The Army believes self-driving vehicles could be ideal during humanitarian relief missions in a natural disaster or for resupplying troops in the field, recognizing opportunities for cost savings and fewer crashes.
Autonomy-enabled systems will deploy as force multipliers at all echelons from the squad to the brigade combat teams. Future robotic technologies and unmanned ground systems (UGS) will augment Soldiers and increase unit capabilities, situational awareness, mobility, and speed of action. Artificial intelligence will enable the deployment of autonomous and semi-autonomous systems with the ability to learn. Decision aids will reduce the cognitive burden and help leaders make rapid decisions.
Driverless technology could reduce the risk of injury or death for convoys traveling through territory with hidden roadside bombs, said Bernard Thiesen, technical manager for Autonomous Mobility Applique Systems at TARDEC.
The self-driving car industry has made great autonomy advances, but mostly for well-structured and highly predictable environments. In complex militarily-relevant settings, robotic vehicles have not demonstrated operationally relevant speed and aren’t autonomously reliable. While vehicle platforms that can handle difficult terrain exist, their autonomy algorithms and software often can’t process and respond to changing situations well enough to maintain necessary speeds and keep up with soldiers on a mission.
“Unlike the commercial sector, we have to develop systems that can manoeuvre off-road, that can manoeuvre in all elements…. that can navigate obstacles, whether they be trees or gullies or rocks or whatever they may be,” US Secretary of the Army, Mark Esper said. Earlier, Michael Griffin, the undersecretary of defense for research and engineering had claimed that United States Army will have self-driving vehicles operating on the battlefield long before they’re on U.S. streets and highways, “But the core technologies will be the same.” “At a minimum, performance at par with a human driver should be achieved.”
DARPA launched Robotic Autonomy in Complex Environments with Resiliency (RACER) program in Oct 2020 with aim to make sure algorithms aren’t the limiting part of the system and that autonomous combat vehicles can meet or exceed soldier driving abilities. Over a four-year timeline, RACER will develop new algorithm technologies that maximize utilization of the sensor and mechanical limits of Unmanned Ground Vehicles (UGVs) and constantly test these algorithms in the field at DARPA-hosted experiments across the country on a variety of terrain. DARPA will provide advanced UGV platforms that research teams will use to develop autonomous software capabilities through repeated cycles of simulations and tests on unstructured off-road landscapes.
“In order to achieve RACER goals of increased speed and resilience, we need to embrace learning approaches that automatically tune system parameters in real time,” said Stuart Young, program manager leading the RACER project. “Successful software will extract features from sensor data and use that information to make on-the-spot driving decisions.”
RACER program goals include not only autonomy algorithms, but also creation of simulation environments that will support rapid advancement of self-driving capabilities for future UGVs. DARPA is slated to provide advanced UGV platforms research teams will use to develop autonomous software capabilities through repeated cycles of simulations and tests on unstructured off-road landscapes. Over a span of four years, RACER would develop new algorithm technologies maximizing utilization of the sensor and mechanical limits of UGVs while consistently testing the algorithms in the field at DARPA-hosted experiments nationwide on a variety of terrain.
The Army Reseach Lab has also been focusing on algorithms to advance UGVs. Researchers from ARL and the University of Texas at Austin are working on a suite of algorithms, libraries and software components that intelligent systems can use for navigation, planning, perception, control and reasoning when performing specific tasks. The goal is to teach ground robots to learn by doing, rather than responding to verbal commands, which will improve how autonomous systems move through rugged and unfamiliar terrain.
ARL’s Scalable, Adaptive and Resilient Autonomy program is looking to improve how autonomous ground systems travel through increasingly complex off-road environments. Software developed by participants will be integrated into testbed platforms and ARL’s autonomous systems software repository so it will be more broadly available.
DARPA’s Robotic Autonomy in Complex Environments with Resiliency (RACER) program is picking up speed after three teams were selected to go to the starting line in October 2021. The program is focused on advancing off-road autonomy of combat vehicles and seeks to demonstrate the ability of these platforms to travel at speeds that maintain pace with manned combat vehicles in complex terrain typical of that seen in real situations. Awards for Phase 1 were given to Carnegie Mellon University, NASA-Jet Propulsion Laboratory, and University of Washington.
The DARPA RACER program’s goal is to develop and demonstrate autonomy technologies that enable unmanned ground vehicles (UGVs) to maneuver in unstructured off-road terrain at the limit of the vehicle’s mechanical systems and at, or beyond, human-driven speeds and efficiencies. “RACER is intended to disruptively advance the integration and fielding of autonomy for robotic combat vehicles into the Army, Marine Corps, and Special Forces communities,” said Stuart Young, RACER program manager in DARPA’s Tactical Technology Office.
In November 2021 , the selected teams received the first of the DARPA-provided RACER Fleet Vehicles (RFVs) – a high performance all-terrain vehicle outfitted with world-class sensing and computational abilities – that they are using to develop platform-based autonomy for testing at upcoming DARPA-hosted field experiments.
RFV robots include 360o range and image sensing such as multiple LIDARs, stereo camera pairs, color and infrared imaging cameras, RADAR, event sensors, and inertial measurement sensing. Computation tools have multiple best-of-class graphical processing units (GPUs) in an environmentally protected, shock/vibration proof, and thermally managed Electronics Box (E-Box) that’s specifically engineered for the demands of the RACER high speed, off-road terrain expected in DARPA’s tests.
The sensor and E-box combination currently collects four terabytes of sensor data per hour to support artificial intelligence, machine learning-based autonomy algorithms and stack approaches required of fast-paced combat maneuvers in complex terrain. Modifications for roll protection, sensor/E-box integration, autonomous control, and increased 7kW of power are included in each RFV. The RFVs were integrated by Carnegie Robotics LLC (CRL) and are housed on a Polaris RZR S4 1000 Turbo base drive-by-wire platform.
Four RFVs have been completed, with three already delivered to RACER Phase 1 performers in November 2021. Four more are expected to be built prior to DARPA’s first RACER-hosted field experiment, scheduled for March of 2022 at the National Training Center in Ft. Irwin, California.
DARPA-hosted field experiments will provide the teams with a place to demonstrate the full capability of their autonomy stacks in complex environments. At Ft. Irwin, teams will demonstrate their ability to navigate courses with a variety of terrain and distances up to five kilometers.
To further support software development, DARPA has also collected over 100 terabytes of RFV-based sensor data from more than 500 kilometers of terrain in the Mid-Atlantic and West Coast. Shared with teams and managed within a RACER development tool for efficiency and security, this data will assist with learning approaches. Conceptual government baseline stacks and autonomy architectures will also be provided. They leverage recent products of Combat Capabilities Development Command-Army Research Laboratory initiatives in collaborative robotics, learning, and intelligent systems activities in partnership with basic research university and industry consortia.
The RACER program also has awarded two contracts to develop simulation environments and capabilities to enable development of off-road autonomy algorithms. “The RACER-SIM portion of the program looks to expand current simulation capabilities and physics-based models to support the testing of off-road autonomy in virtual environments,” said Young. “These simulation environments will allow teams to test and validate portions of their autonomy stacks without having to spend large amounts of time and money with field testing.”
The two primes for the RACER-SIM program are Duality Robotics and Intel-Federal.
RACER’s second round of off-road testing
The Defense Advanced Research Projects Agency’s (DARPA’s) Robotic Autonomy in Complex Environments with Resiliency (RACER) program has entered its next experimental phase. The program aims to produce autonomous off-road combat vehicles while traveling at speeds that keep pace with those driven by people in realistic situations.
The teams involved in the program have one experiment under their belts and will focus on even more difficult off-road landscapes at Camp Roberts, California, from September 15-27. Carnegie Mellon University, NASA’s Jet Propulsion Laboratory, and the University of Washington have each developed autonomous software stacks for the DARPA-provided robot systems tested in Experiment 1 earlier this year at Fort Irwin, California.
Experiment 1, executed March-April 2022, involved tests on six courses of combat-relevant terrain. The team completed more than 40 autonomous runs of about 2 miles (3.2 km) each and reached speeds just under 20 mph (32 km/h).
The terrain at Fort Irwin provided a number of obstacles, including rocks, bushes, ditches, etc., that were capable of severely damaging the robotic vehicles. The course also involved the desert environment designed to test the combat vehicles’ ability to identify, classify, and avoid obstacles at higher speeds.
The next series, Experiment 2, will require teams to go beyond the environmental features found in the desert environment to primarily test their perception algorithms on larger, steeper hills. This will also stress the robotic vehicle’s ability to maintain control, particularly going down steep slopes, on slippery surfaces, and navigating ditches over long distances than Experiment 1. The teams must also create longer-range plans amid driving through or around varied obstacles to successfully navigate courses.
Future RACER program activities include continuation into a RACER Phase 2 effort with a more representative 10T combat vehicle surrogate. The plan is to evolve in speed, scale, and mobility beyond the RFVs, as well as add a research track exploring tactics-based derivation of the new platforms
References and Resources also include: