Papers
Topics
Authors
Recent
Gemini 2.5 Flash
Gemini 2.5 Flash
149 tokens/sec
GPT-4o
7 tokens/sec
Gemini 2.5 Pro Pro
45 tokens/sec
o3 Pro
4 tokens/sec
GPT-4.1 Pro
38 tokens/sec
DeepSeek R1 via Azure Pro
28 tokens/sec
2000 character limit reached

Body Design and Gait Generation of Chair-Type Asymmetrical Tripedal Low-rigidity Robot (2404.05932v1)

Published 9 Apr 2024 in cs.RO

Abstract: In this study, a chair-type asymmetric tripedal low-rigidity robot was designed based on the three-legged chair character in the movie "Suzume" and its gait was generated. Its body structure consists of three legs that are asymmetric to the body, so it cannot be easily balanced. In addition, the actuator is a servo motor that can only feed-forward rotational angle commands and the sensor can only sense the robot's posture quaternion. In such an asymmetric and imperfect body structure, we analyzed how gait is generated in walking and stand-up motions by generating gaits with two different methods: a method using linear completion to connect the postures necessary for the gait discovered through trial and error using the actual robot, and a method using the gait generated by reinforcement learning in the simulator and reflecting it to the actual robot. Both methods were able to generate gait that realized walking and stand-up motions, and interesting gait patterns were observed, which differed depending on the method, and were confirmed on the actual robot. Our code and demonstration videos are available here: https://github.com/shin0805/Chair-TypeAsymmetricalTripedalRobot.git

Definition Search Book Streamline Icon: https://streamlinehq.com
References (17)
  1. M. Shinkai, “Suzume,” [Film], 2022.
  2. C. Liu, Q. Chen, and D. Wang, “Cpg-inspired workspace trajectory generation and adaptive locomotion control for quadruped robots,” IEEE Transactions on Systems, Man, and Cybernetics, Part B (Cybernetics), vol. 41, no. 3, pp. 867–880, 2011.
  3. T. Tanikawa, Y. Masuda, and M. Ishikawa, “A reciprocal excitatory reflex between extensors reproduces the prolongation of stance phase in walking cats: Analysis on a robotic platform,” Frontiers in Neurorobotics, vol. 15, p. 636864, 2021.
  4. G. Bledt, M. J. Powell, B. Katz, J. Di Carlo, P. M. Wensing, and S. Kim, “Mit cheetah 3: Design and control of a robust, dynamic quadruped robot,” in 2018 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS).   IEEE, 2018, pp. 2245–2252.
  5. J. Di Carlo, P. M. Wensing, B. Katz, G. Bledt, and S. Kim, “Dynamic locomotion in the mit cheetah 3 through convex model-predictive control,” in 2018 IEEE/RSJ international conference on intelligent robots and systems (IROS).   IEEE, 2018, pp. 1–9.
  6. J. Lee, J. Hwangbo, and M. Hutter, “Robust recovery controller for a quadrupedal robot using deep reinforcement learning,” arXiv preprint arXiv:1901.07517, 2019.
  7. J. Lee, J. Hwangbo, L. Wellhausen, V. Koltun, and M. Hutter, “Learning quadrupedal locomotion over challenging terrain,” Science robotics, vol. 5, no. 47, p. eabc5986, 2020.
  8. X. Huang, Z. Li, Y. Xiang, Y. Ni, Y. Chi, Y. Li, L. Yang, X. B. Peng, and K. Sreenath, “Creating a dynamic quadrupedal robotic goalkeeper with reinforcement learning,” arXiv preprint arXiv:2210.04435, 2022.
  9. T. Haarnoja, B. Moran, G. Lever, S. H. Huang, D. Tirumala, M. Wulfmeier, J. Humplik, S. Tunyasuvunakool, N. Y. Siegel, R. Hafner, et al., “Learning agile soccer skills for a bipedal robot with deep reinforcement learning,” arXiv preprint arXiv:2304.13653, 2023.
  10. K. Kawaharazuka, K. Okada, and M. Inaba, “Realization of seated walk by a musculoskeletal humanoid with buttock-contact sensors from human constrained teaching,” in 2022 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS).   IEEE, 2022, pp. 5774–5780.
  11. K. Sims, “Evolving Virtual Creatures,” in Proceedings of the 21st Annual Conference on Computer Graphics and Interactive Techniques, 1994, pp. 15–22.
  12. S. Ha, J. Kim, and K. Yamane, “Automated deep reinforcement learning environment for hardware of a modular legged robot,” in 2018 15th international conference on ubiquitous robots (UR).   IEEE, 2018, pp. 348–354.
  13. M. Azumi, K. Ayaka, Y. Hironori, H. Jun, N. Jason, and S. Shunta, “Improvised robotic design with found objects,” in Proc. 3rd Conf. NeurIPS Workshop Mach. Learn. Creativity Des., 2018.
  14. J. Bongard, V. Zykov, and H. Lipson, “Resilient machines through continuous self-modeling,” Science, vol. 314, no. 5802, pp. 1118–1121, 2006.
  15. J. Schulman, F. Wolski, P. Dhariwal, A. Radford, and O. Klimov, “Proximal policy optimization algorithms,” arXiv preprint arXiv:1707.06347, 2017.
  16. V. Makoviychuk, L. Wawrzyniak, Y. Guo, M. Lu, K. Storey, M. Macklin, D. Hoeller, N. Rudin, A. Allshire, A. Handa, et al., “Isaac gym: High performance gpu-based physics simulation for robot learning,” arXiv preprint arXiv:2108.10470, 2021.
  17. E. Todorov, T. Erez, and Y. Tassa, “MuJoCo: A physics engine for model-based control,” in Proceedings of the 2012 IEEE/RSJ International Conference on Intelligent Robots and Systems, 2012, pp. 5026–5033.
Citations (1)
View on Semantic Scholar

Summary

  • The paper introduces two gait generation methods—manual posture interpolation and reinforcement learning via PPO—to tackle asymmetry challenges in a low-cost tripedal robot.
  • The posture-connection approach creates predictable, cyclic walking gaits, while the RL method yields adaptable, variable-speed motions.
  • Experimental results confirm that both techniques enable reliable walking and stand-up motions, suggesting promising avenues for further robotic control optimization.

Design and Gait Generation Techniques for an Asymmetrical Tripedal Robot Inspired by Cinematic Imagination

Overview of Robot Design

This paper embarks on the development of a chair-type asymmetrical tripedal low-rigidity robot inspired by a three-legged chair character from the movie "Suzume." The robot's structural uniqueness, characterized by three asymmetrically positioned legs to its body, elevates the challenge in maintaining balance and movement. Utilizing low-cost servo motors for actuation and limited sensory feedback through a single posture quaternion sensor underlines the design philosophy aimed at simplicity and cost-effectiveness.

Gait Generation Methodologies

Two distinct methodologies are introduced for gait generation, each showcasing a separate approach in overcoming the robot's intrinsic limitations.

  1. Connecting Essential Postures: In this method, essential postures necessary for motion are discovered through hands-on experimentation. By linearly interpolating between these key postures, the gait enabling walking and stand-up motions is formulated. However, this process heavily relies on trial and error, making it intuitive but less adaptable to dynamic changes.
  2. Reinforcement Learning (RL) Approach: Contrasting the first method, the RL approach utilizes Proximal Policy Optimization (PPO) in the Isaac Gym simulation environment to generate gaits. This method capitalizes on a comprehensive representation of the robot's model in MJCF format, enabling detailed simulations. Actions are defined as servo motor command angles, with observations comprising servo motor command angles and quaternion outputs over time. The RL model learns to generate gaits for walking and standing up from various postures, indicated by varied robot orientations.

Evaluation of Gait Patterns

An experimental setup evaluates both gait generation methods, measuring walking and stand-up motions. Results reveal:

  • The walking gait generated through connecting postures showcases a predictable, cycle-repeating motion. This offers consistent forward movement but lacks adaptability to unscripted environments.
  • The RL-generated walking gait demonstrates greater unpredictability with variable speed, indicating adaptability but potentially less consistency.
  • Stand-up motions for both methods underline a common strategy of leveraging a counter-motion to initiate standing up. The RL method stands out by leveraging learning from various initial postures, suggesting a composite understanding of effective movement strategies dependent on initial conditions.

Discussion

The findings illustrate that even with a robot having imperfect and asymmetrical body structures, viable gaits for locomotion and corrective movements can be generated through both manual posture connection and RL. The RL-generated gaits exhibit behaviors that consider the robot's asymmetry, indicating a complex understanding learned through iterative trials. Reset conditions within the RL training play a crucial role in avoiding unwanted behavior patterns, underscoring the importance of thoughtful condition setting in gait learning processes.

Conclusion

By introducing different gait generation methods for a uniquely designed asymmetrical tripedal robot, this paper highlights potential pathways to achieve effective motion. Despite inherent structural challenges, the robot successfully performs desired actions, underlining the promise of both hands-on discovery and machine learning approaches in robotic design and control. Future research could focus on enhancing the adaptability and efficiency of these gaits, possibly by combining the strengths of intuitive manual design and the robust adaptability offered by machine learning techniques.

File Document Download Save Streamline Icon: https://streamlinehq.com PDF File Document Download Save Streamline Icon: https://streamlinehq.com Markdown
Youtube Logo Streamline Icon: https://streamlinehq.com

YouTube