MPC-CBF with Adaptive Safety Margins for Safety-critical Teleoperation over Imperfect Network Connections

TL;DR

The main novelty of the paper is a method to make the CBFs robust against the uncertainties caused by the network delays affecting the system's state and do so in a less conservative manner.

Abstract

The paper focuses on the design of a control strategy for safety-critical remote teleoperation. The main goal is to make the controlled system track the desired velocity specified by an operator while avoiding obstacles despite communication delays. Control Barrier Functions (CBFs) are used to define the safety constraints that the system has to respect to avoid obstacles, while Model Predictive Control (MPC) provides the framework for adjusting the desired input, taking the constraints into account. The resulting input is sent to the remote system, where appropriate low-level velocity controllers translate it into system-specific commands. The main novelty of the paper is a method to make the CBFs robust against the uncertainties caused by the network delays affecting the system's state and do so in a less conservative manner. The results show how the proposed method successfully solves the safety-critical teleoperation problem, making the controlled systems avoid obstacles with different types of network delay. The controller has also been tested in simulation and on a real manipulator, demonstrating its general applicability when reliable low-level velocity controllers are available.

Figures (6)

• Figure 1: High-level problem architecture. The two larger dashed boxes represent the two sides of the network. The controller executes on the same side as the operator controlling the system, while the system and the low-level controller are on the opposite side. The thicker red arrows represent the information shared between the two sides through the remote connection, while the thinner ones represent the information exchanged between elements operating in the same location. The highlighted block symbolizes the main focus of the paper, which is an optimization-based controller in charge of modifying the desired velocity specified by the operator to enhance the safety guarantees of the system.
• Figure 2: Node architecture. Graph showing the connections between the different entities involved in the project. Each entity corresponds to a ROS 2 node and is represented with an ellipse, while the topics are in rectangles. The graph has been redrawn from that obtained using the rqt_graph tool.
• Figure 3: Desired velocity magnitude law. Relationship between the desired velocity magnitude specified by the tester and the system's distance from the target position. When the distance is smaller than 0.15m, its magnitude is at the lower limit 0.05m/s. On the other hand, when the distance is greater than 0.50m, the magnitude is at its upper limit 0.50m/s.
• Figure 4: Simulation results. (a) $xz$-plane projections of the paths followed by the system for a task in simulation with obstacles in red and the target points as yellow circles. The communication is affected by a Gaussian delay with 50ms mean and 20ms standard deviation. (b) A minimum distance to obstacle for both methods and the same task. (c) Statistical results for all tasks showing minimum obstacle distance. The whiskers denote the most extreme values.
• Figure 5: Hardware setup. (a) 3D rendering of the UR5 manipulator and the target task in rviz. The red sphere represents the obstacle the end effector has to avoid, while the yellow ones target points the end effector has to reach. (b) The real UR5 manipulator used in the experiment.