GM3: A General Physical Model for Micro-Mobility Vehicles
Grace Cai, Nithin Parepally, Laura Zheng, Ming C. Lin
AI summary
Problem
Existing micro-mobility simulators rely on the Kinematic Bicycle Model or mode-specific physics that ignore tire slip, load transfer, and rider lean, making them inadequate for training autonomous systems or simulating complex urban interactions.
Approach
The authors developed GM3, a unified tire-level physics model using the brush tire representation to calculate forces for arbitrary wheel configurations, paired with an interactive framework for real-time control and trajectory visualization.
Key results
- Unified tire-level physics model supporting arbitrary wheel layouts and capturing slip, load transfer, and lean
- Interactive model-agnostic evaluation and visualization framework for real-time control and trajectory comparison
- Reduced Average Displacement Error on real-world trajectories compared to the Kinematic Bicycle Model baseline
- Modular support for diverse control interfaces including manual steering, lean-to-steer, and differential drive
Why it matters
Enables realistic simulation and training of autonomous systems for mixed urban traffic by accurately modeling the complex dynamics of diverse micro-mobility vehicles.
Abstract
Modeling the dynamics of micro-mobility vehi- cles (MMV) is becoming increasingly important for train- ing autonomous vehicle systems and building urban traffic simulations. However, mainstream tools rely on variants of the Kinematic Bicycle Model (KBM) [1]–[4] or mode-specific physics that miss tire slip, load transfer, and rider/vehicle lean. To our knowledge, no unified, physics-based model captures these dynamics across the full range of common MMVs and wheel layouts. We propose the “Generalized Micro-mobility Model” (GM3), a tire-level formulation based on the tire brush representation [5]–[7] that supports arbitrary wheel configurations, including single/double track and multi-wheel platforms. We introduce an interactive model-agnostic evalua- tion and visualization framework that decouples vehicle/layout specification from dynamics to compare the GM3 with the KBM and other models, consisting of fixed step RK4 integration, human-in-the-loop and scripted control, real-time trajectory traces, and logging for analysis. We also empirically validate the GM3 on the Stanford Drone Dataset’s deathCircle (roundabout) scene [8] for biker, skater, and cart classes.