Real-Time Tactile Impact Localization on Embedded MCU using Intensity Ratio and Correlation

Abstract

Achieving high-precision, real-time tactile perception on embedded platforms is challenged by limited computational resources and stringent timing constraints. This work presents the implementation and optimization of a previously proposed tactile localization algorithm-based on a hybrid criterion of Signal Intensity Ratio (SIR) and waveform correlation-on a low-cost ESP32-S3 microcontroller. To meet realtime requirements, a dual-core task-scheduling architecture is designed, isolating the high-priority data acquisition pipeline from the computationally intensive localization solver to ensure stream continuity. Key platform-aware optimizations, including a lightweight envelope detector, pre-computed lookup tables, and a parabolic interpolation strategy for efficient peak correlation estimation, significantly reduce the computational burden. A comprehensive evaluation shows the system maintains an average localization accuracy of $\mathbf{1. 8 1} \pm \mathbf{1. 7 7 ~ m m}$ with a total system latency of approximately 57.4 ms, demonstrating real-time capability suitable for interactive applications. Its latency is competitive with some PC-based solutions, as shown in a cross-platform comparison. The system also exhibits low power consumption and high resource efficiency. This work validates the feasibility of executing this algorithm in real-time on a general-purpose MCU, providing a practical step towards its cost-effective deployment.