Active-Inference Control and Sensor-Fusion Simulation for Bicycle Stabilization in Aging-Rider Mobility Systems

Summary

Low-speed fall events are the leading cause of cycling injuries among older adults, generating 25,000?40,000 annual emergency-department visits and more than $50B in national medical costs. These incidents primarily occur during mounting, slow riding, deceleration, and stopping?contexts where traditional steering-based stability controls, passive safety devices, and prior gyroscopic concepts fail to prevent loss of balance. StaeblTECH has developed a dual-Control-Moment-Gyroscope (CMG) stabilization prototype designed to proactively prevent these falls. Its success depends on a predictive, adaptive control architecture that can learn individual rider characteristics and manage uncertainty. Active Inference (AIF), a unified probabilistic framework for perception, prediction, and action, offers capabilities not available in conventional PID or Model Predictive Control (MPC) systems. This project develops the simulation-based control and sensor-fusion foundation required to integrate AIF into StaeblTECH's stabilization platform. Leveraging MnRI's robotics simulation environment, researchers will: 1) build a high-fidelity digital twin of bicycle, rider, and dual-CMG dynamics; 2) implement and tune AIF controllers using message-passing variational inference; 3) integrate IMU, optical-flow, and load-sensor data to evaluate latency, noise sensitivity, and perceptual accuracy; and 4) benchmark AIF performance against PID and MPC baselines using a seven-degree-of-freedom bicycle model. All outputs including AIF controllers, sensor-fusion models, stability metrics, and digital-twin datasets directly support StaeblTECH's NIH Direct-to-Phase II proposal by providing validated simulation results and de-risking subsequent hardware development. This project strengthens Minnesota's leadership in AI-assisted mobility, addresses a critical aging-transportation challenge, and positions the UMN-StaeblTECH partnership for future federal funding.