Development and Characterization of an Interprocess Communications Interface and Controller for Bipedal Robots

TR Number



Journal Title

Journal ISSN

Volume Title


Virginia Tech


As robotic systems grow in complexity, they inevitably undergo a process of specialization whereby they separate into an array of interconnected subsystems and individual processes. In order to function as a unified system, these processes rely heavily on interprocess communications (IPC) to transfer information between subsystems and various execution loops. This thesis presents the design, implementation, and validation of the Valor ROS Controller, a hybrid IPC interface layer and robot controller. The Valor ROS Controller connects the motion control system, implemented with the internally created Bifrost IPC, developed by Team VALOR for the DARPA Robotics Challenge (DRC) with the high level software developed by Team ViGIR that uses the Robot Operating System (ROS) IPC framework. The Valor ROS Controller also acts as a robot controller designed to run on THOR and ESCHER, and is configurable to use different control modes and controller implementations. By combining an IPC interface layer with controllers, the Valor ROS Controller enabled Team VALOR to use Team ViGIR's software capabilities at the DRC Finals.

In addition to the qualitative validation of Team VALOR competing at the DRC Finals, this thesis studies the efficiency of the Valor ROS Controller by quantifying its computational resourceful utilization, message pathway latency, and joint controller tracking. Another contribution of this thesis is the quantification of end-effector pose error incurred by whole-body motions. This phenomenon has been observed on both THOR and ESCHER as one of their arms moves through a trajectory, however, it has never been studied in depth on either robot. The results demonstrate that the Valor ROS Controller adequately uses computational resources and has message latencies in the order of 50 ms. The results also indicate several avenues to improve arm tracking in Team VALOR's system. Whole-body motions account for approximately 5 cm of the end-effector pose error observed on hardware when an arm is at near full extension.



Humanoid Robot, ROS, Bifrost, Interprocess Communications, Robot Controller