Indian Defence Review

Courtesy: Image source – https://politicalminefields.com/

After the advent of Artificial Intelligence (AI), few advanced nations are busy in developing and deploying an array of autonomous weapon systems to attain better speed and accuracy. All fatal missions are likely to be undertaken by unmanned systems in future. AI based Tactical Decision Support Systems (TDSS) could replace a commander in the war rooms. AI triggered smart weapons will replace the conventional warheads. 

Fully autonomous weapon systems are the ones which can process and exploit available data to arrive at decisions with the help of pre-programmed algorithms to make choices and act logically in unanticipated circumstances independently, more importantly, within the ethical framework rendering minimum collateral damage. The design and development of fully autonomous systems is a challenge to the leading military powers view underlying complexity in making it fail safe. Fully autonomous systems imply zero human intervention once let off. Advances in computational techniques and sensor technology accelerated the development of fully autonomous systems.

Accountability of collateral damage inflicted by fully autonomous weapon systems is a matter of contention. Machines can be made as scapegoats for all failed missions.Fully autonomous weapon systems are very effective in Narrow AI circumstances where the application is calibrated. Currently risk aversion has become a big barrier for enhancing the degree of autonomy of weapon systems.

Fully Autonomous Vs Supervised Autonomous Vs Semi-Autonomous Systems

Fully autonomous weapon systems are the ones when activated, can select and engage targets without any further intervention by the humans. They are less supervised autonomous weapon systems where the human operators cannot override operation of the weapon system. The system can select and engage targets without any human intervention after activation.  Whereas supervised autonomous weapon systems are the ones designed to intervene by human operators and even to terminate the engagement in case of system failure.  Semi-autonomous weapon systems are the one intended for specific target group. It has the provision to override the weapon delivery system post-delivery by humans.

Advantages of Fully Autonomous Weapon Systems

Only human beings show the indiscipline by deliberately deviating from the set rules and regulations which they are supposed to follow. Fully autonomous weapon systems are less emotional than human beings during combat or in a hostile environment. A robotic machine cannot hate or fear or be emotional or become tired. It has no survival instinct. The robot can delay the use of force for the most appropriate moment to destroy the target. They can sense and process more information simultaneously faster than human brains and hence it is wiser for humans to delegate many tasks to machines to avoid multitasking and information overload during wartime.

Fully autonomous systems are most appropriate for Narrow AI. Their requirements are more desirable for distinguishing between military and non-military assets and determining the legitimacy of targets. They are fool proof in comparison and can take better proportionality decisions than humans. They are adaptable to changing circumstances and are capable to face unanticipated circumstances from enemy side. These systems are controlled by preset algorithms catering all eventualities.Whenever a new situation arises, the system would switch over to the learning mode and the new learning would be used for future engagements.

Choice of Autonomous System

Majority of the AI based autonomous systems are currently semi-autonomous in nature. If these systems are rendering results to the satisfaction, there is no need to upgrade them to fully autonomous systems.There are certain applications where both manned systems and unmanned systems can team up as peers. The decision to migrate from semi-autonomous to fully autonomous system should be taken after detailed deliberation since the right to abort a mission is impossible.

Identifying Use Cases

Identifying use cases for fully autonomous system needs detailed deliberations and input from various domain experts. Fully autonomous systems found to be ideal for defending airspace, land borders and coastal lines. But developing these systems exclusively for penetrating enemy defence needs extensive research and studies. Meeting military objectives by giving immunity to civilian population is most important factor being considered while shortlisting use cases for fully autonomous military systems.

Challenges in Design and Development

•  Irreversible System

The very fact that we cannot recall a fully autonomous system post launch makes it more challenging for the weapon developers.  Since these weapons are irreversible, the reach of fully autonomous systems should be calibrated within the frame work of safety of mankind and magnitude of human casuality. Owning up responsibility for malfunction, accidents and collateral damages by the nation who launched the weapon is a big challenge.Precautions are mandatory to avoid civilian casualities

•  Limitations of Algorithms

The algorithms used for high resolution image analysis, information storage and decision making are very complex and resource intensive. The system should have adequate redundant channels to boost the reliability of system. The quantum of algorithms needs to be executed consecutively is limited by the computational capability and system memory.

•  Limitations of Sensors

The accuracy of algorithms being executed is depended on the quality of weapon sensors. Weapon sensors includes Optical Camera, LASER, GPS, Infrared and Light Detection and Ranging (LIDAR). The size of sensors limits the capability of autonomous systems. Adding better sensor could make the system more bulkier and hence optimization of size and volume could become a trade-off. Moreover, high resolution sensors warrants large memory capacity and better computational powers.

•  Hardware Limitations

High resolution terrain mapping of autonomous systems is resource intensive. Same is applicable for the underwater sonic mapping as well as for airspace surveillance and mapping. Processing of images is throttled by the capacity of system memory and speed of Graphic Processor Unit (GPU). Hardware capability limits the speed of autonomous system and the level of automation. Only few manufacturers exists in the world who can produce the ‘mil std’ hardware with high processing capacity.

•  Band width of Data Link

Limitations of band width of wireless communication network is another choking point for data link between weapons and controlling platforms. Band width capacity limits the speed of system and the level of automation. Military uses 30 to 88 Mhz band for close air support and associated data links. Limitations in communication hardware limits the band width to a large extend.

•  Availability of Domain Experts

Technologies associated with fully autonomous systems are robotics, optics, artificial intelligence, machine learning, deep learning, validation, quality assurance, sensors calibration, cyber security and material science. Availability of domain experts in each field under one roof is a big challenge for weapon manufacturers.

•  Harsh Battlefield Environment

The process of terrain mapping of land, under water and airspace is an exhaustive resource intensive exercise. Unpredictability of terrains and moving weapon platforms of enemy, adds to the complexity of algorithms for fully autonomous platforms. Making them weather proof and adding more redundant channels are mandatory to avoid mission failures. Autonomous platforms are designed to overcome unprecedented obstructions in transit during missions.

•  Interference from Other Systems

Cross Interference of sensor signals is another big challenge for fully autonomous systems. Currently lasers and radars are used for navigation of autonomous platforms. Spurious Signals from external systems can always play spoilsport and implications could be disastrous. There can be deliberate attempts by the enemy in some form of Electronic Counter Measures (ECM) to misguide the autonomous systems.

•  Time Constraints

Time plays a major role in autonomous system efficiency. Algorithms are capable of successful planning and re-planning of mission activities within the time frame. Interpretation of onboard and offboard sensor information should happen in time for speedy decision-making process. Effective command-and-control squarely depends on timely processing of mission data and sharing them to stake holders.

•  Testing Stage

Testing is one of the most difficult stage in fully autonomous weapon system development. The process is time consuming and the designers may have to go back to the design board many times. Wrong choice of use cases and the flaws of design can have serious ramifications in testing stage. The testing and trial process is humongous since there is a need to mobilize massive military resources.

Conclusion

Protest by few nations against autonomous weapon systems at UN is a big challenge for fully autonomous weapon developers since the stakes are high. The fear of miscarriage is the biggest barrier in the development of fully autonomous systems. Nevertheless, days are not far where we humans will team up with machines as peers for military operations