Robust Ladder Climbing with a Quadrupedal Robot

2024-09-26

arXiv.org

AI Summary

Quadrupedal robots struggle with robust vertical ladder climbing in industrial environments, limiting autonomous inspection in hazardous areas. This paper proposes a reinforcement learning–driven end-to-end locomotion control framework integrated with a novel hook-and-hang end-effector, enabling zero-shot cross-ladder-type transfer and stable climbing under strong external disturbances. The method overcomes longstanding limitations in adapting to extreme ladder inclinations (70°–90°) and geometric variations, leveraging co-optimization of simulation-to-reality training and a comprehensive robustness validation pipeline. Hardware experiments achieve a 90% climbing success rate and a climbing speed 232× faster than the state of the art, while maintaining stability under unmodeled dynamic disturbances. To our knowledge, this is the first systematic demonstration of reliable, real-world ladder climbing by quadrupeds on actual industrial infrastructure—significantly reducing human intervention risks and enhancing operational efficiency.

Technology Category

• Intelligent Robots: Manipulation
• Planning, Routing, and Scheduling: Learning for Planning and Scheduling
• Search and Optimization: Learning to Search

Application Category

• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Economics, Online Markets and Human Computation: LLM based quality controls for crowd work

Abstract

Quadruped robots are proliferating in industrial environments where they carry sensor suites and serve as autonomous inspection platforms. Despite the advantages of legged robots over their wheeled counterparts on rough and uneven terrain, they are still yet to be able to reliably negotiate ubiquitous features of industrial infrastructure: ladders. Inability to traverse ladders prevents quadrupeds from inspecting dangerous locations, puts humans in harm's way, and reduces industrial site productivity. In this paper, we learn quadrupedal ladder climbing via a reinforcement learning-based control policy and a complementary hooked end-effector. We evaluate the robustness in simulation across different ladder inclinations, rung geometries, and inter-rung spacings. On hardware, we demonstrate zero-shot transfer with an overall 90% success rate at ladder angles ranging from 70{deg} to 90{deg}, consistent climbing performance during unmodeled perturbations, and climbing speeds 232x faster than the state of the art. This work expands the scope of industrial quadruped robot applications beyond inspection on nominal terrains to challenging infrastructural features in the environment, highlighting synergies between robot morphology and control policy when performing complex skills. More information can be found at the project website: https://sites.google.com/leggedrobotics.com/climbingladders.

Problem

• Quadruped robots cannot reliably climb industrial ladders.
• Inability to climb ladders limits inspection and productivity.
• Learning ladder climbing via reinforcement learning and hooked end effectors.

Innovation

• Reinforcement learning-based control policy for climbing
• Complementary hooked end effector design
• Robust simulation across varied ladder conditions