A SIMPLE YET EFFECTIVE MOVING OBJECT TRACKING AND HANDLING TECHNIQUE FOR SELF-DRIVING CARS
Name: LUAN FERREIRA REIS DE JESUS
Type: MSc dissertation
Publication date: 27/12/2018
Advisor:
Name |
Role |
|---|---|
| CLAUDINE SANTOS BADUE | Advisor * |
Examining board:
| Name |
Role |
|---|---|
| CLAUDINE SANTOS BADUE | Advisor * |
| PATRICK MARQUES CIARELLI | External Examiner * |
| THIAGO OLIVEIRA DOS SANTOS | Internal Examiner * |
Summary: "I this work is presented a simple yet effective moving object tracking and handling tech-
nique for self-driving cars we called Symotha. Symotha uses an obstacle-distance grid map
(ODGM), in which each cell holds the coordinates and the distance to its nearest obstacle, for
inferring the state of relevant moving objects. It does so by successively searching for occupied
cells across the self-driving-cars path and its neighboring left and right lanes. Successful de-
tections are accumulated over time to produce location and velocity estimates of moving ob-
jects. Symotha handles moving objects by changing the self-driving cars longitudinal speed in
order (a) to keep proper distance from moving objects and (b) to allow a smooth driving expe-
rience. In addition, it changes the ODGM seen by other modules of the self-driving car in such
a way as to hide moving objects from then, which simplifies their implementation significantly.
Experiments demonstrate that Symotha can safely handle both static and dynamic obstacles
without compromising on driving smoothness."
