What Is Extended Object Tracking? | Autonomous Navigation, Part 5

See the other videos in this series: • Autonomous Navigation
In many practical scenarios, there are other objects that may need to be observed and tracked in order to effectively navigate within an environment. This video will show extended object tracking-objects that return multiple sensor detections. It will cover a basic overview of extended object tracking, what makes it challenging, and briefly provide intuition around some of the algorithms that have been developed to solve the problem.
Check out these other references:
- Sensor Fusion and Tracking Tech Talk Series: • Understanding Sensor F...
- (MATLAB Documentation) Multiple Extended Object Tracking: bit.ly/39k9lvQ
- (MATLAB Documentation) Partition Detections: bit.ly/2ZSDZcP
- (MATLAB Documentation) Extended Object Tracking and Performance Metrics Evaluation: bit.ly/2OWayQt
- Extended Object Tracking: Introduction, Overview, and Applications by Granstrom et al. arxiv.org/abs/1604.00970
- Random Finite Sets by Lennart Svensson. • Part 4: Random Finite ...
--------------------------------------------------------------------------------------------------------
Get a free product trial: goo.gl/ZHFb5u
Learn more about MATLAB: goo.gl/8QV7ZZ
Learn more about Simulink: goo.gl/nqnbLe
See what's new in MATLAB and Simulink: goo.gl/pgGtod
© 2020 The MathWorks, Inc. MATLAB and Simulink are registered trademarks of The MathWorks, Inc.
See www.mathworks.com/trademarks for a list of additional trademarks. Other product or brand names may be trademarks or registered trademarks of their respective holders.

[@justarandomguy916]

I had my first control class last semester and I learnt more from Brian and Norman Nise than I ever did from my lecturer

[@oldcowbb]

damn, brian always cover topics that i'm interested in, from control to autonomy

[@HyunukHa]

Great insights, Thanks Brian as always.

[@joaoantunes9756]

Great video! Thank you!

[@matlablatexsoftwares]

Good job sir

[@sherlypagcaliwangan4064]

Hi, ive been looking for a Software that can be use for Automated driving Testing software that can be use for diving school to check their performance on driving before they can have actual diving Licence

[@jiayuanli5189]

the best！

[@BrianBDouglas]

Thanks!

[@kjr5034]

Hi, drawing an ellipsoid or rectangle around a bunch of points that fits the distribution of points like shown in 5:10 seems to be a hard mathematical problem and not as straight forward as shown here. I could not find a function for that in Matlab. Only idea to solve this on my own is to find the center of the point distribution, draw the ellipsoids and than iterate through size and rotation until it fits the best. The computational cost for this is way to high though.
Any ideas?

[@kalebakeitshokile1366]

This might be a bit of a wild card but could you maybe use elliptical regression. So once you have been able to determine which data belongs to which object, you fit the data using elliptical regression.

[@kjr5034]

@@kalebakeitshokile1366 looks promising thx

[@victorvanderdrift5006]

You could use Principal Component Analysis to determine the eigenvectors (principal axes, local x-y-z axes) and their eigenvalues (corresponding sizes)

[@okanozturk2349]

👍

[@shivaa3ps177]

Maybe a dumb question...
Instead of taking thousands or lakhs of detection points from the same object and getting confused about which point belongs to which, can't the sensor identify color of the object and "fix" the 2 border points to get an indication the size of the obstacle?

[@Shafeefable]

Colour is not a valid way to identify clusters, because LIDARs don't have vision like a camera, they return the distances to points on a detected obstacle in its field of view. Moreover, the same object can have different colours like different paints/graphics on the car, making it unreliable. Instead, if we have a camera on-board, we could leverage edge detection and the LIDAR data points to form the clusters.

[@shivaa3ps177]

@@Shafeefable Makes sense...
I need to work on this tech...
Lidar vs computer vision as a future of self driving card is interesting to watch and work towards...
I'd love to see Zero drunk and drive disasters.

[@jeffsteyn7174]

When i drive it doesn't matter if it doesn't matter if it's a one object in my path. Or two or 4 objects that look like one object. Even if it's a part of an object that's all of a sudden moving or changing shape (like a car standing still and the door openning) The only thing that matters is that , is the object in my path and/
Or Is it moving into or out of my path and how quickly. Everything else is irrelevant.
It's all well and good you understand the algorithms. But you missing the fundamental understanding of how driving occurs.