Robotics Homepage

Robart I (1980-1985)

One of the first behavior-based autonomous robots, ROBART I was my thesis project at the Naval Postgraduate School in Monterey, CA.  This robotic security system was fully autonomous with no RF link or operator control unit (OCU).  If desired, a small trouble-shooting panel with four toggle switches and a pushbutton for entry of two four-bit nibbles could be used to invoke various diagnostic routines.  Output responses were conveyed via speech synthesis using the National Semiconductor DigiTalker board, probably the first instantiation of speech synthesis on a mobile robot.

 

ROBART I during filming for ABC World News Tonight at the Naval Postgraduate School, Monterey, CA, 1982.
ROBART I during filming for ABC World News Tonight at the Naval Postgraduate School, Monterey, CA, 1982.

The robot’s mission was to patrol a home environment, following either a random or somewhat predetermined pattern from room to room, checking for unwanted conditions such as fire, smoke, flooding, or intrusion.  The security application was chosen because it demonstrated performance of a useful function that did not require an end-effector or vision system.  Passive infrared (PIR), optical, ultrasonic, and hearing sensors were used to detect intruder motion, with other sensors monitoring for vibration, fire, smoke, toxic gas, and flooding.

Intruder-detection sensors on ROBART I included a PIR motion sensor atop the head, three optical motion sensors above the beacon-tracking collimating tubes on the front face, two microphones for ears, and a vibration sensor.
Intruder-detection sensors on ROBART I included a PIR motion sensor atop the head, three optical motion sensors above the beacon-tracking collimating tubes on the front face, two microphones for ears, and a vibration sensor.

For automatic recharging, the beacon-tracking system on ROBART I employed a photocell array to keep the head pointed at the charging station so the head pan angle could be used to position the steerable front-wheel for homing.  The recharger beacon was initially located by panning the head through its full extent of travel while digitizing the perceived light level as a function of the pan-axis encoder value.  The optical beacon on the charger was then activated using a garage-door-type RF link, after which another full scan was performed.  Subtracting these two linear arrays yielded a peak intensity differential corresponding to the beacon’s relative bearing. 

This early 1981 version of ROBART I has activated the optical beacon on its recharging station and is preparing to dock.  Analog photocell output is seen as a function of head scan angle.
This early 1981 version of ROBART I has activated the optical beacon on its recharging station and is preparing to dock. Analog photocell output is seen as a function of head scan angle.

ROBART I’s autonomous-navigation scheme featured a layered hierarchy of behaviors that looked ahead for a clear path (high-level), reactively avoided nearby obstacles (intermediate-level), and responded to actual impacts (low-level).  A basic tenet of this strategy was the ability of certain high-level deliberative behaviors to influence or even inhibit the intermediate and low-level reactive behaviors, such as disabling collision avoidance just prior to docking with the recharging station.

LEVEL

BEHAVIOR

RESULTING ACTION

High

Radar
Survey
Dock

Look ahead for potential obstacles
Look for opening in forward hemisphere
Home on recharging-station beacon

Intermediate

Wander
Wall Hugging

Seek clear path along new heading
Follow adjacent wall in close proximity

Low

Proximity Reaction
Impact Reaction

Veer away from close proximity
Veer away from physical contact

In support of the high-level Radar and Survey behaviors, a custom-designed near-infrared proximity sensor on the head provided reliable detection of diffuse wall surfaces for ranges out to about 6 feet.  Lateral resolution was sufficient to reliably locate the edge of an open doorway to within 1 inch of arc at 5 feet.  No distance-measurement capability was provided, other than any detected target was somewhere within the effective range of approximately 6 feet.  This sensor, which could be oriented up to 100 degrees either side of centerline by panning the head, was extremely useful in locating open doors and clear zones for travel.

ROBART I was equipped with a variety of proximity sensors, feeler probes, tactile bumpers, and an LM-1812-based sonar for obstacle detection and avoidance
ROBART I was equipped with a variety of proximity sensors, feeler probes, tactile bumpers, and an LM-1812-based sonar for obstacle detection and avoidance

The previously mentioned layered-behavior structure enabled the robot’s hallway-navigation scheme, with the recharging-station beacon being suitably positioned to assist the robot in locating the hallway.  Once in the hall, the robot would move parallel to the walls in a reflexive fashion, guided by ten short-range near-infrared proximity sensors on the front and sides, plus the previously discussed head-mounted proximity scanner.  Orientation in the hallway could be determined from which direction afforded a view of the beacon.  With an a priori linked-list representation of where the rooms were situated relative to this hallway, the robot could proceed to any given room by counting off the correct number of open doorways on the appropriate side.

General floor plan of our residence on Rickett’s Road in Monterey, CA, showing the recharging station at the right-most end of the hallway.
General floor plan of our residence on Rickett’s Road in Monterey, CA, showing the recharging station at the right-most end of the hallway.

The side-looking proximity sensors, with maximum range set to about 16 inches, kept the robot centered while transiting the 3-foot-wide hallway.  The room-entry behavior was tuned by trial-and-error adjustment of head scan angle θ for doorway detection.  The forward-looking proximity sensors enabled last minute heading adjustments as needed to ensure collision-free doorway penetration.  All proximity sensors prevented the robot from entering congested spaces, as for example through the open doorway of a small closet.

Turn initiation was tuned by varying head scan angle θ.  For θ2, the head-mounted proximity sensor has just detected the door opening at wall position B, which will result in the optimal arc shown by the dashed line.
Turn initiation was tuned by varying head scan angle θ. For θ2, the head-mounted proximity sensor has just detected the door opening at wall position B, which will result in the optimal arc shown by the dashed line.

In 1983, ROBART I was loaned to the Naval Surface Weapons Center (NSWC), White Oak, MD, entrusted to the watchful care of MIT AI Lab co-op student Anita Flynn, a later pioneer in the field of micro-robotics.  In late 1985, ROBART I was shipped to Vancouver, BC, for yearlong exhibition in the Design 2000 pavilion at the EXPO ’86 world’s fair. 

Related Posts

Site Visits

3,001,967