satoshi88818/david-

David is a software architecture for an autonomous humanoid robot, a robot shaped like a person that can move and interact in human environments without direct human control. The codebase provides all the reasoning and control layers that sit between raw sensor readings and physical movement. The system is organised into cooperating modules. A Vision-Language-Action model translates what the robot sees and hears into sequences of physical movements using diffusion sampling, generating many possible motion paths and selecting the safest, most useful ones. A World Model stores a constantly updated picture of the robot's surroundings, using a lock-free memory technique so reading the latest state never delays the motor control thread. A Social Grid predicts how nearby people will move using Level-K reasoning: rather than assuming a person walks in a straight line, it models how they would react to the robot changing course, preventing the "hallway dance" where both keep stepping into each other's way. A Dreamer loop filters every planned trajectory using mathematical tools called a Control Barrier Function and a Control Lyapunov Function to guarantee the robot stays balanced and makes progress toward its goal. A Diagnostic Reasoner estimates payload weight and floor friction from joint sensor readings via an Unscented Kalman Filter, a statistical technique for tracking hidden quantities from indirect measurements. All modules are coordinated by a Planner Node running at 10 Hz, ten decision cycles per second. The full README is longer than what was provided.