Autonomous compaction

Self-operating vehicles and construction equipment seems like one of those technologies that is always 10 years away. Ten years ago, it looked like by now we’d all be taking a nap on the way to the office while our automated Tesla did the driving. 

Instead, at the last ConExpo where Tesla was supposed to be showing off its self-driving technology to ferry people through Elon Musk’s tunnel under the convention centre, the cars still had drivers. 

That didn’t stop the exhibitors at ConExpo from leaning hard into automated versions of their equipment. Excavators, loaders, skid steers, graders, aerial equipment…just about every major manufacturer had a vehicle that didn’t require an operator sitting right in it. But when you asked them, it usually turned out that the “autonomous” device was actually remote-controlled by an operator standing off to the side with a tablet or control box.

Truly autonomous construction equipment that can carry out jobsite tasks on its own still seems distant. But when and if it comes, it looks likely that the first applications will be in compaction, for the simple reason that ground rolling happens slowly on relatively flat ground and is mostly just a matter of driving rather than controlling the complex manoeuvres of booms and buckets. Trench compactors, especially, are going to have a hard time steering off course. 

To get a sense of the state of autonomous technology in compaction, we turned to the company that has probably done as much as any other in researching
compaction autonomy and bringing prototypes and demonstration models out into public: Bomag. John Gravatt, product marketing manager for Bomag’s asphalt division, kindly provided a great deal of background for this story.

We can see that compaction is ahead of the rest of the industry when we look at how common remote-controlled versions already are. Every major manufacturer offers a remote-controlled version of at least some of its models for commercial sale. They are not an uncommon sight on jobsites everywhere. But the dream is to free up an operator entirely. It might be safer for the operator to be away from the machine and not down in the trench with it, but their time is still dedicated to that job.

Another level of automation is “office controlled” where the operator is sitting in a comfy room controlling the machines on the jobsite over screens. Gravatt says these environments are hard to set up and tend to be for specialized applications where worker safety is a concern, like mining or working in areas contaminated with toxic substances.

“Fully autonomous systems are in their infancy, generally speaking,” Gravatt says. “These machines incorporate the latest sensor technologies and machine vision systems, along with some deep learning software systems powering them.”

MOTION CONTROL
We’re all familiar with global positioning systems that use radio communication with satellites to locate a signal on a map. GPS can tell an autonomous roller where it is in a general area, but it’s only accurate to about 16 feet. That’s good enough to keep the roller from trundling out into the highway, but not much use for close manoeuvring on a jobsite. It’s role with controlling construction equipment is to set up a geofence that will make sure the machine doesn’t leave a prescribed area. Or get stolen, but that’s another matter. If the area is flat and clear of obstacles (think of a large parking lot), and the roller doesn’t need to closely approach the boundaries, a geofence could suffice combined with a route programmed into the machine’s control. 

“In the future, autonomous machines will still need to be ‘told’ where to perform their duties, just like an operator,” Gravatt explains. 

But to navigate more complex jobsites, and to allay safety concerns, autonomous rollers have to use a combination of more sophisticated technology to detect and avoid obstacles. Three of these are LIDAR, ultrasonic sensors and vision systems.

Vision systems are based on regular video cameras tied to software programmed to recognize shapes that represent obstacles. Vision is good at seeing a large area and detecting any obstacle within it. But it can be hampered both by low and high light conditions, insufficient contrast between an obstacle and its background and conditions that reduce visibility like dust or rain. 

Ultrasonic sensing is similar to the SONAR used by submarines to locate objects at a distance under water. Ultrasonic sensing is not as precise as vision at picking up the exact size and position of objects, but it’s better at penetrating atmospheric obstructions and determining the distance to the target. 

LIDAR – laser imaging detection and ranging – is the gold standard for obstacle detection and the technology that has made self-driving cars possible. Ambient light conditions don’t matter and LIDAR penetrates mist and dust better than vision. Of course, it’s the most expensive of the obstacle sensing solutions and the price likely needs to come down before LIDAR-equipped autonomous rollers become commercially viable. 

Demonstration rollers introduced at trade shows have used combinations of these systems to balance out the strengths and weaknesses of each and generate a working solution.

COMPACTION QUALITY
One area that used to be a barrier to automated compaction but really is not any longer is ensuring the soil is compacted to the correct condition. Vibration sensors have been able to pick up feedback from rollers for a long time and feed it into systems to gauge soil stiffness and density. Combined with modern software and control systems, rollers can automatically adjust their amplitude to address the sensed conditions. The last step in a fully automated roller would be to have the AI monitor this data and predictively adjust speed, amplitude and the number of passes accordingly. An AI could theoretically do this a lot faster and more reliably than a person, resulting in a more consistent final soil condition achieved faster. 

Asked if there are complex scenarios where a human operator would be superior to an autonomous roller, Gravatt says there are but not many. “One that comes to mind would be compacting fresh asphalt next to a live lane of traffic.  There are lots of variables changing continuously in that scenario, so the machine must constantly monitor and adjust to that.  But autonomous machines may also have an advantage in these scenarios, as they have different ways to ‘see,’ like LIDAR, precision GPS, ultrasonic sensors and certain camera systems. They may not be as sensitive to headlight glare, for example, and potentially lead to less risk on the jobsite.”

REPAIR AND MAINTENANCE
Enough about all this benefit-to-the-customer stuff, let’s talk about what a rental store really needs to know: how hard would it be to keep these things out on rent without frequent service calls? “I wouldn’t think an autonomous machine would be more difficult to repair,” Gravatt predicts. “It may actually be the exact opposite. When you consider that there are several systems that are no longer needed on autonomous machines, they may become less complex and more reliable.  We won’t need things like cabs, electrical systems for the user interface (switches and displays), air conditioning systems, radios, and the list goes on. All those things add additional complexity to the machine. From a design perspective, there is one less major variable to design for: the operator.  Now, the focus can be shifted to serviceability when designing the machine layout. Of course, we will see the addition of some new sensing technology. But these sensors typically have failure symptoms that the on-board computers can detect and diagnose relatively easily.

“Additionally, there’s potential for a reduction in user error with autonomous machines, which can also reduce the need for repairs. Take our Asphalt Manager 2 for example: the automated vibration amplitude control limits the user’s ability to use too much vibration force by sensing the feedback from the asphalt, preventing over-compaction and eliminating drum jumping and thus preserving the equipment.”

Axel Romer, head of R&D at Wirtgen’s HAMM group, is quoted in an online article speculating about some addition benefits engineers could fit into automated rollers if they didn’t have to accommodate human operators. “We could construct autonomous rollers with significantly larger drum diameters, bigger water tanks and more space for the batteries of electrically powered rollers. This offers advantages in terms of quality, environmental friendliness and efficiency.” Think of a roller that compacts more surface area and is capable of working longer, in addition to not needing an operator. HAMM is at the stage of testing protoptype automated rollers on experimental test tracks. Current designs include a nine-metric-ton machine with rollers just under two meters in diameter, but with a lower overall height than standard machines at the same weight.

WHEN?
“I would estimate the industry is about 10 years away from fully autonomous machines entering the market,” Gravatt says. “Regulations and design standards around the technology are still in their infancy.  I also don’t think it will necessarily happen with a ‘light switch’ approach. More and more automated tasks or functions will begin to appear, like automated vibration control, automated braking, integrated obstacle detection and object avoidance. These types of functions begin introducing some of the hardware that will be needed for autonomous machines, like ‘steer by wire,’ or machine vision systems, for example.”

The developers at HAMM have been testing their new developments under reproducible conditions and in around-the-clock operation for weeks at a time with no driver at the wheel. In these trials, the machines autonomously complete a specified program, drive themselves to refueling points and park themselves when the test is completed. “We have now clocked up over 10,000 hours on this test track and learned a great deal about autonomous driving in the process,” explains Hans-Peter Patzner, one of the HAMM developers.

Hitachi has prototypes as well. “Looking ahead to commercialization,” a company release reads, “we are conducting verification tests of prototypes applied with the ZCore system platform for autonomous construction machinery. The prototype has no driver seat. The machinery is moved to the construction site via wireless control, without an operator on board, and then switched to the autonomous operation mode when it arrives at the intended location. During autonomous operation, the machinery automatically drives along a set route to compact the soil and automatically stops if there are any obstacles along the way. The operation status and health of the machinery are indicated with an LED panel, flashlight and buzzer to ensure that the machinery can operate in collaboration with nearby workers. This system configures the area for compaction and manages/displays the driving route of the machinery and the number of times that compaction is to be performed. The operator can check the compaction status in real-time using a tablet device connected to the machinery via wireless LAN. The travel data of the machinery is recorded using a satellite positioning system, and any areas that were missed or areas with insufficient compaction can be checked at a glance via the colour-coded display of compaction counts. The system also enables compact records to be retrieved from the Cloud to check past work and create forms based on work history.”

With prototypes out there, autonomous rollers may be like self-driving cars: a technology out there waiting for the rest of the world to catch up.