Digit social-navigation paper combines MPC with zonotope neural networks

Evidence notes
- The paper proposes cascaded zonotope-based neural networks for pedestrian prediction and bipedal social path planning.
- The learned social planner is integrated with MPC footstep planning for Agility Robotics Digit in simulation.
- The work gives Digit a benchmark for socially aware bipedal navigation among moving pedestrians.
Company context
US robotics company focused on humanoid systems for logistics and warehouse automation. Digit is designed for real-world material handling tasks in structured environments.