MIT Department of Electrical Engineering & Computer Science

E E C S

EECS Fall 1998 Catalogue Supplement

6.291 Dynamic Vision for Intelligent Vehicles (H)

MW 11-12:30, 5-134
Visiting Prof. Ernst Dickmanns
For info contact Prof. Sanjoy Mitter, 35-308, x2160
Prerequisite: see below
3-0-9

Prerequisites: basic knowledge in differential equations and in linear algebra. Some background knowledge in video imaging and systems dynamics may be advantageous, but a brief introduction will be given when needed.

1. Introduction: General background: biological vs. technical bases. What is meant by "intelligent?" Types of intelligent vehicles: autonomous vs. assistant systems. Modeling of relevant aspects of the world; visual environment on earth: lighting conditions, dependence on time and location. Structures, objects, buildings; vehicles and living beings.

2. Processes: Differential and integral representations; specific properties of spatial and temporal domains. Constraints to temporal changes; differential equations, state and control variables, parameters. Multiple scales, and local integrals. Vehicle dynamics: longitudinal and lateral degrees of freedom, controls. Sensors and actuators as interfaces to the real world.

3. The 4-D approach to dynamic machine perception: Recursive estimation techniques; extension to perspective imaging; orientation toward generic classes of physical objects and expectations. Intelligent control of image processing steps; data fusion through dynamical and measurement models. Homogeneous coordinates, translations, rotations, perspective mapping.

4. Image feature extraction: Area- (region-) vs. edge based features; gradients and their extreme values, the CRONOS software package geared to the 4-D approach. Area-based features, intensities, colors, textures; multiple scales and image pyramids; the TRIANGLE algorithm for real-time dynamic scene understanding. Intelligent control of feature extraction in the 4-D approach.

5. Roads and relative vehicle states: Planar road models: averaged moving model, locally fixed models; extensions to 3-D road model. Vehicle state and road parameter estimation (road perception); imaging geometry, range and multi-focal vision. Recursive estimation.

6. Obstacle recognition and relative state estimation: Systematic search for feature detection, controlled additional search for hypothesis generation. Multi-object recognition and tracking: object detection, hypothesis testing, state estimation. Shape recognition.

7. Vehicle control and behavioral capabilities: Longitudinal degrees of freedom: throttle and brakes; lateral control: steering angle and trajectory curvature. Feedback-based behavioral capabilities, feed-forward control elements, superposition for robust maneuver elements. Symbolic representations of capabilities and maneuver elements.

8. Mission performance: sequencing of behavioral capabilities: Mission decomposition (planning), monitoring; special task domains. General aspects of system integration. Examples: VaMP, functionalities required for driving on freeways; VaMoRs, driving on minor road nets. System structure, implementation aspects, experimental results.

9. Summary and outlook: Autonomous performance vs. driver assistance (haptic warnings, but no override) vs. survival functions (override for safety reasons only); development trends. What are the limits?


URL of this page: http://www-eecs.mit.edu/AY98-99/fall-cat/6291.html
Editor: Mibsy Brooks  |  Created: Jun 31, 1998  |  Modified: Jan 11, 1999
Related page: EECS Fall 1998 Catalogue Supplement
To MIT EECS home page  |  Your comments and inquiries are welcome.