ABERDEEN PROVING GROUND, Md. -- Army researchers recently expanded their research area for robotics to a site just north of Baltimore.
The U.S Army Combat Capabilities Development Command, known as DEVCOM, Army Research Laboratory [link.mediaoutreach.meltwater.com] now uses land on the Graces Quarters peninsula, as part of the ARL Robotics Research Collaborative Campus, known as R2C2, for experimentation and research operations.
The laboratory uses about 200 acres of the approximately 700-acre site belonging to APG in Middle River, Maryland. Almost a century ago, this area was used to test munitions and armaments. Today, it’s a dedicated, reconfigurable infrastructure that supports scalable, multi-domain battlefield operations accessible to extended research communities focusing primarily on autonomy, artificial intelligence and robotics.
“The one-of-its-kind research campus was established to advance Army knowledge of autonomy and intelligent systems through basic and applied research of unmanned technologies that integrate artificial intelligence, autonomy, robotics and human teaming elements in complex environments,” said Jeffrey Westrich, program manager, DEVCOM ARL R2C2 initiative.
The laboratory is focused on performing experiments at the military operations in urban terrain, or MOUT site, which was completed in November, as well as the natural environment that comprises the majority of the R2C2 site, Westrich said
Earlier this year, Army researchers performed the first fully-autonomous tests onsite using an unmanned ground vehicle testbed platform, which serves as the standard baseline configuration for multiple programmatic efforts within the laboratory. As a means to transition from simulation-based testing, the primary purpose of this test event was to capture relevant data in a live, operationally-relevant environment.
Westrich said the tests served to preliminarily prove performance of the ARL Autonomy Stack for future, extended field testing. The ARL Autonomy Stack is the software framework and collection of algorithms that define the intelligence of the system, and manifest its ability to perform operationally relevant tasks like localization, planning, sensory data analysis, mission-specific behaviors, and communication.
The tests captured sensory data that depicts actual field conditions, including elements that present challenging problems for robots operating in natural, unstructured environments. Challenging natural elements include frequent and erratic terrain changes, fallen branches, large rocks, loose biomass, dense shrubbery, bodies of water and other natural debris and growth.
“These data sets can be offloaded from the platform following testing for labeling, analysis, and incorporation into machine learning applications that iteratively improve on the ground platform’s ability to navigate environments like this in the future,” Westrich said.
Before recent field testing, Army researchers heavily relied on computer-based modeling and simulation to perform qualitative checks on functionality of the ARL Autonomy Stack.
“Simulation-based work was most prevalent throughout the past year because of COVID-19; however, reliance on simulation alone presents its own challenges,” Westrich said. “Due to the nature of its intended use as a qualitative means to verify functionality, the simulated environment presents only a rough approximation of physical elements. Consequently, the accuracy and fidelity of those modeled components influence the accuracy and robustness of autonomy algorithms–which can drastically change how the robot identifies, recognizes, or responds to stimuli in the environment.”
The recent and upcoming tests at the MOUT site and in the surrounding natural environment, will help the Army identify how the platform responds in a real setting, he said.
“We can utilize those results as a comparative metric for improving simulation, and informing research and development through an iterative improvement approach,” Westrich said. “This location promises a substantial role in accelerating our understanding of robotics research, and pursuing significantly more complex experimentation in the future.”
Army researchers have integrated algorithms from recipients of the first Scalable, Adaptive, and Resilient Autonomy program sprint into the ARL Autonomy Stack, with experiments ongoing at R2C2 through the spring. Outcomes from these extramural efforts feed directly into the laboratory’s Artificial Intelligence for Mobility and Maneuver essential research program, as added capability and improved functionality for the laboratory’s robotic systems.
SARA is focused on developing and experimentally accelerating emerging research in autonomous mobility and maneuverability, scalable heterogeneous and collaborative behaviors, and human agent teaming to realize adaptive and resilient intelligent systems that can reason about the environment, work in distributed and collaborative heterogeneous teams, and make op-tempo decisions to enable autonomous maneuver in complex and contested environments.
