Goals:

	Estimate the complexity of on-road autonomous driving task
	Compare symbolic rule-based approaches with iconic map-based search methods
	Integrate rule-based and map-based methods within each node of the 4D/RCS architecture

Approach:

Rule-based

1. Use DOT manual* as guide to task definition and to delimit scope
   
2. Define a hierarchical taxonomy of driving tasks and express in terms of state graphs
3. Identify all stimulus events as WM variables
4. Define a 4D/RCS architecture of Finite State Automatas and implement in ControlShell**
5. Debug, expand, and analyze behaviors using a driving simulator
6. Develop complexity analysis measures for design complexity, system complexity, and resource requirements

Graph based

1. Cost evaluation guides search through state space
2. Use to encourage/discourage behaviors, e.g.,
   
   

   	Multiple lane changes discouraged
   	Erratic velocity changes discouraged

   
   
3. Dynamic feature layer uses cost to prevent collisions and traffic signal violations
4. A provably optimal graph planning algorithm has been developed that is capable of construction superior plans to traditional grid-based approaches
5. The system is able to implement both hard and soft constraints on plans through the use of knowledge base layers
6. A simulation environment is being developed to test and evaluate this approach

Static Road Rule Knowledge Layer Reads road center line and database id from OTBSaf Obtains additional static information from relational database Places hard constraints on road driving behavior Controls lane changes, speed, static road signals, etc.
A Priori Knowledge Layer Visibility graph analysis to determine additional nodes Nodes used for emergency maneuvers Detects small features that may be missed by traditional grid
Dynamic feature layer Works with traffic, dynamic traffic signals Used only in cost evaluation
Planning in space and time. Graph includes time dimension to allow velocity planning and to include dynamic objects, such as other vehicles	Movie (no sound in movie clip) description of the movie clip: first frame: Planning nodes shown along road lanes 2nd frame: nodes created at salient features in terrain map 3rd frame: bridge picture: moving traffic on a bridge traffic light: traffic light with changing states 4th frame: car passing another moving vehicle and avoiding obstacles

*McKnight, A.J, & Adams, B.B. (1970). Driver Education Task Analysis: Volume 1. Task Descriptions (HumRRO Technical Report 70-103)