CPS196 - Fall 1999

Subsumption

Reading: Brooks

Notes: By Ralph Bodenner

LegOS on one NT machine in C:\legos, rotation sensor patch almost ready (Luis).

Group 2 (web compiler):

• rubber band drive failed, couldn't move wheels fast, since they slipped
• single back wheel turned well, but stayed stuck turning
• surface-dependent: turns wider or tighter depending on friction
• to turn, reverse works best on the opposite wheel, then brake

• bottom not secured, it falls out
• applied brake, ran on smooth surfaces
• using rotation sensor would have been best way to make motion friction-independent

• used tank treads, (somewhat) consistent action on different surfaces
• seemed to go more slowly on each trial, may need new batteries

• stopped after each square side
• geared, large drive wheels in front, Lego piece resting on surface in rear for balance and turning
• sleep() caused a problem for no good reason, robot would freeze
• must destroy each thread individually
• adding do-nothing loop at the end lets button-press events stay active so you can kill the robot with a button or a sensor event
• does reset_rom() work?

• stuck in loop at end, must remove batteries after each run, then re-upload program

• passing a message down the hall
• bird flocking
• battles between robots
• laser tag
• robots generally working together

• layers of control: sensors to layer 0 control to actuators (motors, etc.)
• upper levels subsume lower levels (overrule lower levels), take control of actuators
• good for debugging, but why not just use the highest level?
• modules communicate with each other; for example (Fig. 5) forward activates/deactivates turn