Artificial intelligence (AI) and computer vision enable modern assistive systems. Within the AktiMuW project, we present a multi-sensor rollator designed under strict latency, energy, and cost constraints to assess posture and provide guidance for safe, comfortable walking. We evaluate three single-board computers (SBCs): NVIDIA Jetson Nano, NVIDIA Jetson Orin Nano, and Raspberry Pi 5 with AI HAT+, by running the same YOLO-based detector on an identical rollator-perspective video, using each platform’s native acceleration (TensorRT on Jetson, Hailo runtime on Raspberry Pi 5) and PyTorch as a baseline where applicable. We report accuracy, end-to-end latency and inference time, frames per second (FPS), CPU/GPU utilization, energy per inference, and cost per FPS. Results show clear trade-offs: TensorRT on Jetson Orin Nano delivers the highest throughput and lowest latency; Raspberry Pi 5 with Hailo offers the best energy efficiency and cost per FPS; Jetson Nano achieves moderate real-time performance at low device cost. These findings provide practical guidance for selecting embedded AI hardware in assistive mobility, healthcare robotics, and smart environments, and they inform the design of an energy-aware, cost-effective prototype for the AktiMuW rollator.
Keywords
NVIDIA Jetson Orin NanoNVIDIA Jetson NanoRaspberry Pi 5YOLOTensorRTHailo AI HAT+Embedded AIComputer VisionAssistive MobilityRollatorBenchmarkingEnergy EfficiencyCost-Performance.
References
World Health Organization, “Ageing and Health,” Oct. 2025, [Online]. Available: https://www.who.int/news-room/fact-sheets/detail/ageing-and-health, [Accessed: Feb. 24, 2026].
Eurostat, Ageing Europe: Looking at the Lives of Older People - 2023 Edition. Luxembourg: Publications Office of the European Union, 2023, [Online]. Available: https://ec.europa.eu/eurostat.
Y. LeCun, Y. Bengio, and G. Hinton, “Deep learning,” Nature, vol. 521, pp. 436-444, May 2015, doi: 10.1038/nature14539.
E. J. Harris, I. Khoo, and E. Demircan, “A survey of human gait-based artificial intelligence applications,” Frontiers in Robotics and AI, vol. 8, p. 749274, 2021, doi: 10.3389/frobt.2021.749274.
A. Macie, T. Matson, and A. Schinkel-Ivy, “Age affects the relationships between kinematics and postural stability during gait,” Gait & Posture, vol. 102, pp. 86-92, 2023, doi: 10.1016/j.gaitpost.2023.03.004.
E. Blasch et al., “Machine learning/artificial intelligence for sensor data fusion – Opportunities and challenges,” IEEE Aerospace and Electronic Systems Magazine, vol. 36, no. 7, pp. 80-93, Jul. 2021, doi: 10.1109/MAES.2020.3049030.
S. M. Alfayeed and B. S. Saini, “Human gait analysis using machine learning: A review,” in Proc. International Conference on Computational Intelligence and Knowledge Economy (ICCIKE), Dubai, UAE, pp. 550-554, 2021, doi: 10.1109/ICCIKE51210.2021.9410678.
S. Twieg, R. V. Menghani, S. Rajanayagam, and V. Kurumbaparambil, “Sensor-based gait analysis: A comparative study of ultrasonic and laser sensors for gait monitoring in rollator-assisted walking,” 2025, [Online]. Available: https://www.researchgate.net/publication/392553747_Sensor-Based_Gait_Analysis_A_Comparative_Study_of_Ultrasonic_and_Laser_Sensors_for_Gait_Monitoring_in_Rollator-Assisted_Walking, [Accessed: Feb. 24, 2026].
R. V. Menghani, S. Rajanayagam, and S. Twieg, “Optimized indoor and outdoor SLAM for a quadruped-based robot walking partner to safely promote elderly mobility and support nursing staff,” in Proc. Irish Signals and Systems Conference (ISSC 2025), Letterkenny, Ireland, Jun. 9-10, 2025.
NVIDIA, “tegrastats – System Monitoring Utility,” Jetson Linux Developer Guide, Jan. 16, 2026, [Online]. Available: https://docs.nvidia.com/jetson/, [Accessed: Feb. 24, 2026].
NVIDIA, “nvpmodel and jetson_clocks – Power and Performance Controls,” Jetson Linux Developer Guide, [Online]. Available: https://docs.nvidia.com/jetson/, [Accessed: Feb. 24, 2026].
·
This work is licensed under a
Creative Commons Attribution-ShareAlike 4.0 International License
