universal printed circuit cards. The heart of the control system is the soft- 

 ware program stored presently as firmware in 6 kbytes of UV-erasable program- 

 mable read-only memory (PROM) and 2 kbytes of random-access memory (RAM) in 

 the vehicle microprocessor and 24 kbytes of RAM in the topside Intecolor 8051 

 colorgraphic display terminal with associated mini-floppy disk drives and key- 

 board. The primary programming language is PL/M, a microprocessor-compatible 

 subset of PL-1. The topside terminal and the control system on the submer- 

 sible are connected by an umbilical cable which is disconnected before launch. 

 The use of a lightweight, deployable, low-drag, single-fiber strand as a 

 real-time high-bandwidth data link to the submersible after launch is being 

 investigated. This would allow the use of relatively inexpensive data sensors 

 such as TV cameras and sonar systems without the disadvantage of high cable 

 drag. The onboard microprocessor is used to compare programmed run time, 

 heading, depth, and run sequence input data with measured data originating 

 from a clock, fl uxgate-updated gyrocompass, depth sensor, and run sequence 

 pointer, respectively. The microprocessor generates digital 8-bit error 

 signals between the programmed and measured values and issues them to the 

 appropriate linear motor controller. A trajectory design program allows the 

 operator to choose a series of preprogrammed tracks such as taxi (out and 

 return), figure-eight, parallel path search patterns, a square, or a hexagon. 

 If he desires, the operator also can generate new patterns for the vehicle to 

 execute. The total trajectory can be displayed graphically to help the 

 operator in visualizing the chosen trajectory. 



1.2 SUPERVISORY CONTROL 



The basic software and hardware architecture for the vehicle described in 

 section 1.1 is that of supervisory control by means of a two-computer con- 

 figuration. This control concept is shown schematically in fig 2. Pioneering 

 work on this concept was conducted by Ferrell and Sheridan (ref 4). Under ONR 

 sponsorship, Dr. Sheridan is currently conducting basic research on super- 

 visory control at MIT, and with the technical cooperation of NOSC is inves- 

 tigating the application of these concepts to advanced undersea teleoperator 

 systems under supervisory control. The human operator communicates with a 

 given teleoperator system such as a vehicle through the intermediary of a 

 computer. The operator is responsible for higher-level functions such as 

 specifying a trajectory and receiving status information from the remote 

 computer on the submersible. The remote computer, meanwhile, controls 

 continuously the sensors and thrusters of the vehicle by breaking the task 

 commands down to primitive functions such as increasing the current to the dc 

 motor which drives the left stern-mounted thrusters. In general, the human 

 operator communicates continuously with the local computer which, in turn, 

 commmunicates intermittently at a low data rate with the remote computer that 

 continuously completes a previously given set of instructions. 



Although the main reason to adopt a supervisory-controlled architecture is 

 to be able to use lower communication data rates, experiments at MIT (ref 5 

 and 6) have shown that such a manipulator is more efficient and effective than 

 switch rate control, joystick rate control, and master-slave position control. 

 Since the experiments were performed under ideal conditions, it is predicted 

 that supervisory control in a two-computer configuration will show to even 

 more advantage when used with degraded sensor or control loops; eg, time 

 delays, limited bandwidth, etc. Operating a remotely controlled manipulator 

 on a free-swimming untethered submersible with a limited data transfer through 



