A high level of autonomy is desired for future unmanned combat systems because lethality and survivability can be improved with much less communication bandwidth than would be necessary for preprogrammed or remotely operated systems. However, there are a number of technical challenges that must be addressed prior to implementation.
Air Combat Maneuvering (ACM) has been described as the art of maneuvering a combat aircraft in order to obtain a position from which an attack can be made on another aircraft. During ACM, pilots use their knowledge of maneuvering strategies and tactics to determine the best course of action. However, there are many factors to consider (such as relative airspeed, altitude, heading/position, energy, and maneuvering/weapons capabilities), and there are numerous different tactics that can be used to gain an advantage over an opponent.
Most air combat simulations use heuristically-based artificial intelligence (AI) methods to control enemy aircraft. These rule-based approaches have the advantage of being very fast, and have reached a point where they can perform very close to how a textbook flying human would react under a similar situation. However, these gaming environments typically involve a single static tactic within a controlled environment. Other optimization methods have the advantage of solving for an optimal solution; however, these methods typically require substantial computational time and therefore are difficult to use for time-critical applications. Since future air combat missions will involve both manned and unmanned aircraft, the primary motivation for this research is to enable unmanned aircraft with intelligent maneuvering capabilities.
An artificial immune system approach is used to select and construct air combat maneuvers. These maneuvers are composed of autopilot mode and target commands, which represent the low-level building blocks of the parameterized system. The resulting command sequences are sent to a tactical autopilot system that has been enhanced with additional modes and an aggressiveness factor for enabling high-performance maneuvers.
The autonomous ACM system constructs motion-based plans, in the form of maneuver sequences that are composed of one or more autopilot commands, along with the scheduling times for command execution. Each autopilot command consists of a mode identifier and corresponding target. The maneuver selection system contains autopilot mode-dependent performance models for predicting the motion-based path of maneuver sequences. These maneuver sequences are constructed from basic piloting maneuvers, which are stored in a maneuver database. Artificial immune algorithms are used to select the appropriate maneuvers from the database, and to augment them as necessary in order to achieve tactical objectives. Once these maneuver sequences are generated, they are sent to a specialized autopilot system for execution.
These maneuvers can in turn be combined with other maneuvers to form more complex maneuver sequences. The collection of these maneuvers is stored in a maneuver database, and represents a combination of randomly created and/or manually constructed maneuvers, as well as maneuvers that are generated through immunized maneuver selection.
The Artificial Immune System (AIS) combines a priori knowledge with the adapting capabilities of a biological immune system to provide a powerful alternative to currently available techniques for pattern recognition, learning, and optimization.