A Method of Joint Angle Estimation Using Only Relative Changes in Muscle Lengths for Tendon-driven Humanoids with Complex Musculoskeletal Structures

TL;DR

This work tackles joint-angle estimation in tendon-driven humanoids with complex structures where encoders are infeasible. It introduces polyarticular-JMMs by reducing DOFs and partitioning joints and muscles into manageable groups, enabling accurate estimation of multiple joints with cross-crossing polyarticular muscles. It further extends the approach to relative-JAE, allowing estimation using only relative muscle-length changes and avoiding calibration of absolute lengths, with an EKF framework adapted accordingly. Validation includes simulation and real-hardware experiments on the Kengoro humanoid, demonstrating feasibility, practical computation times, and improved robustness, albeit with remaining model-mismatch challenges. Overall, the method advances scalable, calibration-free joint-angle estimation for tendon-driven systems and offers a path toward automatic grouping and more accurate JMM learning in future work, with potential applicability to similar robots and exoskeletons.

Abstract

Tendon-driven musculoskeletal humanoids typically have complex structures similar to those of human beings, such as ball joints and the scapula, in which encoders cannot be installed. Therefore, joint angles cannot be directly obtained and need to be estimated using the changes in muscle lengths. In previous studies, methods using table-search and extended kalman filter have been developed. These methods express the joint-muscle mapping, which is the nonlinear relationship between joint angles and muscle lengths, by using a data table, polynomials, or a neural network. However, due to computational complexity, these methods cannot consider the effects of polyarticular muscles. In this study, considering the limitation of the computational cost, we reduce unnecessary degrees of freedom, divide joints and muscles into several groups, and formulate a joint angle estimation method that takes into account polyarticular muscles. Also, we extend the estimation method to propose a joint angle estimation method using only the relative changes in muscle lengths. By this extension, which does not use absolute muscle lengths, we do not need to execute a difficult calibration of muscle lengths for tendon-driven musculoskeletal humanoids. Finally, we conduct experiments in simulation and actual environments, and verify the effectiveness of this study.

Figures (11)

• Figure 1: Complexity of tendon-driven musculoskeletal humanoids. This figure shows that the neck and shoulder girdle of Kengoro have very complex structures like human beings.
• Figure 2: Overview of the joint angle estimation methods: table-search icra2010:nakanishi:table, EKF with polynomial regression humanoids2015:okubo:muscle-learning, and EKF with neural network ral2018:kawaharazuka:vision-learning.
• Figure 3: Correspondence between joints and muscles of Kupper limb of Kengoro. The joints that each muscle can move are painted orange. NT is the neck top 1-DOF joint, N is the neck 3-DOF joint, SC is the sternoclavicular joint, AC is the acromioclavicular joint, GH is the glenohumeral joint, E is the elbow joint, and RU is the radioulnar joint.
• Figure 4: Details of the scapula structure. The right figure shows the change in DOFs arrangement.
• Figure 5: An example of joint-muscle mapping construction.