cloud to use it for localization.

From my limited testing it works good enough (I don't have the full prototype yet), but I would really like the redudancy of importing a point cloud.

(Or should I think about multi-session mapping as my redudancy?)

Is there better library for my use case or should I just suck it up and hope for the best?

(Bonus question: Any navigation libraries that I should look into other nav2?)

Chinese startup UBTECH Robotics just snagged a $37M contract to deploy its Walker S2 humanoid robots at the China-Vietnam border in Fangchenggang, where they'll guide passenger queues, direct vehicles, patrol corridors, and inspect cargo containers. The deployment follows China's 2023 directive to build a national humanoid robotics innovation system by 2025.

The company just rolled out Qingtian Rent, a rental platform that lets you hire robots for weddings, concerts, trade shows, and business meetings. AgiBot’s advanced Yuanzheng A2 model rents for $1,380 per day, while budget options like the Unitree Go2 robot dog start at $138 daily. The move could potentially reshape the global robotics market: China's rental sector hit $140M in 2025 and is projected to explode to $1.4B next year.

Google DeepMind is rolling out an on-device version of its Gemini Robotics AI model that allows it to operate without an internet connection.