Work on the second-generation ROBART II began in mid-1982, with four general objectives:
- Make the system more modular to facilitate maintenance and upgrades.
- Employ a parallel-processing hierarchy of distributed microprocessors.
- Incorporate a more sophisticated mix of sensors in support of advanced autonomy.
- Provide a more finished look to the modular body structure.
The physical structure of ROBART II consisted of a cylindrical upper body that mated with a rectangular mobility base, with a removable electronics cage serviced via a pair of access doors. The upper housing was fashioned from a 30-inch section of 12-inch-diameter plastic pipe, with an acrylic cake cover supported by a Lazy-Susan bearing that formed the head pan axis. The initial mobility base was a functional plywood mockup fitted with a pair of A-BEC wheelchair motors, front and rear castors, and a lead-acid gel-cell battery. This temporary plywood base was soon replaced by a sturdy aluminum version with a black fiberglass shroud.
ROBART II performed essentially the same functions as its predecessor ROBART I, but with a multiprocessor architecture that enabled parallel real-time operations. Improved performance was further addressed through significantly increased perception and more precise motion control, the latter supported by phase-quadrature optical encoders attached to the drive-motor armatures.
In 1983, ROBART I and II were loaned to the Naval Surface Weapons Center (NSWC), White Oak, MD. As seen below at far left, the collision-avoidance sensor suit was upgraded from that of ROBART I with the addition of a five-element Polaroid electrostatic sonar array, and a sixth sonar transducer mounted on the head. Note the partially installed tactile-bumper strip around the bottom of the mobility base.
During this timeframe, ROBART II was featured in a number of popular books and magazines, to include National Geographic, and often performed for visitors at the NSWC Robotics Lab. One of the more impressive autonomy demonstrations was a robust person-following behavior. The new sonar-transducer array helped discriminate a near-field human target from other reflective surfaces that should be avoided. Lateral displacement of the moving target was used to proportionally adjust the robot’s heading. This algorithm could reliably follow a person through a cluttered room and transit a narrow 28-inch doorway.
Following my transfer to the Naval Ocean Systems Center (NOSC) in 1986, ROBART II became a concept-development surrogate in support of indoor robot autonomy, with initial focus on two specific technology needs. The first of these addressed the navigational shortfalls that were hindering successful implementation of a number of robotic applications requiring mobility. To enable successful traversal of congested surroundings, the robot was upgraded with additional proximity and ranging sensors for autonomous mapping, localization, collision avoidance, and navigational planning.
A cell-based map representation was chosen for the robot’s world model, with free space indicated by a cell value of zero. This approach offered the following advantages:
- The operating area was a bounded interior space where a relatively coarse grid (i.e., 3-inch resolution) could be used.
- The traversability of a square could be statistically represented and easily changed.
- Objects of unknown configuration were easily added.
- A simple Lee maze router could be used for path planning.
- Unique coding of predefined entities (e.g., doorways, recharging station) was supported.
An improved charging station, compatible with the entire ROBART series, was constructed in 1986. As with ROBART I, the homing beacon was activated by a garage-door RF link, whereupon a current-limited sense voltage was applied to the recharging contacts so a valid connection could be perceived by the robot upon docking. The charging station also detected this connection and activated the charger power supply after the mating contacts had debounced.
The second thrust was aimed at producing a robust automated security system exhibiting a high probability of detection, with the equally important ability to distinguish between actual and nuisance alarms. ROBART II was equipped with a multitude of environmental sensors that monitored system and room temperature, relative humidity, barometric pressure, ambient light, noise levels, toxic gas, smoke, and fire. Intrusion detection was addressed through the use of infrared, optical, ultrasonic, microwave, and video motion detection, as well as vibration monitoring and discriminatory hearing.
To increase the probability of detection and reduce nuisance alarms, two new sensor modalities were added to the Intelligent Security Assessment System. A line-based video-motion-detection scheme allowed a 6502-based single-board computer to digitize any three horizontal lines of a composite video image. The software would monitor each of these lines for changes indicative of motion, reconfigure line selection to focus on suspected anomalies, then compare the perceived aspect ratio of the disturbance to a human target. The second sensor upgrade was a directional head-mounted acoustic array that determined the relative bearing to the source of any impulse noise, such as breaking glass or a dropped object.
The Intelligent Security Assessment System achieved a high probability of detection through fusion of a variety of motion-detection sensor outputs, while simultaneously reducing the nuisance-alarm rate through cross-correlation. In addition to the basic smoke and gas sensors, seven different kinds of intrusion-detection modalities were ultimately employed on ROBART II, to include passive infrared, microwave, optical, vibration, acoustical, sonar, and video. Time-stamped sensor status as well as environmental conditions were displayed as shown below, and could be overlaid on live video from the robot’s camera.
In 1989, a reflexive-teleoperated driving mode (now commonly known as “guarded motion”) was added to ROBART II to test mobility behaviors that could reduce the driving burden imposed upon military operators of more simplistic man-portable UGVs. The collision-avoidance sensors, originally intended to provide an envelope of protection during autonomous transit, were called into play during teleoperation to minimize the possibility of operator error. Although ROBART II was never intended to be remotely driven by a human operator, reflexive teleoperation was one of the first UGV behaviors for which the system served as a software-developmental surrogate. The commanded speed and direction of the mobility base were servo-controlled in response to local sensor inputs to keep the robot from running into obstructions.
Also on the agenda was the pursuit of localization techniques to better support autonomous navigation in indoor environments. In the late-80s to early-90s timeframe, before the ready availability of scanning lasers and simultaneous-localization-and-mapping (SLAM) algorithms, a number of different approaches were implemented for subsequent evaluation:
Ultrasonic transponder triangulation
Polarized optical heading reference
Video image processing
Ultrasonic signature matching
Ultrasonic wall referencing
Tactile wall referencing
Realtime wall following
Doorway transit referencing
Lateral retroreflective sensing
Overhead retroreflective sensing
With the advent of ROBART III in 1992, the role of ROBART II as a concept-development surrogate came to an end. The robot continued to provide demonstrations for visitors, however, and remained on line without a power interruption from 1988 to 2002, when a support contractor disconnected its recharging station over a weekend, allowing the onboard battery to go dead.