Biomimetic Walking Trajectory in Kinaesthetic Learning Sustaining Memory Dependencies in AI Patterns of Humanoid Robot using LSTM in RNN

Abstract

The training results during the trajectory mobile robot obtained and the overcome In this chapter, of obstacle using the fusion method with fuzzy logic controller in several reactive learning environments is presented. Analysis of the state vector and the result obtained for the scheduling problem by fuzzy based algorithm and enumeration method is plotted, The Tracking results are arrived are also discussed. Most of the state transition functions are required for achieving the consistency with causality with a response function for fuzzifying the system. Classical dynamic systems by using the time funcitons and concatenation in the function and the desired properties for consistency by using linguistic approximation and the optimization done and the improvement produced using LSTM where by short term memory can withstand over recurrent neural networks as a LSTM, or Long Short-Term Memory, is a type of recurrent neural network (RNN).

Introduction

• Long Short-Term Memory (LSTM) is a specialized Recurrent Neural Network (RNN) designed for handling sequential data (e.g., text, time series, audio).

• LSTMs solve the vanishing gradient problem in standard RNNs, making them ideal for trajectory generation in robotics.

• This research introduces a biomimetic walking trajectory generation method for the biped humanoid robot ARHR, using ZMP (Zero Moment Point) and CoG (Center of Gravity) as key components.

2. Walking Trajectory Generation Method

• The goal is to generate human-like walking patterns on a flat surface using:

  • Fourier series for analytic trajectory generation.

  • ZMP to CoG transformation via a time-segmented approach.

• The system ensures continuous trajectory functions by calculating spline interpolation coefficients.

• The walking model uses the Simple Inverted Pendulum Model (SIPM) to derive ZMP and CoG paths in horizontal and vertical directions.

3. Kinaesthetic Learning for Motion Imitation

• In kinaesthetic teaching, a human demonstrates motions which the robot replicates.

• The ARHR robot, with multiple serial kinematic chains, is programmed to mimic leg and foot motion while ensuring balance and stability.

• Trajectories are generated at specific points using 3rd-order spline interpolation, focusing on swing motion of legs while arms play a passive role for balance.

4. Gait Planning and Parameters

• Key gait parameters:

  • Initial step gap = 0.1 m

  • Step length = 0.2 m

  • Max foot lift height (Zmax) = 0.03 m

• The robot maintains double support phase for stability.

• The center of mass (GCoM) is kept within the support polygon to avoid falls.

5. LSTM Integration for Gait Stability

• LSTM is used to generate hip and CoM trajectories ensuring balanced walking under ZMP criteria.

• The hip trajectory during single and double-foot stance is trained using LSTM to maintain balance and stability.

6. Obstacle Avoidance via Sensor Fusion

• ARHR is equipped with:

  • 4 infrared (IR) sensors

  • 1 electronic compass

• A fuzzy logic system processes sensor data for obstacle detection and avoidance.

• The robot can autonomously choose from five basic motion behaviors to navigate environments and avoid collisions.

7. Simulation and Experimental Validation

• Simulations (ROS and physical tests) on various surfaces show high performance of the proposed method.

• Path planning integrates:

  • Stereo-vision systems

  • Occupancy grid mapping

  • Plane extraction for environment modeling

• Dynamic stability is ensured through ZMP constraints, transforming static paths into stable, full-body motion plans.

8. Contributions and Applications

• Presents a dynamically-stable, collision-free trajectory planner for humanoid robots.

• Combines traditional trajectory methods (ZMP, SIPM) with LSTM-based learning and sensor integration.

• Demonstrates effective motion planning and balance control, applicable to any legged robot operating in complex environments.

References

[1] Carpenter G A, Grossberg S, Markuson N, Reynolds J H, Rosen D B (1992) Fuzzy ARTMAP:A neural network architecture for incremental supervised learning of analog Multidimensional maps. IEEE Trans. on Neural Networks 3(5): 698-713 [2] Watanabe,N and T. Imaizume[1993],” A fuzzy statistical test of fuzzy hyphotheses”. Fuzzy Sets and Systems, 5392), pp. 167-178. [3] Carpenter G A, Grossberg S, Markuson N, Reynolds J H, Rosen D B (1992) Fuzzy ARTMAP:A neural network architecture for incremental supervised learning of analog Multidimensional maps. IEEE Trans. on Neural Networks 3(5): 698-713. [4] Ashok Kumar R, Marimuthu K (2013) Investigation on data concurrency for sensory fusion in humanoid robot. Int J Adv Mech Aeronaut Eng 1:45–49 [5] Pitsch K, Kuzuoka H, Suzuki Y, Sussenbach L, Luff P, Heath C (2009) The first five seconds: Contingent stepwise entry into an interaction as a means to secure sustained engagement. In: IEEE international symposium on robot and human interactive communication (ROMAN) [6] Lee J, Kiser J, Bobick A, Thomaz A (2011) Vision-based contingency detection. In: ACM/IEEE international conference on human-robot interaction (HRI) [7] Chao C, Lee J, Begum M, Thomaz A (2011) Simon plays Simon says: The timing of turn-taking in an imitation game. In: IEEE international symposium on robot and human interactive communication (ROMAN) © 2025 JAAFR | Volume 3, Issue 4 April 2025 | ISSN: 2984-889X | JAAFR.ORG JAAFR2504013 Journal of Advance and Future Research (www.jaafr.org) [8] Sumioka H, Yoshikawa Y, Asada M (2010) Reproducing interaction contingency toward open-ended development of social actions: Case study on joint attention. In: IEEE transactions on autonomous mental development [9] Tranzatto M, Miki T, Dharmadhikari M, Bernreiter L, Kulkarni M, Mascarich F, Andersson O, Khattak S, Hutter M, Siegwart R, Alexis K (2022) Cerberus in the darpa subterranean challenge. Sci Robot 7 (66) : 9742. https://doi.org/10.1126/scirobotics.abp9742 [10] Yang H, Shi J, Carlone L (2021) Teaser: fast and certifiable point cloud registration. IEEE Trans Rob 37(2):314–333. https://doi.org/10.1109/TRO.2020.3033695 [11] Dr.AshokKumar Ramadoss,“Parametric Investigation on Neuromorphic Humanoid robotic control system using deep learning over Recurrent neural network algorithm for nonlinear regression dynamic fitting in Neural schema”, International Journal of Novel Research and Development, IJNRD, vol.10, Issue 1, pp. c493-c500, 2025, E- ISSN : 2456-4184, 2025. Impact Factor 8.76,Paper DOI : https://doi.org/10.5281/zenodo.14892934 Published URL : https://ijnrd.org/viewpaperforall.php?paper=IJNRD2501259 [12] Dr.Ashok Kumar Ramadoss “Designing A Prototype for the Multisensory System in the Generative AI with Dynamics on A Multi Layer Perception (MLP) of A Real-Time Uncertainty Protocol for Humanoid Robotic Control System”, International Journal for Multidisciplinary Research, IJFMR, vol 7,Issue 1, pp.1-12, 2025, E-ISSN: 2582-2160, 2025. Impact Factor 9.24,Paper DOI : https://doi.org/10.36948/ijfmr.2025.v07i01.37150 Published URL: https://www.ijfmr.com/research-paper.php?id=37150 [13] Dr.Ashok Kumar Ramadoss,“Modelling and Recognizing Grounded response as events in Human Robot Interaction through iterative sensory data integration using semi autonomy algorithm and neuronally implementation with Kinematics Control using Deep Belief Network (DBN) in Deep Learning”, International Journal of Novel Research and Development, IJNRD, vol.10, Issue 1, pp. c569-c587, 2025, E- ISSN : 2456-4184, 2025. Impact Factor 8.76, Paper DOI : https://doi.org/10.5281/zenodo.14892951 Published URL :https://ijnrd.org/viewpaperforall.php?paper=IJNRD2501267 [14] Dr.Ashok Kumar Ramadoss, “ Agentic AI with Dynamics and Kinematics Control Incorporated with RPA Attributes in a Humanoid Robotics with Predictive Machine Learning Over Edge Computing” International Journal of Science and Technology, IJSAT, vol.16,Issue 1, pp.1-13,2025, E-ISSN : 2229-7677, Impact Factor 9.88, Paper DOI: https://doi.org/10.71097/IJSAT.v16.i1.1957 Published URL:https://www.ijsat.org/research-paper.php?id=1957 [15] Dr.Ashok Kumar Ramadoss,“An Approach in Recognizing stumble threshold point for the disturbances during the stumble recovery in a Dynamics of Humanoid Robot incorporating Convolutional Neural Network CNN in large learning model LLM”, International Journal of Novel Research and Development, IJNRD, vol.10, Issue 1, pp.C748-C759, 2025, E- ISSN : 2456-4184, Impact Factor 8.76, Paper DOI : https://doi.org/10.5281/zenodo.14892959 Published URL : https://ijnrd.org/viewpaperforall.php?paper=IJNRD2501286 [16] Dr.Ashok Kumar Ramadoss Overcoming the Instability Defects with Reliability Using Alpha Ramda Humanoid Robot for Flexible Trajectory with High Precision Using Generative Adversarial AI Model, International Journal for Multidisciplinary Research, IJFMR, vol 7, Issue 1, pp . 1-10, 2025, E-ISSN : 2582-2160, Impact Factor 9.24 Paper DOI : https://doi.org/10.36948/ijfmr.2025.v07i01.37821 Published URL: https://www.ijfmr.com/research-paper.php?id=37821 [17] Dr.Ashok Kumar Ramadoss,” Cognitive protocol with neural schema implementation using Palletizing approach for humanoid robotic proportional model incorporating hierarchical clustering technique”, IRJMST, vol.16,Issue 2, pp 32-41,2025, E-ISSN : 2250-1959, P-ISSN 2348-9367. Impact Factor 8.1026, Journal DOI : https://doi.org/10.32804/IRJMST, Published URL: http://www.irjmst.com/Artical_details?id=21038 [18] Dr.Ashok Kumar Ramadoss, Fuzzification using Approximation with Fuzzy to Extract Structure and Motion Environments in Dynamic Humanoid Robotic System from the Opted and Uncertainty by Quantization in non-monotone neural networks, International Journal of Advance in Computer Science & its application IJCSIA ,Vol.7:Issue 2, pp 258-262, 2018, ISSN:2250-3765, Institute of Research Engineers and Doctors, Switzerland, Impact Factor 5.0, Published URL: https://journals.theired.org/journals/ijcsia.html ISSN portal : https://portal.issn.org/resource/ISSN/2250-3765. [19] Ashok Kumar Ramadoss& Marimuthu Krishnaswamy , Embedding inference engine in fuzzy expert robotic system shell in a humanoid robot platform for selecting stochastic appropriate fuzzy implications for approximate reasoning, Artif Life Robotics, Springer SG. Germany, vo Springer Nature Link, vol. 20, Issue. 1, 2015, pp. 13-18. ISSN:1433-5298, 2015, Imact Factor 0.8,Paper DOI: https://doi.org/10.1007/S10015-014-0189-2 Published URL: https://link.springer.com/article/10.1007/s10015-014-0189-2. [20] Ashok Kumar Ramadoss, Marimuthu, Krishnaswamy “Prediction of stochastic cognitive neural schema with neural network paradigm of latent semantic nodes in robot humanoid robot using nonlinear regression of Gaussian sigmoidal curves in Boltzmann normalization”, International Information Institute, Inf Int InterdiscipJ,Tokyo, Japan, vol.17, Issue.8, pp.3971-3983,2014, P-ISSN 1343-4500, E-ISSN 1344-8994, Impact factor 2.5, ISSN portal : https://portal.issn.org/api/search?search[]=MUST=allissnbis=%221343- 4500%22&search_id=58751955 Published URL: http://www.information-iii.org/abs_e2.html#No8-2014. [21] Ashok Kumar Ramadoss , Investigation of Data Concurrency of Sensory fusion in Humanoid Robot, IJAMAE - International journal of advancement in mechanical and aeronautical engineering, Vol 1, Issue. 1, pp 45-49, 2013, IRED, Institute of Research engineers and Doctors, NEW YORK,USA..doi:10.3850/ 978- 981-07-6261-2_61: E-ISSN: 2372-4153, Impact Factor 7.3 ISSN portal: https://portal.issn.org/resource/ISSN/2372-4153 Published URL: https://journals.theired.org/journals/en-ijamae.html [22] Ashok Kumar Ramadoss, Transparency in Agentic AI Collaboration in XAI Trustworthy computer applications in Deep Multi language model DMLM for robotic system, JAAFR- Journal of Advance and Future Research, vol 3, Issue 4, pp 105-111, 2025, ISSN:2984-889X, Impact factor : 8.57.Paper DOI: https://doi.org/10.5281/zenodo.15241545 Published Url: https://rjwave.org/jaafr/viewpaperforall.php?paper=JAAFR2504013 [23] Ashok Kumar Ramadoss, Artificial Intelligence for Enhanced Autonomy in Robotics vision for object recognition and tracking using sensor integration for situational awareness, JAAFR- Journal of Advance and Future Research, vol 3, Issue 4 , 2025, pp 117-124, ISSN:2984-889X, Impact Factor : 8.57,Paper DOI: https://doi.org/10.5281/zenodo.15241551 Published Url : https://rjwave.org/jaafr/viewpaperforall.php?paper=JAAFR2504015.

Copyright

Copyright © 2025 Dr. Ashok Kumar Ramadoss. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.