Path and Motion Optimization for Efficient Multi-Location Inspection with Humanoid Robots

TL;DR

The paper tackles the challenge of efficient, precise multi-location inspection with humanoid robots in industrial settings. It introduces a hierarchical pipeline that combines IK-based feasible-region sampling, tolerance-circle abstraction, and mixed-integer programming to optimally select standing positions and sequencing, followed by a model predictive controller with a hybrid velocity/step strategy for millimeter-level end-effector tracking. Key contributions include (i) a co-designed standing-position generation via MIP, (ii) a time-optimal trajectory framework for whole-body motion, and (iii) a hybrid MPC that maintains high tracking accuracy while enabling fast locomotion, validated on the Kuavo 4Pro in both simulations and real automotive inspections. The approach significantly improves task completion time and robustness for dense, multi-location inspection tasks in industry, demonstrating practical potential for scalable automation.

Abstract

This paper proposes a novel framework for humanoid robots to execute inspection tasks with high efficiency and millimeter-level precision. The approach combines hierarchical planning, time-optimal standing position generation, and integrated \ac{mpc} to achieve high speed and precision. A hierarchical planning strategy, leveraging \ac{ik} and \ac{mip}, reduces computational complexity by decoupling the high-dimensional planning problem. A novel MIP formulation optimizes standing position selection and trajectory length, minimizing task completion time. Furthermore, an MPC system with simplified kinematics and single-step position correction ensures millimeter-level end-effector tracking accuracy. Validated through simulations and experiments on the Kuavo 4Pro humanoid platform, the framework demonstrates low time cost and a high success rate in multi-location tasks, enabling efficient and precise execution of complex industrial operations.

Figures (7)

• Figure 1: Performing an inspection task with a humanoid robot. The Dimension Technical Specifications (DTS) task in automotive quality check is utilized as an example, where the humanoid robot needs to hold a high-accuracy 3D scanner to reach multiple targets efficiently to measure the gap between different installation parts.
• Figure 2: The proposed MIP-based standing position planning pipeline. Given a sequence of target end-effector poses, concave feasible base regions are computed via IK sampling. Within each region and their overlaps, the largest valid tolerance circles are generated to capture robust standing areas. A mixed-integer program then selects an optimal subset of circles and computes a walking path that minimizes both the number of standing positions and total travel distance, enabling efficient inspection.
• Figure 3: The experimental setup and system diagram. (a) The inspection task scenario in simulation and experiment, together with the hardware configuration of the bipedal humanoid robot KUAVO 4Pro. (b) The control block diagram of the inspection task with both velocity control and single-step control modes.
• Figure 4: Single-step control illustration. If the robot's position is near the distance circle, single-step control will be used to generate new footsteps for more accurate tracking of target location $\pmb{p}_{b_{x,y}}^{des}$.
• Figure 5: The comparison of planning and execution time with the proposed MIP method and naive point‐to‐point planning method.