TLDR: This research introduces a novel method for optimizing three-dimensional wire arrangements in tendon-driven robots, explicitly accounting for wire crossings and ensuring sufficient joint torque along a defined trajectory. Using multi-objective black-box optimization, the study explores complex wire configurations for a 3D link structure, demonstrating how to achieve stable robot movement while preventing wire interference. The findings highlight the generation of unique, often asymmetric, designs for odd wire counts and the trade-offs between preventing crossings and maximizing performance.
Tendon-driven robots, which use wires to control movement, are becoming increasingly popular due to their ability to be lightweight, offer variable stiffness, and provide redundant control. These robots are commonly found in areas like robotic hands, wrists, and waists. However, designing the intricate arrangement of these wires has traditionally relied on trial and error, a method that becomes incredibly difficult as robot structures grow more complex.
Previous attempts to optimize wire arrangements often simplified the problem. Many studies focused on two-dimensional movements, assumed a constant distance between the wire and the joint (moment arm), or completely ignored the critical issue of wires crossing each other. In real-world, complex three-dimensional robots like the SAQIEL manipulator’s shoulder or the Musashi humanoid’s waist, wire crossings are a significant problem, leading to tangling, increased friction, and even wire breakage. This highlights the need for a more sophisticated design approach.
A new research paper, available at arXiv:2507.04235, introduces a groundbreaking method for optimizing three-dimensional wire arrangements in tendon-driven robots. This study specifically addresses the challenge of wire crossings, ensuring that wires do not interfere with each other while still allowing the robot to generate sufficient force (joint torque) along its intended path.
The core of this new method involves a multi-objective black-box optimization approach. This means the system explores many different wire arrangements to find the best balance between two main goals: minimizing wire crossings and maximizing the robot’s ability to exert force at its joints. The researchers designed a two-link, three-dimensional robot structure for their experiments, where wires connect different points on the links to actuate the joints.
To detect wire crossings, the method calculates the minimum distance between any two wire segments as the robot moves. If this distance falls below a tiny threshold, a crossing is detected. For efficiency, they use a sophisticated mathematical technique called Brent’s method to quickly find potential crossing points along the robot’s movement trajectory. The system also checks for wires crossing the robot’s joints or different segments of the same wire crossing if it folds back on itself.
Evaluating the robot’s ability to generate joint torque is another crucial aspect. The researchers calculate the “feasible joint torque space” – essentially, the range of forces the robot can produce at its joints. They ensure that this space is large enough to allow the robot to perform its tasks stably. The optimization aims to maximize this torque capability across all the robot’s movements.
The optimization process uses an algorithm called NSGA-II, which is well-suited for problems with multiple conflicting objectives. Through extensive simulations, the researchers tested various configurations, including different numbers of wires (M) and different ways wires are routed (N, representing relay points). They observed how the design solutions changed based on these parameters and the number of degrees of freedom (DOFs) of the joints.
Key findings from the simulations revealed interesting trends. When the robot had an odd number of wires, the optimized designs often lacked symmetry and were highly complex – arrangements that would be very difficult for a human designer to come up with manually. In contrast, an even number of wires tended to result in more symmetrical and visually appealing designs. The study also found that while preventing wire crossings is essential for building functional robots, allowing some crossings in the design phase could sometimes lead to better performance, though this depends on the specific robot and task.
Interestingly, simply increasing the number of relay points for wires (from N=2 to N=3) often resulted in the wires simply folding back on themselves, which significantly increased the potential torque but didn’t drastically alter the overall wire path. This suggests that while more relay points can boost performance, they might not always lead to fundamentally different or more efficient wire arrangements in terms of complexity.
Also Read:
- Bridging the Gap: High-Performance Robot Control for Low-Cost Manipulators
- Improving Reinforcement Learning Performance with Constrained Heuristic Optimization
This research marks a significant step forward in the design of tendon-driven robots. By providing a systematic way to optimize complex three-dimensional wire arrangements while explicitly considering wire crossings and ensuring stable joint torque, this method paves the way for more robust, efficient, and capable robots that can perform intricate movements without mechanical interference.
