In the United States, trucking companies and the Army are both developing systems to automate moving groups of trucks. While trucking companies are mostly interested in “platoons” of trucks drafting off of each other to save fuel, the Army wants its “convoyed” trucks to be hundreds of meters apart to improve their chances of surviving an enemy attack.
Battlefield challenges
While the biggest danger for platoons of commercial trucks is crashing, military convoys can be threatened by attacks with improvised explosive devices (IEDs) or rocket-propelled grenades.
Civilian truck drivers also benefit from a robust infrastructure, said Bernard Theisen, division chief for Ground Vehicle Robotics at the U.S. Army’s Ground Vehicles Systems Center (GVSC). For example, nearly all platooning trucks are limited to using roads and highways that have been mapped at centimeter-level resolution with lidar, can communicate over 3G or 4G networks, and have excellent GNSS signals. “I would love to have all that information,” Theisen said, “every time I send a robotic convoy vehicle out there.”
By contrast, the military must design a system that assumes “no comms, no prior data, and no infrastructure,” Theisen explained. “Sometimes a bridge that used to be there has been blown up or we may have put a new bridge across the water overnight. A building that was there yesterday got blown up and is now blocking the road. You cannot pre-plan that in the map and expect it not to change.”
Nevertheless, the civilian and military efforts share some challenges, Theisen acknowledged, including “perceiving the world, understanding it, processing the data, and making the right decisions.” Unlike robots, humans are very good at coping with the unexpected. “You can only train a robot so much, there will always be situations that it does not know how to handle.”
In a military convoy, every fourth or fifth truck may have a mounted gun to protect the convoy. The convoy will typically include one or more ambulances, wreckers and fuel tankers. “It is a different application than for Amazon or FedEx to send a couple of trucks down the highway,” Theisen said.
Leader-follower
In leader-follower applications, GVSC installs the same hardware on all its trucks. “This facilitates software maintenance, because you don’t need to have different versions,” Theisen said. If the convoy’s leader is disabled from a mechanical or battlefield issue, it is easy for a soldier on the next truck to authorize his truck to take over as the convoy’s new leader. “We have also created cases where the leader takes that road months ahead of the followers,” said Alberto Lacaze, co-founder and president of Robotic Research. “So, the leader does not need to be a part of the convoy.”
Rough terrain doesn’t affect navigation, except that in hilly terrain trucks might have more side-to-side drift than in a flat area. “We often use a three-axis IMU [inertial measurement unit] instead of a two-axis IMU, which might be all you need for a commercial application on flat roads,” Theisen said.
“The commercial problem is almost like carrying a group of trailers that are not mechanically connected,” Lacaze said. It is crucial to be able to tie in the vehicles’ low-level controls so that they maintain a very short separation. If those vehicles were trying too hard to maintain those very close distances by frequently accelerating and decelerating, the fuel-savings advantages from drafting would go away. By contrast, for military applications the exact distance between the trucks doesn’t matter much, but their side-to-side error does. “You would like all vehicles to be driving within one tire width of the lead vehicle’s tracks,” Lacaze said. “That has many advantages — for example, if that road has been demined.”
While commercial and military software largely overlap, their sensor requirements are fundamentally different. “Most commercial vehicles are not checking to see whether there is a crater in the middle of the road,” Lacaze said. Military vehicles need to detect such damage to the infrastructure and respond quickly.
Still, the military is interested in “the gigantic amount of mapping of the available infrastructure” being done by private companies, Lacaze said, because most military convoys are not in war-torn areas, but delivering materiel to bases in areas with some infrastructure.
Robotic modes
GVSC purchases commercial off-the-shelf (COTS) systems and integrates them into its trucks, Theisen explained, producing five robotic modes:
- Warning, which consists of “idiot lights” and buzzers alerting human drivers that, for example, they are straying out of a lane or are about to hit something
- Driver assist mode, which helps drivers brake, accelerate and steer
- Teleoperation, which consists of driving the truck from a remote location
- Waypoint navigation, which uses a GNSS waypoint path that can either be pre-programmed or pre-driven and then replayed
- Leader-follower, in which the first vehicle leads and potentially any number of vehicles follow.
Regarding the driver assist mode, Theisen pointed out that “all these features are very common in high-end cars and you are seeing them coming into many Class 8 trucks. We don’t do any development in the Army from that standpoint.” Regarding the leader-follower mode, the first truck can be driven in any of the other four modes.
GVSC is the lead system integrator for 30 robotic palletized loading systems (PLSs) that the Army has at Fort Polk, Louisiana, and another 30 at Fort Sill, Oklahoma. Nevertheless, a human driver usually leads the robotic convoy. The driver determines the best route, assesses the situation, and is normally followed by three unmanned systems. “That is why we call our system semi-autonomous,” Theisen said.
The role of GNSS
The Olli shuttle, equipped with Robotic Research’s AutoDrive kit, is deployed on busy boardwalks, campuses and public roads. (Photo: Robotic Research)
For both commercial platooning and military convoying, GNSS signals are used for redundancy but not as the primary source of measurement of the distance between the trucks. “None of the systems that we have deployed on the commercial side — for example, with Local Motors vehicles (the Olli shuttle) — rely on GNSS,” Lacaze said, though they will use those signals if available. The high accuracy of their inertial systems make it hard to spoof or jam GNSS receivers, because the system would detect any changes in the GNSS solution and the vehicles would continue running on inertial navigation if the GNSS signal were jammed.
“We assume that we will not have GNSS information because sometimes we are jamming ourselves or are being jammed or the enemy could be spoofing us,” Theisen said. Most of GVSC’s systems use “nav boxes” sold by multiple vendors that enable vehicles to navigate for long periods without GNSS signals. They typically combine one or two GNSS receivers, an IMU or several smaller ones, a combination of wheel encoders or ground sensors to determine ground speed, and a digital compass.
GVSC’s trucks also use lidar to generate voxel maps of their current surroundings, and then share them with the other trucks in the convoy. Each vehicle tracks the vehicle in front of it and can just follow it, if it has insufficient position information or good visual cues.
GVSC looks for the highest possible GNSS accuracy, whether using civilian GNSS receivers or military Selective Availability Anti-Spoofing Module (SAASM) units. “We also take advantage of the future M-code,” Theisen said. “We do have capabilities that the civilian marketplace does not have.”
Remaining obstacles
The remaining bottleneck in the development and implementation of convoys of autonomous military vehicles is the approval process, Lacaze said. “Currently, if we make changes to the autonomy systems, the testing parts of the government are asking us to drive hundreds of thousands of miles before providing approvals. It is still a challenge to figure out at what point these vehicles are safe enough to provide to the soldiers and what the cost of doing so is.”
For these systems to take off, better processors, sensors (cameras, radars and lidars) and algorithms are required, Theisen said. “There is way more sensor data that you can collect and process in real time.”
Featured photo, provided by Robotic Research: Army convoys can stretch for miles. The U.S. Army’s Autonomous Ground Resupply trucks shown here are connected with Robotic Research’s autonomous technology.