A dynamical systems approach to task-level system integration used to plan and control autonomous vehicle motion

摘要: Abstract Autonomous systems with multiple senory and effector modules face the problem of coordinating these components while fulfilling tasks such as moving towards a goal avoiding sensed obstacles. We propose set organizational principles for dealing this problem. The ideas are (a) to plan in terms task-related variables that abstract from degrees freedom peripheral sensor coordinates but succinctly capture behavioral constraints; (b) generate time courses behavior through dynamical system planning variables. Task constraints, targets be reached, obstables avoided, etc. expressed parts dynamics principled fashion invoking concepts qualitative theory systems. System integration is possible sense all information provided by various sensory required becomes part dynamics. Compression achieved second layer which relative strengths different contributions governed competitive separate convergent information, integrated selecting representative, non-redundant kept invariant. capability perform stable planning, make decisions, integrate redundant well complementary demonstrated software simulations. These include simulation control errors on both side.