D4W: Dependable Data-Driven Dynamics for Wheeled Robots

TL;DR

Experimental results show that D4W achieves the best simulation accuracy compared to traditional approaches, allowing for rapid iteration of wheel robot algorithms with less or no need for fine-tuning in reality.

Abstract

Wheeled robots have gained significant attention due to their wide range of applications in manufacturing, logistics, and service industries. However, due to the difficulty of building a highly accurate dynamics model for wheeled robots, developing and testing control algorithms for them remains challenging and time-consuming, requiring extensive physical experimentation. To address this problem, we propose D4W, i.e., Dependable Data-Driven Dynamics for Wheeled Robots, a simulation framework incorporating data-driven methods to accelerate the development and evaluation of algorithms for wheeled robots. The key contribution of D4W is a solution that utilizes real-world sensor data to learn accurate models of robot dynamics. The learned dynamics can capture complex robot behaviors and interactions with the environment throughout simulations, surpassing the limitations of analytical methods, which only work in simplified scenarios. Experimental results show that D4W achieves the best simulation accuracy compared to traditional approaches, allowing for rapid iteration of wheel robot algorithms with less or no need for fine-tuning in reality. We further verify the usability and practicality of the proposed framework through integration with existing simulators and controllers.

Figures (11)

• Figure 1: Diagram of a unicycle-type wheeled robot. It shows the coordinate reference point $P$, the robot orientation $\theta$, and wheel speeds $w_l$ and $w_r$.
• Figure 2: Overview of the data-driven simulation pipeline in D4W. Two framework phases are shown: 1) data gathering: recording the robot poses with input command windows to form the dataset 2) dynamics learning: Evaluating and training the model with a sliding command and pose window along simulation steps.
• Figure 3: Possible architectures of the data-driven model. From left to right: 1) pure data-driven model 2) dynamical hybrid model 3) kinematic hybrid model 4) analytical parameter model
• Figure 4: Training trajectories under various transformations. Egocentric transformation yields an easier training target.
• Figure 5: Robot commands gathered in data collection. The X and Y axes show the longitudinal and the angular speed commands. Commands are clustered into three types: forward, left turn, and right turn.