Military Tank drone, why are they not common yet?

Question

I was wondering--why are there no self driving tanks in the US army?

They seem much easier to build than a flying drone. Most of the space within the tank is wasted on instruments/cabin space for the crew. Removing all of that would allow the tank to be much smaller or have thicker armor. It might even reduce the overall cost of the tank and make it more robust.

Unlike a flying machine, loss of connection or autopilot malfunction is not catastrophic. In the worst case scenario, your tank just sits idle or runs into a wall instead of crashing like a flying drone would. You wouldn't mind sending such a tank to the front line, since there are no human lives at risk.

It would probably aim and shoot much better than human pilots.

I'm really surprised that they don't exist yet.

Answer 1

Unlike a flying machine, loss of connection or autopilot malfunction is not catastrophic, in the worst case your tank just sits idle or runs into wall instead of crashing like a flying drone would.

Totally incorrect. Quite the opposite actually.

Flying autopilots have been common for decades now and they have become even better. If a flying drone loses connection it won't crash, obviously. Its autopilot will kick in with preprogrammed "lost connection" routines. So it can at least evade the enemy and avoid destruction or capture.

A ground drone is the exact opposite. We have yet to create an acceptable "autopilot" for ground vehicles. Especially if you are in unknown and complex terrain like a city or forest. So if a ground drone loses connection, it is only matter of time before it is destroyed or, even worse, captured.

On top of that, it is much easier to lose connection to a ground drone than it is to an aerial drone, magnifying the above problems.

It would probably aim and shoot much better than human pilots.

That is again not true. If such technology existed, it would already be used on tanks even with human crew. Even with such a high level of automation, humans are still needed for command and control. And even with full automation, it is worth having pilots inside just for backup. So there would be no difference in firing ability of the tank with or without crew.

Answer 2

Many good answers here, but I thought I'd add my two cents as a former infantryman.

One of the biggest benefits to using a drone aircraft is the mitigation of risk to human life. We are willing to pay a premium in monetary terms in order to limit risk to human operators of our weapons. This works with UAVs because (as was stated previously)

• Airborne autopilot technology is mature (in case you lose connection with the drone)
• UAVs operate at reasonably high altitudes most of the time so we usually have a good line of sight to the drone from the ground or a satellite, which mitigates the problems of jamming
• There are way fewer things to bump into at 50,000 feet than there are at 0 feet, so even imperfect controls and AI give us a good chance of sucessfully navigating and getting our UAV home

With a UGV (unmanned ground vehicle), we have none of these advantages. Hills or concrete buildings block our signal to home, tons of complex obstacles challenge our ability to navigate, and ground based autopilot is still unreliable.

Beyond all that however, and VERY importantly: WE ARE NOT REALLY ACCOMPLISHING ANYTHING!

Why not? Because human lives ARE at risk no matter what! What most people don't realize is that tanks are not used as "stand alone" weapons. You do not use an MBT the way you might use a B-52 to just cruise over to the enemy, drop some badness, and scuttle away. Tanks cannot operate this way. They hold terrain. Tanks must control terrain by use of their powerful weapons and survive by use of their heavy armor. They remain in place or move together in order to cut through enemy formations, form lines, or flank an enemy force. 99% of the time, tanks are NEVER alone! They are surrounded by infantry, either in lighter vehicles or on foot, or a mix of the two. This infantry keeps the tank alive by spotting infantry with anti tank weapons and eliminating them. They are eyes and ears for the tank, which has a limited ability to "see" what is around it and no ability to hide in small crannies or be stealthy.

In the 1973 war, one of the major lessons that Israel drew from the annihilation of their (highly skilled and advanced) tank force in the Sinai was that they must NEVER use tanks without infantry support again. Tanks by themselves are nothing but targets for infantry carrying ATGMs.

Given all that, you gain NOTHING by creating an UGV tank, because you still have to have guys on foot all around it in order to be sure it doesn't become an easy to hit rolling target (and probably a slower one than a conventional tank because the AI will be challenged to think it's way through battlefield terrain). So you just spent enormous amounts of money to get rid of the crew of the tank, but you still have tons of guys right up at the front, and you can't use an MBT by itself without them, so there is no political payoff.

In addition to all this, auto loaders have never been reliable. Tanks are so complicated that the more things you automate, the more things break, and the more time the thing spends in the shop. UGV MBTs would be a maintenance nightmare.

Answer 3

Since there are no self driving cars, self driving tanks are right out of the question, because they not only have to cope with terrain that doesn't have neat signs and lines to show the way, but also with actual hostile action.

Question why there are no remote controlled tank is valid, though. There are multiple reasons:

1. Military prefers technology that can be relied on. You don't want your spiffy shiny new tanks bursting in flames because engine didn't broadcast warning signals to operator. Using the absolute cutting edge is a great way to suddenly discover fatal flaw at the most inconvenient moment.

2. Designing and building weapons costs a lot of money. You can't afford to replace your entire armour complement of thousands every 5 years because there's new model. Tanks like Abrams or Leopard were around for 3 and a half decades. T-90 and Challenger are in service for about 20 years. However, those tanks have been upgraded multiple times since their commission. New turrets, modified armour plates, better guns, better ammunition. All that improves their performance WITHOUT requiring major redesign and without incurring excessive costs. Also, all nations keep a stock of mothballed obsolete equipment, for sale and emergency use.

First two factors taken together: you want a lot of working and good enough equipment, not few prototypes that will break down because infancy problems weren't ironed out and those which remain are swarmed by enemy. WWII was decided by tens of thousands of workhorse vehicles: Shermans (USA), T-34 (Soviet) or their Axis counterparts like PzKpfw IV and STUG III, while famous Tiger and Tiger II bore relatively small impact due to small numbers. Nearly 50000 Shermans and 65000 T-34 were produced before the war ended. Tigers? Just a bit over 1300. Tiger IIs were even less numerous.

3. A lot of the technology required to make that happen didn't actually exist until very recently or didn't exist in practical form. Autoloaders exist, but they were problematic. As far as I know, there's no automatic target acquisition and fire control in service anywhere - ballistic computers don't replace gunners, they only assist them in rangefinding, compensating for gravitational drop, weather and target speed. Infra-red imaging makes identification of targets easier but not automatic - modern computers still can't be relied upon to differentiate tank from car so human has to identify target. Remote guidance and digital cameras weren't robust, cheap and reliable enough to be used on a tank.

   Basically, while it is an obvious next step, it's actually much harder than it seems.

   That being said, there are changes in this direction:

   • TUSK (Tank Urban Survival Kit) system for Abrams, introduced in late 2000s includes machinegun controlled remotely from inside.

   • Trophy system introduced in early 2010 includes 360 radar linked to pellet launchers and is capable of shooting down incoming rockets without any input from operators (Soviets were using active anti-missile countermeasures since 80s, but their solutions weren't as precise as current ones).

   • Russian T-14 Armata tank unveiled last year has unmanned turret. Main gun is loaded by autoloading system and remotely controlled from crew compartment inside the hull. Gun control is supposedly able to automatically track and engage targets designated by gunner (IR tracking of signatures). Other systems it's supposed to have include active anti-missile countermeasures (which require no input from crew) and machine-guns operating in both remote-controlled and IR-based autonomous modes.

   T-14 is probably the first design to integrate so much automation into single platform, but it's not actually superior to NATO designs in firepower and protection (it however significantly changes tactics). What it is, is a wake-up call to quickly start work on new tanks. Upgrades to existing vehicles soon won't be sufficient. Technological race never ends - Russians are already working on upgrades for T-14.

4. Jamming. Problem with remote control is, if your signal can be jammed, then your device isn't of much use. That can be partially remedied by directional antennae, you can't really jam them unless you are in the way, and in case of flying vehicle, with horizon at the distance of hundreds of kilometres, it's relatively easy to create a chain of unobstructed links. On ground not so much. Tanks have to plow into enemy territory, behind hills, field and cities, where the only direct unobstructed link to base would be through satellite. Lightspeed "ping" to low earth orbit and back is 1ms (ignoring processing speeds), "ping" to geostationary orbit and back is 0.25s. To operate tanks remotely you would need satellites capable of transmitting real-time image, exceeding 4K by far, from dozens of cameras, from thousands of tanks, and you would need thousands of LEO sats if you want to control your tank for more than a minute per day, or dozens of geostationary sats and you would get 500+ ping to use gaming as comparison. Neither of required networks exists, and neither would be cheap. If you want to try and relay through some AWACS equivalent, you run into the problem that you need total control of airspace, but if you do, then you are against enemy against whom tanks are impractical, you need infantry instead.

5. Logistics. Tanks don't actually drive to the battelfield. They are carried on trucks, trains or ships as close as possible. Due to weight and huge mechanical stress involved in propelling 50+ tonnes at 60+ km/h, tank engines are rated for merely thousand km or so. After that they need service or replacement. Tanks need ammo and a lot of fuel. They also need personnel to resupply and service them. Unless you can automatise resupply and service, you need people nearby anyway.

There's even more, but for now, this is where I stop listing reasons.

Answer 4

• The ground is a more crowded, complicated environment than the air. In the worst case, an UAV out of contact will fly in a straight line or circle. In the worst case, an UGV out of contact might drive into a ditch or into a house.
• Drones are under remote control when it gets exciting. That works because they usually have good commo links, either satcom or line-of-sight. The communications environment on the ground, especially in mountains and urban areas, is much more difficult. Reception may be degraded by obstacles at inconvenient times.

Answer 5

Firstly, the concept of your remote tank is a bit off. A tank that engages in fighting, i.e. firing its main armament close to the enemy requires always friendly infantry support. Without adequate support a tank is easily flanked and destroyed by enemy units. Thus, the drone tank would require regular (or remote controlled) infantry to support it, thus removing the advantage of not sending humans to the fight. A tank that would not need to engage enemy, for example carrying out reconnaissance could be a viable application for automation, but then again a cheaper flying drone is much better suitable for getting better picture of the surrounding areas.

Secondly, there are quite many technical obstacles. Other responses have already argued well the connection issues that would occur due to hills, forrests and buildings. In addition, coming from some experience with tanks, driving a tank requires multiple decisions being made at the same time and close collaboration from the crew. Whereas a flying drone can be set to follow certain path and operator can concentrate on acquiring targets and firing, tank commander has to issue multiple commands simultaneously that all affect the battle effectiveness. The driving instructions, for example, to the driver determine the opportunities to acquire targets, remain hidden from enemy, fire main armament and survive possible incoming fire. All these decisions are made by a human and they all interact (driving, reloading, firing decisions cannot be made indipendently as in a flying drone). As we have no such artificial intelligence in hand in any conceivable near future, the remote tank would need multiple remote operators (at least driver, commander and gunner). This would increase the communication overhead, as they all need enough sensory input from the tank (camera, voice, etc.) and all need to be able to issue near-realtime control commands to the tank. Even in a modern tank situational awareness is a key challenge tankers face, doing everything via laptop and some cameras would be ridiculously difficult. Moreover, when ground fight gets hot, the pace is much faster. Life-or-death decisions have to be made quickly. The flying drones by remaining unseen can take their time to carry out the issued commands.

Answer 6

Because they would serve a different use case, and so far nobody high enough decided that investing in that use case is worth it yet.

Expensive heavily armored remote controlled tanks are inferior to tanks with crews for a variety of reasons, which mostly boil down to lower reliability. At the price point of a tank, lower reliability is not something you want.

The use of remote controlled tanks would have to be different - weakly armored, cheap, even disposable remote controlled tanks. Pretty much cannons on wheels. It's the same as with the flying drones which aren't super expensive remote controlled fighter jets. They can shoot and if they get shot themselves, who cares, they're disposable.

The problem with these disposable remote controlled cannons on wheels is that in large scale engagements they are somewhat unpredictable, especially due to jamming. While in small scale assassinations, their role is already filled by flying drones. Solving this problem requires full automation, which is an unsolved technical problem, as well as an ethical problem.

Long term, due to the very high land speed such cannons can achieve, they will certainly appear, probably some time after high speed scouting drones on wheels. And once they appear they will certainly lead to new tactics.

Answer 7

It's also worth considering the things that an armoured fighting vehicle needs people to do for it. An awful lot of the role of the crew of an AFV is keeping the AFV running.

However, it is worth considering that military unmanned ground vehicles are in development, primarily for logistics where the loss of a relatively inexpensive truck and some cargo is much less important than the crew - take a look at the TerraMax as an example.

Answer 8

Why not on the ground?

-sensor blocking Computers are not good at image processing. It is hard for them to identify objects by sight or deal with foreground objects blocking background objects. From the air it is much easier to see and there are fewer obscuring objects

-navigation as noted earlier autopilot in the air is much easier than on the ground there are no obstacles to avoid, no people to not run over, no mud to get stuck in.

-Delay loop The army wants a human to be "in the loop" when ever a drone kills something due to speed of light limitations to get up to a satellite back to Nevada and back down this can cause a full second delay between an even and the drivers response.

On the ground that second delay is very costly. The enemy can see you coming you are firing unguided munitions(bullets) so you have to lead the enemy its a harder problem. In the air there is a building or a car and a guided missile it still works even with the delay.

Why not Armored Drones mean you don't have to survive. Why is a tank armored? So it can survive being shot. Because a tank crew and their tank are expansive and hard to replace. A destroyed drone is only expensive if the drone is expensive.

This means that drones will tend to be lighter and cheaper than the normal versions. A tank drone is more likely to be a lightly armored golfcart with 2 sidewinder missiles than an Abrams with computers stuck in it.

Answer 9

There has actually been a lot of research and debate in this area for a while. Both small scale (eg. packbot, TALON) and larger scale (e.g. MULE robots). Full sized robots or drones on a tanks scale are prohibitively expensive (witness how long between Abrams upgrades, much less redesign, and the time and money spent on things like Future Combat System and the Joint Strike Fighter), both of which included significantly more automation than today's fleet of vehicles, but neither were anywhere close to a full-on drone.

Additionally, while I agree that the computing power required to generate a firing solution for a tank is not prohibitive, what is prohibitive is the moral side of things. I made a comment above but figured it deserved expounding upon.

First, the US does not want to relegate the decision to take a life to anything other than another human being. This includes sketchy feelings about even humans pulling the trigger when a robot aims, and I can't find the article now but there was a follow-on piece to that involving a malfunction that wasn't catastrophic but damned scary.

Second, when it comes to large robots, there are still a lot of safety concerns (even on the battlefield). On modern battlefields, there are a lot of civilians / noncombatants still, and you can't just ask them to get out of the way. A drone typically has some non-trivial delay from command to execution due to bandwidth and signal travel limitations, and the last thing the US wants to do is run over someone's kid with a robot tank by accident, because the signal was delayed and the autonomic sensors didn't sweep the right spot of ground to "see" the kid in time.

The other answers are also mostly great, just wanted to cover things I thought were missing.

Answer 10

Flying drones have a lot of major advantages over piloted aircraft

• Potentially much cheaper. This is less true in the case of fighters, but reconnaissance can be done by tiny drones that cost less than even training a pilot.
• Expendable. You can send thousands of recon drones over hostile territory and it's not a big deal if many are shot down.
• More agile. Modern fighters are limited by the fact that humans cannot endure really high G forces. Drones can have the same performance characteristics as guided missiles.

Tanks, on the other hand, will always be very expensive, which also means they are not very expendable, and they will never be agile enough that G forces matter. So there is not much incentive to develop a tank drone.

Answer 11

Self-driving cars are hard

Self-driving cars are actually hugely difficult to make. Imagine a camera feed. You have a grid of pixels. Essentially a huge chequerboard full of binary digits.

Now try to write an algorithm that can take that chequerboard and turn it into a meaningful map of the surroundings. It's a very hard problem.

Going off road multiplies the problem

If you are on a road, you can reduce the scale of the problem considerably. You are surrounded by rectangular cars. Slow moving upright objects are probably people. Converging light coloured lines are probably roads.

Going in the air makes it easier still. There is almost nothing to hit. All you need to do is proceed to a GPS coordinate.

In a tank none of these assumptions hold true. Tanks are camouflaged. They may be partially occluded. You might be in a forest or a destroyed city. There might be obstacles.

Now add in tactics

The bridge has been damaged and might be dangerous to cross. The trees that should offer cover have no leaves because it's autumn. The enemy is flying a false flag. There are RPGs on that hill over there, covering the road.

A human with common sense (learned heuristics) can make these complex, unpredictable decisions. A machine can't, because machines can only do what they are programmed to do.

What about a learning computer?

You might try to make a computer that can learn from its environment. This is a very hard problem indeed. You now have to take your chequerboard input and somehow convert that grid into "rules for staying alive". No one is even close to being able to write a program to do this. Such a program may not even be possible. We might need quantum computers, or something exotic we haven't yet thought of.

All current machine learning algorithms require carefully curated datasets, and hand tuned algorithms. You can't just chuck in a real world, noisy video image at them.

Our current best computers give us no more than insect level intelligence if that. You would need mammalian level intelligence to power a free roving autonomous tank.

People are clever. Computers are dumb.

Remote control

So what about remote control? In the air, if your signal is jammed, you can simply fly to a preprogrammed GPS coordinate and land. On the ground, if you lose your signal, you are on your own. You can't drive home in a straight line. Plot a predictable route home and your unpiloted convoy could be taken out tank by tank by a guy making holes with a mechanical digger, or some similarly unpredictable obstacle.

The future?

We may one day be able to make a computer intelligent enough to function autonomously in the world without a human guiding it. That day is a long way off and we have no roadmap of how to get there. We really have no idea about what intelligence actually is, or what problems we need to solve to reach it. True AI is still no more than sci-fi.

(Source: I studied AI at Sussex University.)

Answer 12

Keeping a tank operational in the field requires lots of maintenance. Fuel. Oil. Track tension. Lubrication. Replacing parts. Without those things, a tank will pretty quickly become inoperational. (Look at late WWII German tanks. Magnificent beasts when operational. A nightmare to maintain.)

Tank crews are trained to do this kind of field maintenance. Actually, one of the arguments for a human loader instead of an autoloading system is the additional pair of hands servicing the tank when not actually in combat.

You need hands, right there where the tank stops, to ensure it can perform the next day as well. While you might be able to keep human tankers out of the "shooting" war by using drones, you will need just that many technical personell very close to the frontlines for doing the maintenance on them... people who cannot, if pressed, just jump into the protection of their tank and shoot back...

Answer 13

Communications

When the drone is in the air, the controller can be on the ground a few miles away with a dish antenna and be able to communicate over a simple line-of-sight radio link with the drone. For more distant operations, the controller can be airborne or can use an airborne relay.

When the drone is on the ground, even just a mile or two away, ground clutter and interference absolutely requires a relay. And even then, things like uneven terrain (hills, ditches, etc), vegetation, and gouts of dirt, dust, smoke, and water from incoming artillery will significantly interfere with the signal.

Answer 14

Can't believe someone hasn't added that a Tank Drone on the ground is ripe for the taking. Find a weakness, overwhelm it with targets and now YOU have a cannon you can turn against your oppressor.

If you take a flying drone down, you have a bunch of junk. If you capture a tank, you have a FREE weapon you can now use.

Answer 15

Reliability, and to a certain extent, culture. The americans, and british for example do not like autoloaders, with a human loader and an isolated compartment for shells being less likely to cook off and kill the crew.

In addition, at the end of the day its a human who makes the decision to pull that trigger, and kill someone, not a machine, least in this day and age. You still have to decide if you want to fire, or get the enemy to surrender, and occationally act on gut instinct.

Unlike a flying machine, loss of connection or autopilot malfunction is not catastrophic. In the worst case scenario, your tank just sits idle or runs into a wall instead of crashing like a flying drone would. You wouldn't mind sending such a tank to the front line, since there are no human lives at risk.

Which the enemy can capture and take apart, potentially reverse engineer - the americans lost one drone to the iranians and well, the chinese and russians probably got a peek inside. Imagine that happening to your newest tanks - the americans destroyed mission-killed but otherwise 'intact' tanks in battle

That said, eventual automation makes sense. A remotely controlled turret might be useful - the russians have extended the concept of the autoloader to build an unmanned turret into their latest tank. Considering the reliance on GPS, and that many of the technologies used in self driving cars would be useful in close quarters navigation like LIDAR, non combat self driving feels potentially natural, as does smaller crews. You might also have slightly expendable/cheaper semi autonomous tanks acting as decoys or supporting tank platoons

Answer 16

Just to clarify: by tank you mean actual tank (i.e. Abrams) or any armored vehicle (i.e. AFV Bradley or Stryker)? Might be a good idea to clear that up...

Contrary to most other answers it is very simple to produce a UGV that's quite... reliable, let's call it. You can now remotely hijack and control most newer cars, so highly computerized tank is even easier to autonomate*. For the same weight and space you can equip tank with recognize-and-follow-the terrain unit that's installed in Tomahawks or SpecOps aircraft. Add to it INS or other stuff you can have pretty good self-driving unit. Not to mention that for smaller obstacles what you need is buffed-up and pimped-out parking sensors...

Same with gun crew - aiming and firing is already highly automated, as is the loading. But here's where complications begin. You can have essentially two types of autoloader: Soviet-style carousel or multiple-queue-feeder. As last 30 years have demonstrated carousel type is highly susceptible to catastrophic hit (i.e. "golden bb hit", where tank explodes pretty spectacularly with turret flying high and far). Multiple-queue-feeder is too big to fit into a turret of an MBT - compare tank to Navy 155mm destroyer cannon and you'll see the size of the whole unit is... bit different. There's a reason why even MC-130's 105mm gun is manually loaded, even if there's a lot more space there than in MBT's turret. Of course there are smaller options/solutions, but they all will be bulky (to say the least). And might be vulnerable to a hard hit - that is if hit directly (even without penetration) it might jam as it will have much to many moving parts. Same argument goes for all kinds of electronics, actually - as much as you'd want it's still not great in "shock-proof" department.

I'd say that main concern is (cue in dramatic music jingle)... Communications! Jamming signal is a concern, but not that much as one would think. Most of military traffic goes through secure satellite communication which is not exactly easy to jam. It's not difficult, too, technically, because if there's any satellite redundancy, it's pretty much impossible (you have to target the satellite, and for that you have to know where it is). But that's why there's redundancy. Not to mention intra-unit tac-net including BFS etc. Also signal repeaters are not exactly new... The problem is signal latency. For standard sat comm (geosynchronous) it will be at least half a second (including encryption/decryption), maybe more. This is the same time sabot needs to travel almost a kilometer down the range... In other words: half a second is difference between hit and miss in most combat situation. "Moving target is harder to hit" is no longer valid axiom - it's now "Randomly moving target is harder to hit". You can't have random with such delays...

Last point: as mentioned in other answers - MBT is never alone on the battlefield. There are Scouts, Screening units... Remoting them all will definitely jam the bandwidth and signal delays mentioned earlier makes the whole point moot, anyway. If we're talking supporting other units (i.e. infantry) it's even bigger issue. It could be remedied by making the unit autonomous, but it there's any malfunction results can be... catastrophic. Not to mention the fact that being around Abrams firing it's main gun is neither pleasant nor safe... Tight fire control is strongly suggested...

From what I've read on tank crew operations (by actual Abrams crewmen) manual loading is much more reliable and faster than autoloaders, driving a tank is not the same as cruising in an Escalade... Because it's not technology but training and experience is what makes 5-man crew operate as nearly-one. But they're still autonomous, so tank commander does not have to think to tell driver how to drive and maneuver nor loader how to load. He gives order and crew expands on it according to their training, experience and role. This is the oldest and best understood for many centuries force-multipier, which is also the biggest one...

Also they are cheaper - tomahawk is quite pricey exactly due to it's sensor/guidance package. Add all the equipment needed to make Abrams UGV and you're doubling it's price tag... No one will pay close to $10 mil for one tank... Which is, in my opinion, main reason for not having UMBTs. Or UAFVs, for that matter.

*not an error.

Answer 17

I think that frankly flying war drones are just more effective than ground drones would be regardless. Check out what the Iraqis have done. https://futurism.com/the-iraqi-army-just-deployed-its-cannon-equipped-robot/ Not self driving but remote controlled makes killing like playing Halo. One person steers while the other utilizes the main cannon.

Answer 18

Maybe not a MBT (yet) but the Russian Whirlwind is an unmanned combat ground vehicle (UCGV) armed with a stabilised 30 mm Shipunov 2A72 automatic cannon and other weaponry. I suspect an MBT can't be far behind. And there's the Israeli Guardium as well with a lot less weaponry but featuring a fully autonomous mode.

Answer 19

Several reasons

1. An "Unmanned" vehicle must still receive instructions, that means signals which can be hacked, jammed or intercepted. The defense/tech industry is already experimenting with anti-drone guns that fire interrupting signal beams. https://www.youtube.com/watch?v=ADW63thj-Pg
2. Repair/Recovery: If the vehicle is damaged; you have to send people to fix it anyway. Unless you have a unmanned tow vehicle to recover it.
3. Resupply, upkeep of the vehicle overall must be maintained, By whom?