OTTO: Dynamics and Control of Wheeled Bipedal Jumping Robot
Paweekorn Buasakorn, Supachai Vongbunyong and Kitti Thamrongaphichartkul
AI summary
Problem
Wheeled bipedal robots struggle to navigate complex terrain and execute stable jumps due to uncontrolled angular momentum during flight and a lack of integrated, disturbance-robust jump planning.
Approach
The authors develop a 3D dynamic model and integrate an LQR-based stance controller with a trajectory-optimized jumping cycle, uniquely repurposing the robot's wheels as reaction wheels to actively control aerial posture and absorb flight disturbances.
Key results
- 3D-WIDP dynamic model with equilibrium point analysis for robust stance control
- QP-based jumping cycle planner for takeoff, flight, and landing phases
- Novel flight posture controller using wheels as reaction wheels for aerial stability
- Validated terrain traversal and stair-jumping capabilities in Gazebo/ROS2 simulation
Why it matters
Provides a robust, model-based alternative to reinforcement learning for agile mobile robots navigating unstructured, obstacle-rich environments.
Abstract
This paper presents OTTO, a study of jumping and terrain traversal for a 6-DOF wheeled bipedal robot that addresses the limitations of purely wheeled or legged locomo- tion when navigating complex and challenging environments. We develop a robot model based on a 3D wheeled inverted pendulum (WIP) system equipped with torso degrees of free- dom. A unified framework integrates LQR-based controllers with jumping strategies to enable effective terrain traversal. The system features a novel flight posture controller that exploits wheels as reaction wheels to actively control aerial posture without predetermined activation timing. Additionally, we implement a comprehensive jumping cycle using trajectory optimization that encompasses a terminal trajectory planner for takeoff and landing phases, as well as a flight phase planner. We execute various comprehensive experiments to assess the capabilities of these behaviors utilizing the Gazebo ROS2 simulation environment. The complete system architecture is evaluated through simulation trials of a prototype wheeled- bipedal robot, illustrating the viability and robustness of the proposed approach for terrain locomotion and jumping.