HumanMPC - Safe and Efficient MAV Navigation among Humans

TL;DR

This work tackles safe and efficient 3D MAV navigation in human environments by marrying data-driven 3D human motion forecasting with a novel MPC safety constraint that enforces safety for the initial input $$ while propagating effects over a horizon $T$. The core idea is to use a linear, reachability-based safety condition built on zonotopes for the MAV's reachable set and combined complex/simplified models for human reachability, integrated with MotionMixer forecasts to guide visual servoing and goal-directed tasks. Key contributions include a scalable MPC formulation with a linear safety constraint that avoids heavy precomputation, a hybrid human-reachable-set representation with switching at $t_s$, and extensive simulation and real-world validation showing safety without excessive conservatism and improved efficiency over baselines. The framework enables real-time, safe navigation for MAVs among moving humans and can be adapted to other platforms, marking a practical step toward robust autonomous operation in dynamic human-centric spaces.

Abstract

Safe and efficient robotic navigation among humans is essential for integrating robots into everyday environments. Most existing approaches focus on simplified 2D crowd navigation and fail to account for the full complexity of human body dynamics beyond root motion. We present HumanMPC, a Model Predictive Control (MPC) framework for 3D Micro Air Vehicle (MAV) navigation among humans that combines theoretical safety guarantees with data-driven models for realistic human motion forecasting. Our approach introduces a novel twist to reachability-based safety formulation that constrains only the initial control input for safety while modeling its effects over the entire planning horizon, enabling safe yet efficient navigation. We validate HumanMPC in both simulated experiments using real human trajectories and in the real-world, demonstrating its effectiveness across tasks ranging from goal-directed navigation to visual servoing for human tracking. While we apply our method to MAVs in this work, it is generic and can be adapted by other platforms. Our results show that the method ensures safety without excessive conservatism and outperforms baseline approaches in both efficiency and reliability.

Figures (5)

• Figure 1: HumanMPC: A control framework for safe and efficient MAV navigation among humans. Our method combines reachability-based safety constraints with state-of-the-art data-driven human motion forecasting, ensuring recursive safety guarantees without overly conservative behavior.
• Figure 2: Based on the current MAV state $\mathbf{x}_0$ and the tracked 3D positions of human body joints, we first compute both the MAV's and the human's reachable sets over the entire horizon. We then optimize the control inputs $\mathbf{u}_k$ with the requirement that the initial control input $\mathbf{u}_0$ must not lead to a situation where the MAV’s reachable set becomes a subset of the human’s reachable set at any time. While overlaps between the two sets are safe, as they can be resolved by future control actions, a complete overlap would imply inevitable collision regardless of future control inputs. To run on an MAV, our method is augmented with a real-time perception and forecasting stack (dashed boxes) that provides the current human and MAV state and informs the objective about future human motion.
• Figure 3: Example evaluation of our safety constraint for three different choices for the initial control input. While the left choice leads to unsafe behavior, both the middle and the right choices for the initial control input are regarded safe as they leave options to avoid collision later on.
• Figure 4: Example construction of the MAV reachable set with non-zero initial velocity in forward direction, expanding along the planning horizon from $\Delta t = 0.025s$ to $\Delta t = 0.500s$.
• Figure 5: Example constructions of the human reachable set for discretization steps $\Delta t = 0.025s$ and $\Delta t = 0.500s$. The black dots represent the 3D body joints at $t = 0s$.