Abstract:Complex robotic systems require whole-body controllers to deal with contact interactions, handle closed kinematic chains, and track task-space control objectives. However, for many applications, safety-critical controllers are important to steer away from undesired robot configurations to prevent unsafe behaviors. A prime example is legged robotics, where we can have tasks such as balance control, regulation of torso orientation, and, most importantly, walking. As the coordination of multi-body systems is non-trivial, following a combination of those tasks might lead to configurations that are deemed dangerous, such as stepping on its support foot during walking, leaning the torso excessively, or producing excessive centroidal momentum, resulting in non-human-like walking. To address these challenges, we propose a formulation of an inverse dynamics control enhanced with exponential control barrier functions for robotic systems with numerous degrees of freedom. Our approach utilizes a quadratic program that respects closed kinematic chains, minimizes the control objectives, and imposes desired constraints on the Zero Moment Point, friction cone, and torque. More importantly, it also ensures the forward invariance of a general user-defined high Relative-Degree safety set. We demonstrate the effectiveness of our method by applying it to the 3D biped robot Digit, both in simulation and with hardware experiments.
What problem does this paper attempt to address?
The paper primarily focuses on addressing the following issues:
### Research Background and Objectives
- **Complex Robotic Systems**: For complex robotic systems with multiple degrees of freedom (DOF), especially humanoid robots, there is a need to develop whole-body controllers capable of handling contact interactions, closed-chain mechanisms, and task-space control objectives.
- **Safety Challenges**: In practical applications, safety-critical controllers are essential to avoid undesired robot configurations and prevent unsafe behaviors.
### Specific Problems
- **Coordination Difficulty of Multiple Limbs**: The dynamic coupling between the legs and arms of humanoid robots makes it challenging to control these parts independently without causing instability and performance degradation.
- **Unsafe Task-Space Control**: Existing non-safe task-space controllers may successfully track desired targets but fail to check whether joints are approaching mechanical limits or if there is a risk of self-collision for the robot.
- **Implementation of Safety Constraints**: Current methods struggle to effectively incorporate a safety layer into the control strategy, ensuring that the control system prioritizes safety.
### Main Contributions
- **Novel Inverse Dynamics Controller**: A new safety inverse dynamics controller based on Quadratic Programming (QP) is proposed, featuring:
- Avoidance of mass matrix inversion by directly considering joint accelerations, torques, and constraint forces as decision variables, which enhances numerical efficiency.
- Convenient integration of closed-chain constraints and other critical constraints, such as contact, Zero Moment Point (ZMP), and friction cones.
- Utilization of Acceleration-based Exponential Control Barrier Functions (A-ECBF) to ensure the invariance of predefined safety sets, thereby guaranteeing system safety.
- Extensive simulation and hardware experiments on a three-dimensional humanoid robot validate the effectiveness and safety of the controller in performing complex tasks.
### Key Aspects of the Solution
- **Mathematical Modeling**: A detailed mathematical description of the dynamics of humanoid robots is provided, including floating base coordinates, contact constraints, closed-chain mechanisms, and key dynamic constraints.
- **Task-Space Inverse Dynamics Control Algorithm**: A QP-based task-space inverse dynamics control algorithm is proposed, with the introduction of Acceleration-based Exponential Control Barrier Functions to enhance system safety.
- **Experimental Results**: A series of simulations and hardware experiments demonstrate the effectiveness of the controller, particularly in tasks such as squatting, bowing, arm movements, and walking, and show how collisions can be avoided through A-ECBF.
In summary, this paper aims to solve the design problems of whole-body controllers in complex robotic systems, especially addressing the safety and stability challenges of humanoid robots performing complex tasks. By proposing a novel inverse dynamics controller, it not only improves control performance but also ensures system safety.