A Framework for the Systematic Evaluation of Obstacle Avoidance and Object-Aware Controllers

TL;DR

This work introduces a structured framework for evaluating object-aware controllers (OACs) that react to dynamic obstacles in real time. By centering analysis on three design considerations—kinematics, motion profiles, and virtual constraints—and validating with two simple SRDO/DRDO experiments on a 7-DoF Panda arm, the authors compare Flacco, Ding, and Escobedo within a unified QP-based formulation. They reveal that many OACs neglect robot kinematics, produce discontinuous constraints, and apply unweighted criteria, limiting safe and smooth operation. The framework provides measurable benchmarks and actionable insights to guide the design and comparison of future obstacle-avoidance controllers, with potential for library-style benchmarking in the community.

Abstract

Real-time control is an essential aspect of safe robot operation in the real world with dynamic objects. We present a framework for the analysis of object-aware controllers, methods for altering a robot's motion to anticipate and avoid possible collisions. This framework is focused on three design considerations: kinematics, motion profiles, and virtual constraints. Additionally, the analysis in this work relies on verification of robot behaviors using fundamental robot-obstacle experimental scenarios. To showcase the effectiveness of our method we compare three representative object-aware controllers. The comparison uses metrics originating from the design considerations. From the analysis, we find that the design of object-aware controllers often lacks kinematic considerations, continuity of control points, and stability in movement profiles. We conclude that this framework can be used in the future to design, compare, and benchmark obstacle avoidance methods.

Figures (10)

• Figure 1: This diagram depicts a scenario involving a robotic manipulator and two dynamic obstacles, shown as small black objects. The colored rays emanating from these obstacles represent the distance from the obstacles to various control points along the robot's body. The red ellipsoid illustrates the current range of motion for the end-effector. The ellipsoid and rays are components of the analysis of object aware controllers presented in this work.
• Figure 2: Each column represents one of the three design considerations used to evaluate controllers in this work, introduced in \ref{['sec:evaluation']}. a) High-manipulability configuration as shown from the large ellipsoids. Eigenvectors of $\mathbf{J J}^{\top}$, where $\mathbf{J}$ is the robot Jacobian introduced in \ref{['sec:evaluation']}, are shown as orange arrows. b) An undesirable configuration due to movement restrictions caused by the red object, the smaller the ellipsoids are, the less ability the robot has to move a particular point freely. c) Static control points are determined by the controller designer and placed in predefined locations. Used in flacco2015control. d) Dynamic control points are selected by determining the closest point on the robot body to an obstacle. Used in ding2020collision and escobedo2021contact. e) Low jerk is exhibited and the robot smoothly moves away from the obstacle as seen by the EE orange path. f) The movement restrictions imposed on the EE cause the jittery motion as seen in the jagged path to its final location.
• Figure 3: Flacco flacco2012depth controller diagram with obstacles represented as red circles. The orange arrow originating from the end-effector is introduced in \ref{['eq:repulsive_vector']} as a repulsive force. Blue rectangles show joint restrictions caused by constraints added through \ref{['eq:flacco_body']}.
• Figure 4: Ding ding2020collision controller diagram with obstacles represented as red circles. The rectangle encapsulating the EE shows movements restrictions from \ref{['eq:ding_restriction']}, the orange portion shows restricted movement while the green shows where the robot can still move. This first restriction is also added to the body control points, when obstacles are nearby. The rectangle near the robot body shows the weighted sum distance gradient restriction introduce in \ref{['eq:ding_gradient']}.
• Figure 5: Escobedo escobedo2021contact controller diagram with obstacles represented as red circles. The rectangles shows movements restrictions from \ref{['eq:ding_restriction']}, the orange portion shows restricted movement while the green shows where the robot can still move. The orange arrow shows the scaled EE velocity introduced in \ref{['eq:velocity_scaling']}, the black arrow is the initial velocity.