An integrated behavior tree, quadratic programming, and cartesian compliance control framework for autonomous
door traversal in mobile manipulators

Autonomous door traversal for mobile manipulators poses significant challenges due to environmental uncertainties, precise control requirements, and real-time adaptability. This study proposes an innovative framework integrating Behavior Trees (BT), Quadratic Programming (QP), and Cartesian Compliance Control (CCC)  to address these challenges. BTs provide a modular and flexible decision-making structure, enabling the system to adapt dynamically to changing environments and complex task requirements. However, BTs alone face limitations in precise positioning, which can hinder their effectiveness in high-accuracy tasks.

To overcome these constraints, QP is employed to optimize motion planning under multi-objective constraints, ensuring precise end-effector positioning along designated paths. CCC complements this by dynamically adjusting the system’s compliance, enabling smooth force control and stability during interactions with door mechanisms. Together, these components form a robust framework that enhances the adaptability, precision, and decisionmaking flexibility of mobile manipulators in dynamic environments.

Experiments conducted in both Gazebo simulations and real-world settings validate the framework’s effectiveness. Simulation results demonstrated the system’s ability to adapt to positional uncertainties and varying door states, achieving precise trajectory tracking and force regulation. Real-world experiments further confirmed these findings, with the proposed CCC reducing peak applied force to 4.4 N compared to 11.9 N under traditional velocity control methods. Moreover, integration of QP and CCC improved manipulability metrics and reduced trajectory errors, showcasing the framework’s ability to maintain stability and accuracy under complex interaction scenarios.

Key experimental insights include the framework’s capacity to mitigate errors arising from positioning uncertainties and its ability to handle external disturbances effectively. Statistical analyses and comparative studies highlight the superiority of the proposed approach over traditional and state-of-the-art methods, particularly in terms of force minimization and trajectory accuracy. The inclusion of graphical system overviews and detailed experimental results supports the comprehensive evaluation of the framework (see Figure 1).

This research contributes to advancing robotic manipulation by addressing critical limitations in autonomous door traversal tasks. By leveraging the complementary strengths of BT, QP, and CCC, the proposed framework provides a scalable solution for mobile manipulators operating in humanshared environments such as healthcare and industrial facilities. Future work will explore extending this approach to other dynamic tasks requiring realtime adaptability and precise force control.