Command and Control of Killer Robots

With the explosive growth in drone technologies over the past decade a growing debate has stirred regarding autonomous systems.

This past week a conference to discuss the role of lethal autonomous systems was held in Geneva. Many organizations, such as Human Rights Watch, believe that autonomous systems with lethal capabilities raise legal and ethical concerns. Fears about their proliferation, arms races between nations, and lower thresholds for violence have many nations examining their national policies on the matter. In their closing comments, the International Committee for Robot Arms Control (ICRAC) called for the following:

  • Develop national policies in favor of a ban on autonomous weapons systems,
  • Commit to moratoria on development, production and use, as recommended by UN Special Rapporteur Christof Heyns,
  • Support an international mandate leading to substantive negotiations on a preemptive prohibition, either by a CCW Sixth Protocol or other means, and
  • Ensure that discussions on autonomous weapons systems are open and inclusive especially of women, experts from the Global South and civil society.

However, it is doubtful that the U.S. will completely ban these systems as there is a strong use case for them – particularly in anti-access area-denial problems. Therefore, as military communicators – the mounting evidence that the U.S. military and many of its allies may choose to outsource its most dangerous missions to unmanned systems cannot be ignored. In a near future, providing reliable voice and data communication links to Soldiers, Sailors, Marines, and Airmen on the ground, on the open ocean, or in the air may no longer be our top priority. Getting communications to lethal autonomous robots and drones may be our main effort. Therefore, as the debate on the ethical use of these systems continues – we should not put off our planning.

So where to begin?

  • First, understanding what data and control measures must reside at the device – and what will reside away from these devices must be determined. Concerns over system disconnects, system losses or compromises must be taken into consideration as the command and control means for these systems are engineered. Once these decisions are made – better refinement of the bandwidth and redundancy requirements should emerge. However, its safe to assume that a dramatic increase in bandwidth requirements both in terms of line-of-site (LOS) links, but more likely beyond-line-of-site (BLOS) links, will emerge.
  • Second, a determination of where and when human control will be called for must be delineated. If the intent is to never have a “man” in the middle – and for these devices to be emplaced similar to the use of mines – then there may be no need for communications support. I suspect, however, that the vast majority of “autonomous” systems will go through a series of revisions in which human control is not only mandated by those authorizing the use of these systems but also desired so that human judgment and interaction can enhance the overall capability of these devices. Machines make us better and we make them better. What this means though, is that if operators must be located near the launch site of these devices that will mean one thing for system architects. If operators reside far from the battlefield that will mean another. Either way, now is the time to start considering the risks, costs, and challenges associated with both of these approaches.
  • Third, another important consideration is what autonomous system sensor data will be and where this data will ultimately be housed and processed. Likely, a large swath of data will be pulled from the battlefield and transported elsewhere to data centers or supercomputers performing intense analytics. Most AI like systems evolve their capabilities by learning from their environment and adapting their representative models of their worlds. Super computers are in the early stages of beginning to program themselves. I have no doubt engineers of these autonomous systems will seek the same form of learning for them. As such, we need to consider where these capabilities will reside and how they will be supported.
  • Fourth, the employment of armies of autonomous systems lends even more support to the cause of those working on the development of cognitive radios  and self-healing networks i.e. mobile ad hoc networks. Systems acting as their own communication relay points, capable of shuttling data through other autonomous systems within range until they can reach a gateway to greater bandwidth and primary data sources and controls may be essential. The efforts of those working on these capabilities for vehicles and dismounts could be adapted to this new use case.

So in summary – we need to start considering how much data we will be asked to push to and pull from these systems – we need to consider how new network architectures can be employed – and we need to start developing the backside infrastructure that will be called upon to handle the potential onslaught of data coming to and from these systems.

As always – the views in this piece are mine alone and do not represent the U.S. government, Department of Defense, United States Army, or any other organization with which I have had any association.

If you are interested in collaborating on this topic please let us know and we’ll get you on our distro.

There are no bad comments – please share your thoughts, insights, questions, etc.


Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s