Learning Inverse Kinodynamics for Autonomous Vehicle Drifting

TL;DR

The paper tackles learning an inverse kinodynamic model to improve autonomous drifting on a small-scale vehicle by using teleoperation and IMU data to correct commanded curvature. A compact two-input neural network is trained to predict corrections that align the executed trajectory with the IMU-observed motion, and it is validated on circle and drift tasks. Results show clear improvements in curvature tracking for loose drifts, while tight drifts remain challenging due to limited data and environmental constraints, highlighting the need for richer sensing and more diverse data. Overall, the work demonstrates a data-driven path toward safer, more accurate autonomous drifting with practical implications for kinodynamic planning in robotics.

Abstract

In this work, we explore a data-driven learning-based approach to learning the kinodynamic model of a small autonomous vehicle, and observe the effect it has on motion planning, specifically autonomous drifting. When executing a motion plan in the real world, there are numerous causes for error, and what is planned is often not what is executed on the actual car. Learning a kinodynamic planner based off of inertial measurements and executed commands can help us learn the world state. In our case, we look towards the realm of drifting; it is a complex maneuver that requires a smooth enough surface, high enough speed, and a drastic change in velocity. We attempt to learn the kinodynamic model for these drifting maneuvers, and attempt to tighten the slip of the car. Our approach is able to learn a kinodynamic model for high-speed circular navigation, and is able to avoid obstacles on an autonomous drift at high speed by correcting an executed curvature for loose drifts. We seek to adjust our kinodynamic model for success in tighter drifts in future work.

Figures (31)

• Figure 1: Stanford's MARTYkhana DeLorean drifting on an open space and tracing its path.
• Figure 2: UT AUTOmata F1/10 scale vehicle, used for testing
• Figure 3: PS4 Controller, used for teleoperating the vehicle
• Figure 4: Example sample from our training data
• Figure 5: Diagram of the Anna Hist Gymnasium with example turning trajectories of the vehicle during the 5-min timers.