BioAxis: A Brain Controlled Robotic Arm using EMG Signals and Edge AI

Abstract

The restoration of upper-limb functionality through intelligent assistive systems has become a central focus in biomedical engineering. This study introduces BioAxis, a brain-inspired robotic arm platform that utilizes Electromyography (EMG) signals combined with Edge Artificial Intelligence (Edge AI) for real-time motion control. Unlike conventional brain–computer interface systems that rely solely on EEG, the proposed architecture captures muscle activation patterns from residual limb regions, enabling more stable and intention-driven control.Surface EMG signals are acquired, filtered, segmented, and processed through embedded edge-based machine learning models to classify motor intentions such as grasping, releasing, wrist rotation, and elbow flexion. The integration of Edge AI ensures low latency, improved privacy, and reduced dependency on cloud computing. A multi-degree-of-freedom robotic arm is designed to replicate natural limb kinematics, allowing intuitive interaction.

Introduction

the BioAxis system, an intelligent EMG-based robotic arm designed to improve assistive mobility for individuals with upper-limb impairment using Edge AI and machine learning.

Loss of arm mobility significantly reduces independence, and traditional prosthetic systems often suffer from latency, high cost, and reliance on cloud computing. To address this, BioAxis uses surface EMG signals, which capture muscle electrical activity and provide reliable input for controlling robotic movement.

The system leverages Edge AI, meaning all signal processing and AI inference are done locally on embedded hardware. This reduces delay, improves privacy, and enables real-time control without needing cloud connectivity.

Key components of BioAxis:

• Signal Acquisition Unit: Captures and filters EMG signals from muscles
• Edge Processing Unit: Extracts features and classifies intended movements using lightweight ML models
• Control Mapping Layer: Converts predicted intentions into joint-level robotic commands
• Robotic Actuation System: Executes movements using motors, sensors, and mechanical joints

Literature review summary:

Previous research shows progress in EMG-based prosthetics, EEG brain-controlled systems, and hybrid EEG–EMG models, but most suffer from issues like high computational cost, latency, or dependence on cloud processing. Edge AI approaches are emerging as a solution for real-time performance.

Research gap:

Existing systems often rely on EEG (less stable), cloud computing (high latency), lack personalization, and are too computationally heavy for portable devices.

BioAxis contributions:

• Uses EMG as the primary control signal
• Runs ML models on edge devices
• Supports real-time, low-latency control
• Enables user-specific adaptive learning
• Offers a cost-effective, portable design

Advantages:

The system provides real-time response, privacy, portability, energy efficiency, adaptive control, and affordability, making it suitable for rehabilitation and assistive robotics.

Future scope:

Improvements may include hybrid EEG–EMG integration, optimized deep learning models for edge devices, and self-learning adaptive systems that adjust to long-term user changes.

Conclusion

BioAxis proposes a novel integration of EMG-based intention decoding and Edge AI-driven robotic control. By shifting computation to embedded platforms and focusing on muscular signals for direct motion interpretation, the system enhances responsiveness, portability, and affordability. This approach demonstrates strong potential for next-generation prosthetic and rehabilitation technologies, particularly in resource-limited settings.

References

[1] E. Kim, J. Shin, Y. Kwon, and B. Park, “EMG-Based Dynamic Hand Gesture Recognition Using Edge AI for Human–Robot Interaction,” Electronics, vol. 12, no. 7, Art. no. 1541, Mar. 2023, doi: 10.3390/electronics12071541. [2] K. P. N. V. Satya Sree, T. Bikku, S. Mounika, N. Ravinder, M. Lakshmana Kumar, and C. Prasad, “EMG Controlled Bionic Robotic Arm using Artificial Intelligence and Machine Learning,” in Proc. 2021 Fifth Int. Conf. I-SMAC (IoT in Social, Mobile, Analytics and Cloud), Amaravati, India, 2021, doi: 10.1109/I-SMAC52383.2021.9640623. [3] “EMG Controlled Bionic Robotic Arm using Artificial Intelligence and Machine Learning,” SciSpace PDF, 2020. [4] H. F. Hassan, S. J. Abou-Loukh, and I. K. Ibraheem, “Teleoperated Robotic Arm Movement Using EMG Signal With Wearable MYO Armband,” arXiv, Oct. 2018. [5] “Implementation of a Brain-Computer Interface for Robotic Arm Control,” IEEE Xplore, 2024. [6] “EMG Controlled Bionic Robotic Arm using Artificial Intelligence and Machine Learning — Amrita Vishwa Vidyapeetham,” IEEE Xplore, 2021. [7] R. Li, S.-W. Ye, and Z.-Z. Shi, “SVM-KNN Classifier — A New Method of Improving Accuracy for EMG Gesture Recognition,” in Proc. IEEE Int. Conf. Comput. Sci. & Inf. Technology, Aug. 2009. [8] E. Xu, W. Li, D. Liu, K. Zhang, M. Miao, G. Xu, and A. Song, “Continuous Hybrid BCI Control for Robotic Arm Using Noninvasive EEG, Computer Vision, and Eye Tracking,” Mathematics, vol. 10, no. 4, pp. 618, Feb. 2022, doi: 10.3390/math10040618. [9] “Robotic arm control system based on brain-muscle mixed signals,” Elsevier Biomed. Signal Process. & Control, vol. 77, Aug. 2022, doi: 10.1016/j.bspc.2022.103754. [10] “Hybrid EEG-EMG Based Brain Computer Interface (BCI) System for Real-Time Robotic Arm Control,” Am. J. Mod. Lett. Eng. Tech. Sci., 2019, doi: 10.5185/amlett.2019.2171.

Copyright

Copyright © 2026 Prof. D. A. Kapgate, Roshani Musale, Prerna Khode, Shantanu Jangade, Dharamveer Bhardwaj. 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.