Data-Driven Autoregressive Power Prediction for GTernal Robots in the Robotarium

TL;DR

This paper introduces a lightweight autoregressive neural network model that accurately predicts power consumption of mobile robots in real-time, capturing complex dynamics beyond simple kinematic models, and generalizes well across different robots and behaviors.

Contribution

The paper presents a novel autoregressive power predictor for mobile robots that outperforms traditional models and enables real-time energy-aware control in multi-robot systems.

Findings

Achieves $R^2=0.90$ on held-out data

Demonstrates zero-shot transfer to unseen robots

Runs in 224 microseconds per inference

Abstract

Energy-aware algorithms for multi-robot systems require accurate power consumption models, yet existing approaches rely on kinematic approximations that fail to capture the complex dynamics of real hardware. We present a lightweight autoregressive predictor for the GTernal mobile robot platform deployed in the Georgia Tech Robotarium. Through analysis of 48,000 samples collected across six motion trials, we discover that power consumption exhibits strong temporal autocorrelation ($\rho_1 = 0.95$) that dominates kinematic effects. A 7,041-parameter multi-layer perceptron (MLP) achieves $R^2 = 0.90$ on held-out motion patterns by conditioning on recent power history, reaching the theoretical prediction ceiling imposed by measurement noise. Physical validation across seven robots in a collision avoidance scenario yields mean $R^2 = 0.87$, demonstrating zero-shot transfer to unseen robots and behaviors. The predictor runs in 224 $\mu$s per inference, enabling real-time deployment at 150$\times$ the platform's 30 Hz control rate. We release the trained model and dataset to support energy-aware multi-robot algorithm development.