The first of a two-parter, one for each side of the argument. This is an overtly critical piece challenging the fervent argument for combat UGVs. It's meant to be confrontational. The next part will argue the benefits, so any UGV fanatics out there, please don't get too upset, but please do contribute to the discussion.
UGVs are the future of ground warfare, so they say. Soon we will all see combat UGVs (often termed Robotic Combat Vehicles or RCVs, which I'll use from here to differentiate from the much broader UGV domain) prowling ahead of the FLOT, engaging ahead of human risk. Even when manned vehicles move forward, they will do so with 'unmanned wingmen' alongside, providing additional firepower and autonomous sensor fusion. UGVs represent a game-changing evolution of warfare.
Do they though? The technological hurdles to the credible real-world employment of a combat UGV are enormous, and a long way from being solved. The resource cost to solve them is massive, and even if they are, further problems remain. In many cases the RCV proposition offers no tangible benefit over manned alternatives, and sometimes might make things worse.
To be clear, autonomy is absolutely the future for land platforms and a key enabler of the next generation of AFV. Manned unmanned teaming (MUM-T) and other applications of RCVs are a compelling vision and should continue to see committed and proactive R&D to further their maturity. This post is something of a devil's advocate piece raising a discussion around some of the more significant obstacles to the fielding of RCVs, some of which really don't seem to be getting the recognition they need.
In no particular order, then, some major objections to the RCV argument:
Autonomous Navigation: A colossal cognitive obstacle
Autonomous navigation is an absolutely critical element of RCVs becoming a viable proposition. Yet it seems to be dismissed as a simple concept, for example I saw a piece on the current UGV landscape recently that opened with a remarkably reductive understatement that driving from point A to point B is not difficult for a UGV.
Autonomous navigation within a land environment is colossally more challenging than anything in the air or sea domains, owing to the unhelpful nature of the ground being uneven and covered in things. If a UAV or a USV need to get from A to B, it will normally be in a totally clear space. An RCV, by contrast, will need to navigate through terrain that is hugely complex and fraught with automotive and tactical disaster.
Purely from an automotive perspective, is the ground hard or soft, dry, or wet? Are there water obstacles, of what depth, of what velocity? Are there pits, ditches, or trenches, and are they covered or visible? Are there mines, IEDs, obstacles, defensive positions? What is behind that rock, house or hill? Objects and obstacles may screen the terrain behind so as to render route finding challenging and highly contextual.
Take the simplest task in this domain - driving down a road. Most UGV using a combination of GPS/INS to navigate known routes, with optical and LIDAR type systems scanning and mapping the path ahead. However, something as simple as a cardboard box in the road may appear to the system the same as a concrete post, and be impassable - inferring object material composition visually is a very complex cognitive task that is challenging for a machine. That said, technologies are capable to making this distinction (though a creative opponent could make lightweight dummy obstacles that are designed to spoof RCVs, but that is a different challenge). Even if it is classified as a box as the route determined to be viable, there may be more context that needs to be established.
Is that a suspicious place for a box to be? Is there anything about the box that makes it of note? Placement, shape, adjacent objects, bystanders or lack thereof, its contextual existence in the first place.
All of these are relatively simple concepts for a human mind to factor into its determination that the box is safe, or is an IED or obscuring some other risk to the platform, but these sorts of cognitive calculations are enormously difficult for a computer.
The simple 'cardboard box problem' is something that demonstrates the troubles of asking a computer to undertake such inherently human and contextual actions as driving a vehicle in complex environments.
Say the system is working well, and a clear path is determined. The vehicle comes to a point and has an open plain to cross. How does it determine the best route? The system needs not just to be capable of scanning and avoiding obstacles in the immediate vicinity, but to plan ahead. The best route at the present moment may overtly lead to an impassable position on the far side of the terrain. A human driver may see that one side is overtly more readily transited than another based on an assessment of the distant terrain, but can an RCV make this determination? Can an RCV see the texture and type of terrain a mile or more away in all directions and make snap decisions from this information?
This is all before layering in consideration of the best tactical movement of a vehicle for a combat engagement in the terrain ahead, where advantageous placement may warrant a compromise in route selection from a purely automotive perspective. Is one option inherently more favourable to an ambushing force than another and worth the greater automotive challenge?
We should be under no illusion of the scale of this hurdle. Driverless passenger cars have been in very proactive and heavily funded development by the world's most talented technology companies for decades now, and whilst they are coming on well, there are still plenty of examples of autonomously driving cars going the wrong way or slinging themselves into ditches. Remember, this is on paved roads with high fidelity mapping data where the main challenge is safely navigating traffic. The challenge even in that domain is so great that industry are resetting significant areas of development and adjusting public expectations around timelines and capabilities.
Navigation off road in uncertain terrain at pace within a combat environment is therefore many, many orders of magnitude more challenging. So, unless we consider a Great War era paradigm of AFVs trundling cautiously at 4 mph and then getting stuck in the first unsighted puddle they encounter as a bold new future, there is some considerable distance to go yet.
Assured Communications: An Achilles heel with no immediate solution
Like all remotely operated systems, RCV have assured communications as an absolutely critical capability to be remotely viable. Unless the command link from the control station, typically envisaged to be a partner vehicle moving within the area of operations near to the RCV itself, can be assured, the notion of an RCV becomes very dubious.
The issues of assured communications in a highly digitised contemporary battlespace is by no means isolated to notional RCVs, however it can impact them the most severely. Where manned platforms can degrade to individual vehicle or soldier level operation with drills and procedures for such degraded use, RCVs are only able to operate in any capacity whilst the command link remains open and stable.
An immediate concern is around range and interruption of signal by natural obstructions. Contemporary requirements for RCV are seeking relatively short range - typically 1.5 km to a maximum of 4 km and generally within line of sight of the control platform. The broad approach is that an RCV should remain within range of the control vehicle's primary weapon for mutual support. As such, they are more unmanned wingmen than the more visionary RCV future that is desired, yet even within these very modest requirements there are challenges of connectivity when momentary terrain or buildings screen the RCV and potentially degrade or block the command link.
The issue is not just one of obstruction of the signal by natural means, however. Active interruption is not just possible but highly likely, with peer opponents such as Russia fielding very capable electronic warfare (EW) capabilities.
"Russia’s growing technological advances in EW will allow its forces to jam, disrupt and interfere ... negating advantages conferred on [NATO] by its technological edge" International Centre for Defence and Security
Users envisaging the fielding of RCVs need to have robust plans for how these systems will operate in a contested electronic environment and with active countermeasures being deployed against them, and part of this needs to be a serious discussion of the course of action when a connection is lost with the command platform.
So, what should the RCV do when the link is lost? The broad options to choose from are:
Retrace steps to last point where there was a connection.
Continue following last command until a connection is re-established, stopping at limit of that command.
Go fully autonomous on programmed parameters until a connection is re-established, becoming a fully autonomous combat system.
Stop and wait for a reconnection, relying on the command vehicle to find them.
None of these are good courses of action. At best they pause movement and risk losing the initiative, or worse, being geolocated in a static position and rapidly destroyed. With the option (3) to go fully autonomous and behave in a human-like manner of deferred command authority firmly off the table for the foreseeable future, which of the bad options does an RCV follow?
The US Army, representing the leading edge of thinking and funding into RCVs, foresees that they will at least for the time being pursue option a combination of (2) and (4), "Degrade then auto-stop if the vehicle loses comms w/ the operator."
In the envisaged robotic warfare battlefield, this potentially sees many opportunities for an opponent to stall manoeuvre and pin assets in place for rapid destruction, and exacerbates the extant risk of signals interference to a point where it doesn't just offer the ability to degrade a force's capability, but utterly nullify it.
Taking people out of harm's way: or putting them further into it?
One element of the argument is that RCV platforms mean less personnel in harms way, but again this is something of a false narrative. Whilst a hypothetical future state where RCVs deploy and fight alone is theoretically great, as I've outlined, the reality is far from this.
For at least the foreseeable future, RCV will need a human in the loop - the moral discussion around fully autonomous fighting vehicles is a whole separate topic for another day. The vehicles can conduct very basic route-finding (see above) but more will need teleoperation for actual combat manoeuvre. Similarly there is very low appetite for closed loop automated target identification and engagement, so a human will certainly remain in the loop for lethal force.
"RAS and AI also show great promise in reducing physical risks to soldiers and Marines." US Congressional Research Service
Unlike UAVs, RCVs will remain in the area of operations for the duration of a conflict, and so any support personnel will need to be broadly collocated with them. AFV, including RCVs, need constant daily support and the vast majority of it cannot be automated, so there is an inherent inevitability to humans being in proximity to RCV platforms. However, say RCV proponents, these personnel are fewer in number than the manned alternative, and in a safer location as they are not in the vehicles while they are fighting.
That's not quite true though. Take a practical example in the form of the US RCV programme, specifically the RCV-M. RCV-M has notionally been presented as being controlled from the MET-D platform, which would be a command and control vehicle locally positioned (potentially no more than 1.5 to 3 km away) to control the RCVs connected to it.
At present, that is envisaged to be two operators per RCV (driver/gunner), with two RCV controlled by each MET-D with a supervisor for the overall fire unit. These personnel are in an AFV of their own (the MET-D) which at a minimum will require a driver and commander.
Assuming a 1:1 replacement of M2/M3 Bradley with RCV-M, a platoon has gone from 4 vehicles with a crew of 12, to 6 vehicles with a crew of 12-14. The personnel are more densely clustered into just two platforms rather than 4, increasing risk not least as they are a more static and far more readily identifiable electronic node for attack. The crew also now have 50% more vehicles to maintain in the field without a net increase in manpower, and there are now two distinct vehicle types to support rather than one.
"Signature management will be a challenge." US Army NGCV Cross Functional Team
So what was the benefit? Well, soldiers are not so directly in harm's way now as they are not in the combat platform, but sat a (short) distance back in the MET-D command vehicle. But we have consolidated that crew from 4 to 2 vehicles, so they are now more densely located and in addition more readily identified as a priority target, as their constant high bandwidth emissions will clearly indicate them as an RCV command node to be prioritised for targeting.
The electronic emissions of all contemporary AFV are becoming enormous, from high bandwidth digital communications enabling a sensor fused battlespace to their navigation and situational awareness systems and more recently the increasing adoption of active protection systems (APS), most of which utilise radars that can be detected at ranges of up to 500 km in some instances.
In a big picture sense, those soldiers are now at greater risk than they were in their Bradleys. They are now more easily identified and killed with fewer engagements, including long range precision fires that might otherwise be unable to find them. And if killed (whether a mission kill of their platform or literal kill of the personnel), both of the RCV they control will cease to operate, regardless of their own status. It now takes one hit to eliminate three vehicles from the fight, and to kill far more soldiers too.
Is that a better vision than the current conventional picture?
Logistics, deployment & support: Who exactly keeps RCVs running?
It is all well and good to discuss RCVs in the tactical context, discussing how they would operate either in isolation or as an integrated element of a MUM-T formation, but there is a much larger picture to be considered. When not sallying forth alongside a line of tanks, how are these RCVs deployed, maintained and when broken down or damaged, recovered?
They don't magically reach the battlefield, and will require dedicated transporters. Autonomous convoys are an area of interest, but are not mature yet and in any case would not be able to autonomously load and secure RCVs for transit. So they will need to be handled by humans. This isn't a huge issue - they need to deploy with their command vehicle anyway for the foreseeable future, which itself is manned and being transported, but there does need to be acknowledgement of the burden of transporting entire fleets of new vehicle types and how these will be funded and crewed.
RCVs like any AFV require maintenance and repair, and much of it is very regular and quite onerous - AFV crews spend much of their time checking, adjusting and fixing things, often more so than they do actually manoeuvring in vehicles. RCV by their nature tend to be smaller and lighter than most manned equivalents and so have a lighter maintenance burden, with disruptive technologies like hybrid drive and composite rubber tracks (CRT) able to mitigate that burden - but it is still significant.
As mentioned above, the crew of an RCV fire unit have their own AFV to maintain, but now have a number of RCV to work on too. If they are geographically dislocated during operation as is envisaged and desired, who exactly is checking and tweaking the vehicles, and how is this workload being managed? Perhaps it could mirror arrangements like the French have with their Leclerc tank formations, which attach small teams in VBL 4x4s to provide additional personnel for administration, security and maintenance of the tanks.
If we have removed crews from the vehicles but consequently created an environment where we now need more engineers and mechanics in the combat area, and who are critical to those vehicles functioning, have we again created more risk to those personnel and key nodes of attack to disable the RCVs without engaging with them? And are planners taking account of the cost and additional equipment required for these capabilities?
What about when RCVs break down, are damaged, or simply get stuck? Recovery of an AFV is a regular task, sometimes solved by the crew locally, but sometimes requiring intervention of an ARV or AEV to extract them with specialist equipment. How easily can an RCV operator determine the cause and severity of an immobilisation from their operator screen, unable to get out and look at the vehicle?
Without adequate solutions, the vision of a robotic battlefield is potentially one where logisticians, electrical/mechanical engineers and recovery specialists are the new combat arms, the only ones going into harms way and doing so to resolve issues that may have been readily addressed by crews on their own in a manned platform. What is the net effect of this much higher demand for specialist services, equipment and personnel?
There is a separate answer to this particular challenge - that RCVs be considered expendable like a missile of PGW, and so an immobilisation or damaged vehicle is simply discarded until it can be recovered later. The US RCV programme sees RCV-L and RCV-M as 'attritable' (disposable) and 'degradable' systems, yet has no plans or discussion of how potentially large numbers of these vehicles, which are not insubstantial at over 12 tonnes each in the case of RCV-M, will be moved, stored and deployed.
Though an interesting discussion that will be explored in the second 'pro' part of this post series, the concept of disposable RCV raises its own issues around logistics. If RCV are disposable, and so there is no need for additional ARV/AEV and other capabilities to recover and repair them, then who, what and how is the tremendous burden of transporting and deploying even greater numbers of RCV to allow for replacements and 'one-way' missions being accommodated? Vastly greater numbers of logistics vehicles and personnel to operate them are required, and credible means of deploying them in volume.
Permanence: A flawed value proposition?
One significant value proposition for UAV is that the endurance of an aircraft or boat can radically outpace that of the human operator, and so UAVs offer the ability to transit or loiter on autopilot for prolonged periods whilst cycling remote human crews brought in to operate payloads as required. A manned alternative will have very short comparative endurance and on-station time, and so the UAV offers a unique capability advantage.
The same simply is not true of an RCV. Service leaders and commentators talk of permanence and "all-weather, 24/7" endurance, but in reality, the RCV argument is often the inverse of the UAV proposition. Unlike aircraft, a vehicle or other manned formation is physically on the ground and will remain there. It is not transitory in nature, they arrive with food, water, ammunition, batteries and supplies to stay there. Ground forces operate over or dig in and hold ground for days, weeks or longer.
So then, how is an RCV providing permanence and endurance that the manned formation cannot?
To function, the RCV must be turned on and burning batteries, fuel, or both. Its endurance is entirely constrained by that power generation dependency. If it is to hold a piece of ground under its control, it will sit until its fuel supplies reach a point that it must withdraw and refuel or recharge. For most current and near-future RCV designs, that is around 72 hours at best.
The manned alternative, is a number of vehicles accompanied by crew and infantry. A larger grouping in almost any instance, certainly, but one that at worst is a peer of an RCV, and at best can vastly outlast an RCV 'on station'. Troops can dig in, shut down vehicles, turn off radios, and subsist on rations, which even dismounted infantry without vehicle support can carry up to 10 days with them. As required they power up systems to observe or communicate, and shut back down. The enduring energy burn rate is negligible compared to an RCV, which has to remain on constantly to operate. A formation on position could operate for days, or even weeks, before needing significant support.
An RCV cannot claim to any kind of permanence in this context, certainly not to a greater degree than the conventional situation. Any technologies that enable longer RCV endurance are applicable to the manned equivalent, and so maintain the disparity rather than close it.
Closing comments
Individually these are substantial obstacles that will take many years and many more billions of dollars for the global community to devise solutions to. Collectively, they are an entirely impassable obstacle for the credible fielding of RCV systems in the immediate foreseeable future.
Should we discard RCVs as a foolish idea and a waste of limited developmental resources - of course not. But perhaps the discussions should shift from an aspiration to field these system in the near future to a more measured recognition of the key issues and iterative development to work through the obstacles.
A refocus to aiming to empower the next generation of manned AFV with high level autonomy would likely create better solutions that are actually viable for fielding in the near term, whilst maturing the exact technologies that prevent a fully autonomous RCV from being fielded so that they can be considered in the mid to long term.
Whilst experimentation with RCV and other UGV in a tactical sense should continue, there must be much greater focus not on how an RCV might be employed tactically and any procedural or organisational issues with how they might be integrated with a force, but on the big ticket issues that prevent their use at all.
By way of closing example, this month's Robotics, Uncrewed Systems and Human-Machine Interface day at IQPC's International Armoured Vehicles conference (which is a fantastic event anyone with an AFV taste should attend) has 14 sessions on RCVs and not one addresses any of the issues in this post, all focus on how an RCV might fight, what missions it could do, and how they might integrate with human forces. The mindset and focus of those asking for this equipment and developing it needs to be much broader.
Part two will be posted soon and will cover the other side of the coin - why RCVs are an amazing proposition and perhaps fundamentally change the face of ground manoeuvre warfare. Until then, the issues above need some serious thought and some serious money to overcome.
Good article, I don't see UGVs being ready for prime time any time soon for exactly the reasons you stated. That said, I think an "optionally manned" platform being the way forward, with 1 crewmember in the vehicle to handle local problems, supported by (initially) remote operation of the weapon system etc. by external crew and by increasing levels of AI. Finally you can remove the operator entirely.
Something a bit like this:
https://community.battlefront.com/topic/142139-new-armoured-vehicle-concept-lessons-from-ukraine/
I'm looking for the accompanying "Pro-UGV" article, but cannot seem to find it. I'd appreciate a link or assistance. I'm sorry if I'm simply blind to the obvious.
I would like to mention, that while a UGV capable of successful navigation in any terrain with very high reliability is not here yet, it is not supposed to be ready yet.
UGVs today are providing a lot of value in controlled environments, whereas RCVs are only really part of a concept enabled by more advanced manned combat vehicles, like OMFV and Carmel, which are almost a decade away.
Progression from here to reliable UGVs should be linear. All the physical principles are known, so it is an engineering challenge.
I think we will stumble upon a much greater challenge when it's time for the next generation of RCVs - how do we achieve the same results without the use…