Despite the challenges posed by Covid-19, the U.S. DARPA continues to work on advancing technologies and new programs. Let's look at some of the most recent programs launched in 2020...
Despite the challenges posed by Covid-19, including remote work, facility closures, and other restrictions to keep people safe, the Defense Advanced Research Projects Agency (DARPA) continues to work on advancing technologies. In addition to the agency’s Biological Technologies Office working on solutions to address Covid-19, DARPA has several new programs underway that address new technologies. These range from artificial intelligence for simulated war games and advanced technology for satellite constellation autonomy and space mesh networks to quantum information processing, optical interconnects, and passive 3D sensors for navigation.
Here are some of the most recent programs launched in 2020.
DARPA recently launched the Gamebreaker AI Exploration program to test AI methods that create game imbalance in video games for application in simulated Department of Defense (DoD) war games. The Gamebreaker program, comprised of nine teams, is tasked with applying AI to existing video games to assess game balance and find new capabilities, tactics, and rule modifications to destabilize the game.
DARPA said there are many advanced commercial video games that involve realistic command and control, campaign planning, and strategy development that are used in the military. The idea is to leverage the gaming industry’s AI developments, instead of starting from scratch, and use them for the DoD.
By figuring out a generic method to assess and manipulate balance in commercial video games, the DoD could then apply those AI algorithms to create imbalance in DoD-simulated war games used to train warfighters, said DARPA.
High-speed autonomous networks in LEO
DARPA’s Blackjack program, in partnership with the U.S. Space Force and Space Development Agency, is aimed at demonstrating advanced technology for satellite constellation autonomy and space mesh networks. The program will develop and validate critical elements of global high-speed autonomous networks in low-Earth orbit (LEO). The demonstration flights (later in 2020 and 2021) are rideshares, launching with other missions.
The first demo, Mandrake 1, is a cubesat that will carry supercomputer processing chips, followed by Mandrake 2, a pair of small satellites that will carry optical inter-satellite links for broadband data. These technologies could be used as the foundation for future optically meshed computer networks in LEO.
The program is also evaluating a risk reduction payload called Wildcard, a software-defined radio that will experiment with links from LEO to tactical radios. This includes a data-fusion experiment with the ability to host advanced third-party algorithms for an upcoming Loft Orbital mission.
Focusing on buses, payloads, and an autonomous mission management system, called Pit Boss, the aim of the program is to demonstrate that sensors, low in size, weight, and power, can be mass-produced to fit on different buses from different providers for less than $2 million per payload.
The program will run simulations to test payloads in virtual constellations of all mission types to show interoperability between the commoditized buses and the various payloads being considered. DARPA expects to have the actual hardware by late next spring and by the summer work on satellite-level qualification for launch readiness in late 2021.
DARPA launched the first phase of the Optimization with Noisy Intermediate-Scale Quantum devices (ONISQ) program with the selection of seven university and industry teams. Phase 1, which started in March, runs for 18 months. Exploring quantum information processing, the program is looking at a hybrid concept that combines intermediate-sized quantum devices (hundreds to thousands of quantum bits, or qubits) with classical computing systems to solve challenging problems called combinatorial optimization.
If the program can demonstrate a quantitative advantage of quantum information processing by advancing beyond the performance of classical-only systems in solving optimization challenges, it could be used by the defense and commercial industries in areas such as global logistics management, electronics manufacturing, and protein folding.
Real-time airspace awareness for future battles
Under the Air Space Total Awareness for Rapid Tactical Execution (ASTARTE) program, in partnership with the Army and Air Force, DARPA will evaluate the use of advanced low-cost sensors, AI algorithms, and a virtual laboratory testbed to enable a common operational picture of the airspace above future battlefields. The goal is to allow for long-range fire missions as well as manned and unmanned aircraft operations to occur simultaneously and more safely in the same airspace, said DARPA.
DARPA said that this is critical for implementing its Mosaic Warfare concept for seamless coordination across aerial, ground, and sea nodes. ASTARTE will provide a real-time, 4D (space and time) moving picture of the battlespace for friendly forces and use its sensor network to detect and map adversary locations. The new ASTARTE “engine” or “brain” will be designed for compatibility with existing and future command and control systems (C2) used by the military services.
The three technical areas include developing algorithms for understanding and decision making that can predict airspace usage conflicts, propose de-confliction solutions with associated risk levels, and direct sensors to maintain the airspace picture at any given moment in time. The program will develop or leverage existing low-cost sensors to detect and track, in real time, manned and unmanned aircraft, airborne weapons, and other potential flight safety hazards. The third technical area is the development of a virtual laboratory testbed for modeling, simulation, and virtual experimentation using a combination of current C2 systems and ASTARTE technology.
Under DARPA’s Photonics in the Package for Extreme Scalability (PIPES) program, researchers from Intel and Ayar Labs have demonstrated early progress in improving chip connectivity with photons, or light, by replacing traditional electronic input/output (I/O) with optical signaling interfaces. The program is looking at ways to expand the use of optical components to address the limitations of electrical signaling from the chip package, which impacts overall bandwidth and signaling efficiency.
DARPA reported that the researchers have achieved major improvements in link reach and efficiency to enable higher-performance digital microelectronics. The researchers have replaced the traditional electrical I/O of a state-of-the-art FPGA with efficient optical signaling interfaces by leveraging an optical interface developed by Ayar Labs called TeraPHY, an optical I/O chiplet that replaces electrical serializer/deserializer (SerDes) chiplets and Intel’s advanced packaging and interconnect technology.
The team integrated the TeraPHY chiplet, capable of 2 terabits per second of I/O bandwidth, and the Intel FPGA in a single package, creating a multi-chip module (MCM) with in-package optics. “The integrated solution substantially improves interconnect reach, efficiency, and latency — enabling high-speed data links with single-mode optical fibers coming directly from the FPGA,” reported DARPA. It was built on GlobalFoundries’ advanced photonics process.
The researchers also used technical advances achieved under two other DARPA programs — the Photonically Optimized Embedded Microprocessors (POEM) and Common Heterogeneous Integration and IP Reuse Strategies (CHIPS) programs — and low-power signaling standards and chiplet packaging processes developed by Intel under the DARPA CHIPS program.
An optically connected FPGA board developed by Intel and Ayar Labs (Image: Ayar Labs)
Fully homomorphic encryption
DARPA developed the Data Protection in Virtual Environments (DPRIVE) program to design and implement a hardware accelerator for fully homomorphic encryption (FHE) computations to significantly reduce the current computational burden in order to speed FHE calculations. The goal is to reduce the computational runtime overhead by many orders of magnitude compared to current software-based FHE computations on conventional CPUs, as well as accelerate FHE calculations to within one order of magnitude of current performance on unencrypted data, reported DARPA.
Conventional data-encryption methods or cryptographic solutions, such as Advanced Encryption Standards (AES), protect data as it is transmitted across a network or at rest while in storage, said DARPA. However, processing or computing on this data requires that it is first decrypted, which exposes it to numerous vulnerabilities and threats, the agency added.
FHE provides an alternative solution that enables computation on encrypted data, or ciphertext, rather than plaintext, or unencrypted data, which protects data at all times. The key challenge is that FHE requires enormous computation time to perform even simple operations, making it impractical to implement with traditional processing hardware, said DARPA.
Under previous programs, DARPA developed FHE algorithms that showed what was possible with FHE running on standard CPUs, as well as the compute penalty and limitations of the technology. Under the new program, DARPA said it is re-architecting the hardware, software, and algorithms to make it a practical and usable solution. The goal is to develop a hardware accelerator that can process large arithmetic word size (LAWS) without word size reduction and overhead, natively processing on LAWS of 1,024 bits or more.
Passive 3D sensors
DARPA’s Invisible Headlights program addresses the need for illumination for autonomous and semi-autonomous systems to navigate at night or underground. However, switching on visible headlights or other systems like LiDAR will allow adversaries to detect a vehicle’s presence. DARPA said the goal is to create new passive 3D sensors and algorithms using the information captured from thermal radiation, as everything gives off some thermal energy, and turning it into a 3D scene for navigation.
The three phases of the program include determining if thermal emissions contain sufficient information to enable autonomous driving at night or underground; refining models, experimental designs, and ensuring system feasibility for achieving 3D vision at both low speeds (<25 mph) and high speeds (>25 mph); and then building and testing passive demonstration systems that compete with active sensors.
DARPA is also working on a variety of other technologies and challenges, including security challenges in 5G and future wireless networks by leveraging open-source software and systems, the development of formal safety assurances for autonomous systems, and protecting wideband RF systems from interference and jamming in electromagnetic environments.