Integration of the TIAGo Robot into Isaac Sim with Mecanum Drive Modeling and Learned S-Curve Velocity Profiles

TL;DR

The paper tackles sim-to-real transfer and rapid learning for a dual-arm, mecanum-wheeled mobile manipulator by embedding the TIAGo++ Omni into Isaac Sim. It introduces two drive models—a physically accurate mecanum-wheel dynamics model and a high-efficiency velocity-based alternative—augmented by a neural-network calibration that predicts S-shaped wheel-velocity curves $S_{oldsymbol{\Theta}}$ from commanded velocity changes using minimal data. Key contributions include a complete Isaac Sim integration, a roller-collider wheel model, and a data-driven method to approximate realistic acceleration profiles, enabling RL and perception workflows. The work demonstrates that proportional wheel acceleration is crucial for smooth omnidirectional motion and provides open-source tooling to accelerate future research in mobile manipulation and learning-based control within Isaac Sim.

Abstract

Efficient physics simulation has significantly accelerated research progress in robotics applications such as grasping and assembly. The advent of GPU-accelerated simulation frameworks like Isaac Sim has particularly empowered learning-based methods, enabling them to tackle increasingly complex tasks. The PAL Robotics TIAGo++ Omni is a versatile mobile manipulator equipped with a mecanum-wheeled base, allowing omnidirectional movement and a wide range of task capabilities. However, until now, no model of the robot has been available in Isaac Sim. In this paper, we introduce such a model, calibrated to approximate the behavior of the real robot, with a focus on its omnidirectional drive dynamics. We present two control models for the omnidirectional drive: a physically accurate model that replicates real-world wheel dynamics and a lightweight velocity-based model optimized for learning-based applications. With these models, we introduce a learning-based calibration approach to approximate the real robot's S-shaped velocity profile using minimal trajectory data recordings. This simulation should allow researchers to experiment with the robot and perform efficient learning-based control in diverse environments. We provide the integration publicly at https://github.com/AIS-Bonn/tiago_isaac.

Figures (8)

• Figure 1: Sample view of the TIAGo++ Omni integrated into Isaac Sim.
• Figure 2: (a) Simulation in Isaac Sim and (b) corresponding sensor data visualization in RViz 2.
• Figure 3: Model of (a) the mecanum wheel and (b) the roller colliders.
• Figure 4: Example of an S-shaped velocity curve fitted to wheel velocity data. The velocity command is to move in the x-direction at $0.35$ m/s. Since all wheels behave identically in this scenario, only one wheel's velocity is shown.
• Figure 5: (a) Wheel velocity curves for different commands in the x-direction (red curves), y-direction (blue curves), and rotational motion (green curves). The S-curve shape exhibits some inconsistencies, likely due to noise from the real robot's PID controller. (b) Predicted S-curves from the fitted model.