MARK I Robot

FILM OF MARK I ROBOT (16 nun COLOR SILENT - 5 MINUTES)

Dr . John K. Dixon , Ms . Susan A. Bouchard, Mr. Wi ll iam G. Kennedy, Dr . James R. Slagle

NuvaJ lk:;r-,arch Laboratory, Washington, D.C.

envisioned for the underwater robot, collecting geological samples.

First, the robot calibrates its tactile sensor by touching a straight edge of known location. This tells the robot exactly where the tip of the touch sensor is, in relation to the hand of the robot.

Next it searches and finds a block. Then it touches the block on four sides to determine size, shape, and orientation.

Finally, having determined that the block has acceptable size and shape, the robot discards its touch sensor, picks up the block and places it into a basket.

The robot moves quite slowly in the film. It vas programmed to move at 20% of maximum speed for clear viewing. The robot al3o hesitates between moves. This is caused by conununication delay between a large computer and a small one.

ABSTRACT

This film shows an autonomous robot in action. 'I'he goal of this NRL research effort is to estab­ lish a technical base for an autonomous under­ water robot. It is expected that such a robot would be able t o do only s imple tasks, but would b.e able to do them one or tvo orders of magnitude cheaper than other methods. (The economy is mainly because it i s not necessary for a ship to standby on the surface).

The robot vehicle would be progranuned at a base and proceed to its assigned destination on the floor of the ocean. It would then carry out its vork autonomously, without even a commun­

ications link.

This project is concerned vith softwar e rather than ha rdwar e. The Mark I Robot is a system de­ signed to operate in the laboratory as a research tool. It is a simplified r obot , a first cut at providing the minimal intelligence necessary for an autonomous underwater r obot . The robot per­ formed as expected, successfully carrying out the

task it vas assigned.

The main issues addressed are:

1. Sensor data collection. A touch sensor (a microsvitch) held in the robots hand vas chosen because it works in muddy water, and because it provides three-dimens ional data whi ch is relatively easy to process, Algorithms for searching (to find an object) and tracking (to explore the shape of an object) vere developed and tested.
2. Object recognition. That is, the processing of raw sensor data into a form useful to the robot. In the Mark I robot a solid object model (a rectangular prism or block) is fitted by a kind of relaxation procedure to the data points collected by the touch sensor. This simple recognition scheme proved adequate t o allow the robot to recognize and manipulate objects.
3. System Integration. That is, putting s com­ plete robot together, and testing it in action t o see hov vell the various modules vork together and where the r oueh spots are.

The film shows the robot locating an object and then putting it into a basket . This is the laboratory vers ion of the simpliest task