In this report, RAND researchers explore current and potential military applications of autonomous systems, focusing especially on unmanned undersea vehicles and unmanned surface vehicles. The analysis centered on four areas: the current state of the art of autonomous technology, current kill chains and capabilities, future fleet architecture and its autonomous capabilities, and autonomy in alternative concepts of operation.
- What technologies are being developed to enable autonomy in naval vehicles?
- What capabilities currently are and are not possible?
- Where are there challenges, why do these challenges exist, and why are they difficult to overcome?
The U.S. Navy is interested in developing autonomous capabilities to execute tasks that are increasingly hazardous for humans and to enhance warfighting capabilities. In this report, RAND researchers explore current and potential military applications of autonomous systems, focusing especially on unmanned undersea vehicles and unmanned surface vehicles. The analysis centered on four areas: the current state of the art of autonomous technology, current kill chains and capabilities, future fleet architecture and its autonomous capabilities, and autonomy in alternative concepts of operation. The authors conclude that, although technological advances have occurred, autonomous systems capable of responding to unexpected changes in the environment do not yet exist and may not for some time. Any development of these capabilities will require substantial military investment because commercial systems are unlikely to meet the Navy's needs. The authors recommend that the Navy revisit assumptions about technological progress in autonomy, align the development of autonomy with the development of other capabilities that may be limiting factors, develop new concepts of operation to take advantage of autonomy's key characteristics, and reevaluate force requirements in light of the state of autonomous technology.
Autonomous systems for military applications are not on pace for the proposed future fleet, but some of the more promising uses of autonomy are not being exploited
- Advances in autonomy have been steady, but the transition to systems capable of reacting to unexpected changes in the environment has not occurred and might not occur for several years.
- The military applications contemplated for unmanned vehicles are unlikely to be developed without substantial military investment and development.
- The limiting features associated with the most-complex missions are not necessarily associated with autonomy, and it might not be useful to accelerate autonomy while such issues are still being worked out.
- Under current kill chains and concepts of operation, autonomy is generally employed to replicate items in the kill chain exactly as they are carried out by manned systems.
- The alignment between the expected future fleet and possible future uses of autonomy is not close, and author analyses suggest that some features desired for the future fleet are unlikely to be reached under the known program of record.
- Policy issues related to autonomous systems applying rules of engagement do exist, and it is unlikely that these systems will be engineered in a way that avoids these issues. Slowing the decisionmaking of autonomous systems by interposing a human in the loop will likely cede a critical time advantage in a high-intensity environment. This delay is a choice and cannot be mitigated by technical improvements.
- Some of the more promising uses of autonomy might be in using simple systems with limited autonomy for most functions — but adding the capacity for the systems to coordinate with each other.
- The Navy should revisit assumptions concerning technological progress in autonomy. The research indicates that the capability for autonomous systems to interpret context and make independent decisions, particularly in a dynamic environment, is not realistic in the short term.
- The Navy should align the development of autonomy with the development of other capabilities, such as power generation and storage, that might be limiting factors.
- The Navy should use the unique features of autonomy to enable new concepts of operation. It seems particularly promising to employ simple but numerous systems carrying different kinds of sensors.
- The Navy should accept the reality that autonomous systems will need to make engagement decisions if those decisions are to be effective. Modern weapon system timelines simply preclude human intervention.
- The Navy should develop a mechanism that allows humans to periodically assess whether an autonomous system is misinterpreting its environment.
- The Navy should critically evaluate the viability of complex multimission platforms and consider emphasizing simple but cooperating platforms.
Table of Contents
The State of the Art of Autonomous Technology
Autonomy in Existing Kill Chains
Future Fleet Architecture and Its Autonomous Capabilities
A Different Direction for Autonomous Systems
Conclusions and Recommendations
In-Depth Analysis: Algorithms
In-Depth Analysis: Payloads
In-Depth Analysis: Platforms
In-Depth Analysis: Patents