Note that the robot dog in the diagram below seems to have successfully balanced the seemingly contradictory goals of load capacity, system complexity, and motion flexibility:
It is keenly aware of human intentions, adapts movement accordingly, and carries heavy loads on rough terrain, showing excellent following ability and adaptability.
In the field of robotics, quadruped robots have always been a research hotspot due to their flexible movement ability and the advantages of adapting to complex terrain. However, when it comes to high-load collaborative tasks, quadruped robots often seem inadequate.
Researchers at the Dynamic Leg Systems Laboratory of the Italy Institute of Technology (IIT) have come up with an ingenious solution – the Passive Arm Collaborative Handling (PACC) system. This innovation not only greatly improves the collaboration capabilities of quadruped robots, but also opens up new possibilities for human-machine collaboration.
Let's learn more about this technological breakthrough~
▍ Travel light: the subtleties of the passive arm design
Traditional robotic collaborative handling often relies on actively controlled robotic arms, which not only increases the complexity of the system, but also greatly reduces the load capacity of the robot. The IIT research team came up with the ingenuity to design a passive arm structure with three degrees of freedom. Behind this deceptively simple design lies a deep engineering ingenuity.
The passive arm uses a yaw-pitch-pitch joint configuration, and the torque of each joint is generated by the telescopic and damping elements of the spring. This design not only greatly reduces weight, but also increases the robustness of the system. What's more, it introduces inherent compliance between the load and the robot, which is essential for safe and stable movement on complex terrain.
From a mechanical design point of view, many details of the PACC have been carefully considered. For example, the position of the end effector (hook) is deliberately designed near the height of the robot's center of mass. When carrying heavy loads, the actual working height of the hook will be slightly lower than the resting position due to gravity, thus effectively reducing the adverse effects of horizontal forces on the zero moment point (ZMP) of the robot.
Another ingenuity is the design of the third linkage. Its movement is inspired by a single pendulum, which makes it possible to concentrate the most important external interaction forces in the handling task in a plane perpendicular to the third joint axis. This design facilitates the estimation of load motion and the prediction of external forces, laying the foundation for subsequent motion control.
▍Intelligent navigation: Passive arm-led motion control
Another highlight of the PACC system is its unique motion control strategy. The research team has developed a simple and effective motion guidance method based on the angular displacement of the passive arm. This method is particularly suitable for motion control of follower robots in collaborative handling tasks.
Specifically, the system uses the angular displacement of the third joint to control the forward speed of the robot, while the angular displacement of the first joint is used to control the steering speed. The researchers cleverly designed a neutral angle range and two speed classes to adjust the speed by the deflection angle of the joint. To ensure smoothness of motion, they also applied a second-order low-pass filter to process the command signal.
This passive arm-based motion guidance method is not only simple and intuitive, but also highly adaptable to changes in the external environment. It enables the follower robot to naturally follow the movement of the leader, whether it is another robot or a human operator, without the need for complex sensor systems or communication protocols.
▍MPC: The intelligent brain of collaborative handling
To realize the full potential of the PACC system, the research team developed an innovative Model Predictive Control (MPC) scheme. This controller not only takes into account the dynamics of the robot itself, but also incorporates the dynamic characteristics of the passive arm and the estimation of external forces during collaborative handling.
The core task of the MPC controller is to plan the foot-end trajectory and generate optimal motion instructions. It first generates a preliminary foot-end position based on the desired body speed and gait parameters. Then, the support polygon is adjusted to take into account the external forces at the end effector. This adjustment ensures that the robot remains stable even when subjected to external loads.
A particularly noteworthy innovation is the controller's ability to use the characteristics of the passive arm to estimate the external forces at the end effector. By ignoring the small inertia term and the Coriolis force under low-speed motion, the researchers proposed a simplified but efficient method for estimating the force. This method is not only simple to calculate, but for pure inertial loads, it is not even necessary to know the exact load mass to estimate the oscillation dynamics of the load.
Another highlight of the MPC controller is its distributed nature. Each robot operates its own controller independently, eliminating the need for central coordination. This design greatly improves the scalability and robustness of the system, making multi-robot collaboration or human-robot collaboration more flexible and reliable.
▍Experimental verification: deduction from theory to practice
In order to verify the actual performance of the PACC system, the research team designed a series of challenging experimental scenarios. These experiments not only include collaborative handling between robots, but also involve scenarios of human-machine collaboration, fully demonstrating the versatility and adaptability of the system.
In a robotic collaboration experiment, the researchers tested two different ways to connect loads: rigid and non-rigid. Experiments with rigid connections simulate handling tasks that require precise coordination, while non-rigid connections are closer to rope dragging scenarios that are common in real-world applications. The results showed that the PACC system performed well in both situations, with the robots working together stably and maintaining good coordination even in the face of obstacles such as steps.
These experiments not only verify the technical feasibility of the PACC system, but more importantly, they demonstrate the great potential of the system in practical applications. Whether it's moving materials at a construction site or transferring the wounded during a rescue mission, PACC systems have shown clear advantages.
Overall, this work by the IIT research team provides an innovative and practical solution for collaborative handling tasks in quadruped robots. By cleverly combining passive mechanical design and intelligent control algorithms, it not only expands the application range of quadruped robots, but also opens up new possibilities for human-robot collaboration in the future.
See the paper for details of this study:
https://arxiv.org/abs/2403.19862