Westrack: Putting ITS to Work
On May 30, 1996, three 69-metric ton triple-trailer/vehicle combinations (without drivers at the wheel) circled a 2.9 kilometer (km) test track at 65 km per hour at the Federal Highway Administration's (FHWA's) WesTrack facility. These laps were the first of some 3 million driverless vehicle kilometers planned for the WesTrack project over the following two years. WesTrack is a pavement-testing facility, located at the Nevada Automotive Test Center (NATC), 100 km southeast of Reno.
The primary objectives of the current research at the facility are the development of performance-related specifications for hot-mix asphalt pavement construction and the early field validation of the Strategic Highway Research Program SuperpaveTM performance-prediction models and complete mixture analysis procedures. (See related article in Public Roads, Autumn 1996, page 23.)
In addition, the project has supported the development and use of driverless vehicles to allow safer operation of the repetitive WesTrack pavement loading. The autonomous vehicle operations are an application of intelligent transportation systems (ITS), a subject of major interest to FHWA and other researchers in the highway community. Success in this development has resulted in four driverless trucks operating an average of 15 hours per day at 65 km/h to simulate 10 to 20 years of pavement loading.
The WesTrack Driverless Control System is an example of a truly operational, driverless heavy truck system. This system is addressing issues that are very similar to the real-world requirements of an automated highway system. To achieve these requirements - long-term durability, fully automatic operation, and a robust fail-safe capability - WesTrack implemented a fully redundant control system, continuous safety monitoring of the trucks and their controls, and a fully autonomous fault-resolution capability. One of the basic features is that, in the event of any failure, all of the trucks come safely to a stop without any human intervention.
Over a nine-month period, NATC designed and integrated the driverless vehicle system into four standard trucks. The resulting WesTrack control system includes 14 separate computers and more than 30,000 lines of specially written software code. Sixty-six custom circuit boards were designed and built to perform interface and steering control-related functions; the circuit boards contain more than 16,000 electronic components. Each truck has more than 700 cable connectors and 15 km of wiring related to the automatic control system and the associated instrumentation.
Aboard each truck are two computers: one controls the vehicle, while the second constantly searches for errors in the control. Either is capable of shutting down the truck. The control room has six additional computers. Of these, one serves as traffic manager. A second computer continuously check to ensure that this function is performed correctly. The other four are assigned to the individual trucks; each constantly monitors both the control system and the mechanical system of the trucks for failures that could be hazardous to the operation of the truck. The use of multiple computers is necessary in a system designed for fail-safe operations.
Truck control (speed and steering functions) is performed in the trucks. Track control and truck health checking are control room activities. Track control consists of maintaining truck spacing by adjusting the speed of the trucks. Truck health checks are performed by monitoring each truck's control systems and various mechanical parameters. The system is designed to survive a single hardware or software fault without losing control of the trucks.
Spread spectrum radio-frequency data modems are used to transfer data between the trucks and the control rooms. Each truck has a modem, and four modems are located in the control room building. A separate radio-frequency is dedicated to each of the four trucks.
The truck controls involve four separate functions within each truck; steering, throttling, transmission, and braking. The two computers in each truck are involved primarily with controlling the truck's speed. The first computer monitors the speed and adjusts the throttle, brakes, and transmission to control it. The second computer analyzes the same commands and, using the speed information obtained from separate sensors and a separate data path, continuously checks that the programmed speed is being maintained. The second computer has an independent set of hardware interfaces that allows it to stop the truck on its own.
The steering is performed entirely by the hardware and electronics without computer control. It is a fully redundant guide-wire system that, even in the event of a control system failure, will continue to operate until the truck comes to a complete stop.
Two guide wires, which are embedded in the pavement around the track, provide the transverse position signal to each truck. Each of the two redundant guide wires is powered by a separate amplifier and oscillator combination. Each amplifier and its associated oscillator are powered by a separate, uninterruptible power supply to ensure that the guide-wire system will remain operational in the event of a power failure. The use of two separate frequencies - 3,750 and 4,285 hertz - allows continuous monitoring of the signal integrity in each of the two guide wires. The signal current is approximately 100 milliamps of alternating current in each wire.
Each truck has two separate guidance antennae, which are housed in an attachment to the front bumper. Each of the antennae includes two coils: one senses the vertical magnetic field strength and the other senses the lateral magnetic field strength. The two field-strength signals "tell" the truck where it is located in relation to the guide wire. The field coil for sensing the vertical field is built into a four-layer circuit board. The lateral field coil is generated with the strips that project upward from each of two redundant vertical antenna boards. Each of the antenna sets, which connect the antennae to the guide-wire control system, has two cables. These cables are routed via separate paths through the truck to reduce the likelihood that both cables would be broken at the same time.
The truck is maneuvered by a steering-motor feedback assembly, which is mounted on top of the power-steering gear box. The motor is a two-phase, 0.9-degree-per-step stepping motor that has redundant windings. The motor can be operated by either of the two sets of windings. The motor is connected to the steering-wheel input shaft by a heavy-duty cog belt. Position sensing of the steering system is performed by a set of three potentiometers in the assembly on the end of the stepping-motor assembly. Three separate feedback potentiometers, three connectors, and three cables are independently connected to the control system. The control system contains a voting circuit that chooses the feedback potentiometer with the middle value of the three.
The steering control unit is a four-layer, single-circuit board that is approximately 250 by 300 millimeters (mm) in size. The unit includes three guide-wire receiving circuits: two are used for steering the truck, while the third measures the position of the last trailer in the triple string to monitor the way the trailers are tracking. (The last trailer is also equipped with a pair of antennae.) The circuit board also includes the position feedback voting circuit and two separate motor drive circuits for the two sets of motor-driving windings.
Each of the two motor drive circuits is supplied from a separate, electrically isolated battery. This ensures the truck will remain steerable even if the vehicle's main electrical system suffers a catastrophic short circuit.
The steering control loop updates the truck's steering system approximately 10 times per second. In comparison, because a driver has "look ahead" capability, a driver updates steering approximately once per second while driving a similar Class 8 truck.
This highly redundant steering system has proved satisfactory, and few mechanical failures have occurred among the guide-wire and steering components. The steering system has continued to operate, identify failures, and - when necessary - shut down the overall system.
Throttle and Automatic Transmission Controls
The next group of truck controls are those for the throttle and automatic transmission. Both the Detroit Diesel Series 60 engine and the Twin Disc automatic transmission have electronic control units that govern their respective operations.
In a conventional installation, both the automatic transmission and the engine controller would have separate position sensors on the throttle pedal. That is, the automatic transmission's sensor would tell the transmission when to shift gears, while the engine controller's sensor would independently dictate the engine speed. To integrate these two electronically controlled systems with the control computer, a system of isolating and scaling amplifiers was built to accept a single throttle position sensor on the pedal. This sensor operates in manual mode or in automatic mode, and it supplies a single command voltage to the computer. The computer, in turn, gives directions to both the transmission and the engine.
The other design feature of the transmission control is gear selection. The shift control for the Twin Disc transmission is electronic, and it was necessary to provide electronic isolation between the two systems so that unwanted interference would not occur. This system operates and conveys commands in both the automatic and manual operating modes.
|Vehicle Dynamics Data
A sophisticated data acquisition capability has been built into the control system to capture very high-rate, high-resolution, vehicle dynamics data when the trucks pass over strain gauges located in the pavement. This additional vehicle dynamic data is used at WesTrack to ensure that road roughness and vehicle dynamics in one test section do not affect an adjacent section and to closely tie the dynamics force variations on the axles to the pavement strains. Within an ITS research environment, such an on-board data acquisition system would allow on-board logging and diagnosis of all vehicle health parameters that would normally be checked by a driver. The WesTrack 160-channel data package is checked every half second as an integral part of the on-board automation, and today's advanced vehicle fleets could take advantage of this capability for monitoring maintenance requirements. This measurement of truck health and axle loads is also applicable to weigh-in-motion and moving vehicle inspections for commercial vehicle operations.
The final set of truck controls (for the brakes) performs three separate functions. The first is for routine stops and is performed by computer control of a conventional air-brake system. A proportional valve provides an air pressure that is proportional to an analog command from the control system. In addition to the proportional valve, a parallel solenoid valve conveys the full-system air pressure to the brakes in the event of a detected failure. This solenoid valve is controlled by the secondary computer in the truck. The solenoid valve is normally open, and it must be continuously energized to keep the valve closed and to prevent the brakes from being applied. In the event of a loss of power to the control system, the solenoid valve opens, and the brakes are applied. This system is also used for several emergency braking scenarios.
The second function of the braking interface is to control the whip of the triple-trailers. Steering adjustments, if severe and rapid, can be amplified through the length of the trailers and potentially result in a loss of control. Such adjustments might occur in the event of a steering-tire blow-out. To control whip, the brakes on the last axle of the third trailer are applied by a solenoid valve located on the third trailer and are controlled by one of the computers in the truck.
The final function of the brake controls is anti-lock braking. An anti-lock braking system was included in the trucks to ensure stability of the trucks under emergency braking conditions. It operates on all of the truck axles except the steering axles, where the brakes have been disconnected. To date, the anti-lock braking system has performed excellently. The system was tested in many hard-braking modes, including full air pressure to the service brakes through the backup solenoid valve. This last test produced a very short stop, but, as in all of the other tests, the tires did not lock.
Truck Health Data
To be sure that the trucks are operating safely and without mechanical or electrical problems, the health of the systems in each truck is checked and evaluated every half second. Data are acquired from four sources aboard the truck and then transmitted to the control room. The Detroit Diesel electronic control (DDEC) from the Series 60 engines provides engine data and some vehicle data over an SAE J1708 data bus. The system monitors this bus and extracts data of interest. Additional pressure, temperature, and voltage sensors are installed in the vehicle to measure truck parameters, such as cab temperature and power steering fluid temperature, that are not monitored by the DDEC III system. Thirdly, a tire-monitoring system reports temperatures and pressures in the tires. Finally, one of the trucks is equipped with a set of accelerometers and strain gauges to measure the forces on the axles of the truck. These accelerometers and strain gauges have very severe signal-processing requirements to maintain precise timing alignment with other pavement signals.
The data from the four sources are accumulated by the two computers in each truck, combined with the vehicle-control information, and sent back to the control room over the spread spectrum radio-frequency modem links. The data rate is 9600 baud. In the control room, these data are captured by a data acquisition computer and are downloaded once a week to CD-ROMs for permanent storage.
Track control and safety monitoring are assigned to a pair of computers in the control room. Traffic is managed by referencing very-high-resolution odometer positions that the trucks report twice per second to the control room. The trucks also report their odometer positions once per lap when they pass over a radio beacon located on the track surface. Truck spacing control is maintained by adjusting the directed speed of the individual trucks so that they remain equally spaced around the track. This control system resides in the first of the two computers; it is backed up by the second, which uses that uses differential global positioning system (DGPS) data reported by the trucks once per second. The second computer compares the DGPS data to the odometer-generated positions and uses both to continuously verify that the spacing tolerance between trucks has not been violated.
In addition to controlling the truck spacing, the traffic managment computer periodically, or on manual request, commands the vehicle equipped with strain gauges and accelerometers to take high-resolution dynamics data as it passes strain gauges in the pavement at specific locations along the track. This computer also provides a graphical display of the steering deviation of the trucks relative to the guide wire. For each truck, the positions of the tractor and the last trailer can be displayed so that the tracking performance of the triple-trailer combination can be examined as it circles the track.
The safety-monitoring computer verifies the truck spacing from the DGPS data and interfaces with a lock board that contains safety keys for all of the trucks. The truck control authority is passed to the control system only when the safety key has been inserted and locked. This provides a hardware method of removing a truck from automatic operation. When the truck is in maintenance, the safety key for that truck is removed from the lock board and commands cannot be sent to that truck.
In addition to the traffic management and safety computers, vehicle-monitoring computers are located in the control room. These computers continuously verify that all truck and control systems are operating properly. The screens of the vehicle-monitoring computers are displayed in a "red-yellow-green" format. If a parameter display is green, it indicates that the system is operating well within the tolerances set for that system. If a parameter is yellow, the system is getting close to the upper or lower limit. If the parameter is red, all trucks are stopped automatically, and the control room operator gets an instant, visual reading of the problem.
Lessons Learned From the WesTrack Driverless Trucks
As of May 1997, the four trucks have covered more than 700,000 km in their driverless mode, and the WesTrack researchers have already learned much that will be useful in future driverless truck system design and presumably in automated highway development.
Multiple Failures Will Occur
Redundancy was designed into the WesTrack driverless vehicle system, and there are few potential sites for single-point failures within the system. Where single-point failures are possible - for example, the belt between the stepping motor and steering box - the system is significantly overdesigned. As a result, it was assumed, throughout the design of the system, that it was improbable that two different failures would occur at the same time.
However, Murphy's Laws work, and we learned that multiple failures can occur at the same time. A steering motor failure and a brake system (air line) failure occurred at the same time on one of the WesTrack trucks, resulting in a "slow-speed track exit," and the truck was slightly damaged. A lesson was learned: The system must be able to control the vehicle in multiple failure scenarios.
Vehicle Wander Is Critical
Both WesTrack and other driverless vehicle systems have proved to be extremely repeatable in terms of lateral position control. For example, without the wander control built into the driverless vehicle system, the precise repeatability of the truck paths (less than 2 mm of variation from center per pass) would reproduce the rain grooves of the tire tread in the asphalt.
The concentrated wheel-path loading accelerates pavement wear. One concept for an Automated Highway System dedicates vehicle lanes, similar to current, dedicated high-occupancy-vehicle lanes, for driverless vehicle operation. Given traffic with extremely repeatable guidance systems, this concentrated loading will accelerate the pavement wear and quickly establish rutting or fatigue cracking within the narrow wheel-loading area.
Building in and controlling wander is a lesson learned and a consideration for future designs. WesTrack trucks were initially designed to operate "on the wire," centered over the embedded guide wire, but to reproduce driver variation, they have since been programmed to operate up to 380 mm "off the wire." The new capability allows the track operators to generate a distributed traffic pattern that approaches the wander patterns found on typical open highways. Truck wander and the establishment of a different wander algorithm in each truck may be an important consideration in future designs. A gaussian wheel-path distribution with a standard deviation of 100 mm is the minimal recommendation for an automated highway scenario.
Guidance Sensors Can Move
Two lessons were learned about the WesTrack "wire in the road" design. At the time of construction, the wire was installed in a conduit under the asphalt.
The first lesson learned is that it is nearly impossible to keep a conduit or any other sensor device perfectly aligned underneath an asphalt paving machine. Slight jogs were introduced in the wire during paving and the tightly controlled trucks followed them. Instead of maintaining a straight path down a tangent section of the track, the trucks executed a series of extremely repeatable left and right steering variations. This was eventually corrected by cutting a groove in the top of the hot-mix asphalt (HMA) layer and placing a new straight wire at the surface.
The other lesson learned is that the wire can, and does, move within the conduit. The direction of WesTrack traffic around the track is always in the same direction - counterclockwise. The wire in the conduit also moved counterclockwise and was displaced approximately 200 mm longitudinally within the first 120,000 km of vehicle operation. This tensioned the wire to the point that service loops were pulled tight, and the wires broke. This problem was also corrected by moving the wire to the surface. The lesson learned is that asphalt is a flexible pavement, and given a constant direction of traffic - especially heavy truck traffic - subpavement items will migrate.
Slight Lateral Positions Make a Big Difference
When the wire was installed in the spiral that transitions from the straight tangent section to the 18 percent superelevated curve, it sometimes was placed as much as 25 mm laterally from the theoretical centerline on which it should have been placed. This created an undesirable lateral sway when the trucks exited the curve. The lesson learned is that alignment of a curve transition is critical, and the position of guidance sensors 25 mm left or right of the theoretical path can establish undesirable lateral sway dynamics. Given that sensors can move in HMA, this sensitivity to exact curvature is critical to future ITS designs.
Truck Air-Conditioning Systems Are Not Designed to Operate 24 Hours per Day
The control computers for the WesTrack trucks were located in the sleepers of the trucks, and the air conditioning was directed through ducts to keep the computers operating as in an office environment. During the summer, this required the air-conditioning systems to operate up to 22 hours per day, and current, standard truck air-conditioning systems are not designed for such extended, continual operation.
Steering Lash Increases With Gearbox Wear
At the current 700,000-km point, the track operators have noted the steering lash, the amount of steering wheel play, is increasing as the steering gearboxes on the trucks wear. By design, truck steering gear boxes are designed with lash because overly responsive steering on a large truck is not desirable. Steering wear must be accounted for during steering control design.
Robust Communications Are Necessary
Stable communications systems for longitudinal control are critical. Telemetry communications need to be robust. Loss of communications for as little as one second can cause problems. Of all the redundant subsystems in the WesTrack design, longitudinal control is considered the most critical.
Steering Tire Selection Is Important for Large Trucks
A natural tendency of radial tires is to go straight. This effect is called self-aligning torque. The higher the self-aligning torque, the greater the tendency to go straight.
Initially, steering tires with relatively high self-aligning torque were selected for the steering axles of the WesTrack trucks. As a result of the increased steering demands on the tires negotiating the curves of the track, the tires were wearing high on their shoulders. After the first 100,000 km on each truck, new tires with a lower self-aligning torque were put on trucks. This enabled the trucks to steer with less frequent correction through the curves, and it reduced the demand on the steering motor. Steering tire selection is an important consideration for future ITS vehicle designs.
The WesTrack driverless vehicle program has been successful and is an operational real-world example of heavy-vehicle driverless operation. It has addressed the practical requirements for control system redundancy and for continuous, vehicle safety and vehicle-health monitoring that will be part of practical automated highway systems. WesTrack has been operating with four driverless triple-trailer combinations since June 1996. More than 700,000 kilometers of driverless operation have been recorded to date. Fail-safe controls have been implemented, and a practical level of reliability has been achieved. A system of continuous, vehicle safety monitoring has also been demonstrated.
Colin Ashmore is the co-program manager for WesTrack and is a test engineer with the Nevada Automotive Test Center, the prime contractor for WesTrack. He joined NATC in 1987. He is responsible for various vehicle and pavement/vehicle interface engineering projects at NATC. He received a bachelor's degree in agricultural engineering from Texas A&M University and a master's degree in mechanical engineering from Auburn University.
Terry M. Mitchell is a research materials engineer in the Pavement Performance Division at the Turner-Fairbank Highway Research Center in McLean, Va. He joined FHWA in 1971. He received bachelor's degrees in aeronautical engineering and mathematics and a doctorate in nuclear engineering from the University of Michigan.