In the recently held symposium on AI for Air Warriors in New Delhi, the Indian Air Force's (IAF) Air Chief Marshal RKS Bhadauria's statement reflected the high level of professional approach being taken by IAF in implementing the Artificial Intelligence (AI) concepts for the Air Force.
By Milind Kulshreshtha,
In the recently held symposium on AI for Air Warriors in New Delhi, the Indian Air Force’s (IAF) Air Chief Marshal RKS Bhadauria’s statement reflected the high level of professional approach being taken by IAF in implementing the Artificial Intelligence (AI) concepts for the Air Force. The other contributing expert speakers from India and abroad highlighted various AI solutions being designed and developed in the context of Air Combat operations.
The e-symposium was organized by Society for Aerospace, Maritime and Defence Studies (SAMDeS) in collaboration with the industry body FICCI.
AI for IAF Missions
The information received from heterogeneous sources is fused to enhance the detection capability and early identification of target(s). Multi-platform multi-sensor data fusion is the key technique for achieving information superiority over the adversary. For any effective and efficient AI application, an AI based Decision Support Systems (DSS) architecture needs to emerge for the complex Air combat operation environment. It emerges that the latest generation fighter jets like F-35 are predominantly software-centric (up to 90%) for target detection, target categorization, tracking and target engagement activities. In today’s Air Combat scenarios, it is humanly impossible for a pilot to process the enormous amount of high-speed data being generated by multiple sensors. This kind of activity can only be undertaken by high-end processors manufactured for hard Real-time architecture and running on RTOS (Real Time Operating System).
The OODA (Observe, Orient, Decide, Act) loop is the fundamental basis of any C4I (Command, Control, Communication, Computers and Intelligence) and it is feasible to implement AI at various points in this concept. To enable the C4I process, creation of a Fusion Intelligence platform by using data available from Space based sensors; airborne units, Weather sensors, Surface sensors etc. would be the first step. For these applications it is possible to exploit the mature and well proven commercial AI algorithms, however, complexity of training the AI individually for each use-case is an arduous and time consuming activity. Further, acquiring the correct data-set for training and testing itself is a challenge. The difficulty likely to be faced in molding such commercial solutions to a MIL Grade standard and using AI in a SaaS (Software as a Service) model is yet to be thoroughly researched.
The man-in-the-loop emerged as a necessary criterion for any AI application in combat due to multiple factors, especially while AI is still far from achieving accuracy in threat detection. For example, the image processing algorithms for AI to identify a threat are still not completely developed since the supporting technologies are yet to be innovated. The trust in the AI system is still a long way to leave AI for autonomously engaging a hostile target. Even the present generation C4I systems are designed for AUTO/SEMI-AUTO/MANUAL (viz. Autonomous, Semi-autonomous, Manual) modes and have a human element as a part of the process. This Human element intervention can be in the form of pushing the trigger for FIRE (i.e. weapon shooting), Target selection or FIRE Authorization. Hence, the C4I systems with such ‘man-in-the-loop’ limit the systems to be designated as Soft-real Time instead of Hard-real Time.
AI in Communication Systems
The net-centric tactical ISR information combined with the joint operations in a combat mission requires information collection and transmission amongst the net units (like Satellites, Air Electronic Warfare, AWACS etc.). The moving of Real-time information across multiple systems in the loop always diminishes the `real-time’ quotient within the information, making the data stale for use. Here, AI-Driven multi-access networking and Edge computing architecture emerges as an ideal communication solution. Free-space optical (FSO) communication, 5G and Satcom channels of communication can thus be optimally utilized for achieving flexible and assured bandwidth. Further, the present AI system architecture necessitates bringing the data to the AI engine for processing, however, are being made to bring AI closer to the data end using edge technologies.
AI solutions for UAVs
The research and development work pertaining to the AI application demonstrated the enhanced combat capabilities which UAVs shall be bringing in the Air combat operations. AI in UAVs can be said to be the natural extrapolation making the drones truly autonomous. These Air launched UAVs are today capable of stand-off imaging and extended range communication. UAVs are expected to improve the decision support capabilities on the edge, thus making the DSS systems more efficient and effective.
Predictive Maintenance for Air Assets
The use of AI for predictive maintenance is an already evolved field commercially, and is under various stages of implementation by Multinational firms. It is well proven that AI based predictions maximize efficiency, reduce unplanned downtime and increase equipment reliability. Coupled with a maintenance scheduler application, it provides an ability to manage, schedule and execute maintenance programs for thousands of machines and help a user to manage the full asset lifecycle to finally help in an intelligent strategic planning. It is possible to provide timely alerts via alarms, triggers for email, onscreen or SMS notifications to prompt for action by the user. Aircrafts have a well defined, structured and a strict maintenance schedule to keep these air assets fighting ready at all times. The Ops Logistic Concept can be effectively implemented using such AI based predictive maintenance techniques.
Ethics of AI in Combat Role
The AI systems are designated as ‘true AI’ when they have the intrinsic capability to carry out ‘own’ measurements as per the operational circumstances, continuously train itself and successfully execute the combat mission. The development of a self training Deep Learning Neural Network like this is technically feasible. However, achieving this level of ultimate AI based killing machine raises multiple ethics issues, especially when it comes to the human conscience. It brings back the memory of Oppenheimer’s regret on successful deployment of Project Manhattan to finally conclude World War II.
There is a definitive application of AI in the Air Combat role and it is a matter of time that IAF too shall be looking at inducting AI within its war-fighting capabilities. The need for unbiased data to train and test such combat systems shall be one of the biggest challenges for IAF. Many security aspects like smart cloud servers available in India independently for providing the data confidentiality and cyber security in the support infrastructure too needs to be addressed. In the near future, AI solutions in Air Combat and predictive maintenance shall change the IAF SOPs. However, the contradiction here is whether the existing conventional Defence Acquisition Procedure (DAP) are adequate to allow AI based solutions to be inducted, or whether these procurement procedures too are going to be changed by AI for good.
(The author is a C4I expert who has worked on the development of the indigenous Naval Combat Management (CMS) systems. He has a keen interest in Joint Warfare C4I solutions pertaining to Threat Assessment & Warfare modules, Multiplatform Multi-sensor Data Fusion (MPMSDF). Email: firstname.lastname@example.org Views expressed are personal and do not reflect the official position or policy of Financial Express Online.)