Talk

Implementing a Simultaneous Localization and Mapping (SLAM) system for Robotics/Unmanned Vehicles

Watch Now

45:07

EOC 2020

Login to Watch

Mike Anderson

Mike Anderson is currently lead systems architect for The PTR Group, LLC -- a subsidiary of Huntington-Ingalls Industries. With over 40 years in the embedded and real-time computing industry, Mike works with a number of RTOS offerings for IoT devices including mbed and Zephyr as well as embedded/real-time Linux on a number of CPU architectures. As an instructor and consultant, Mike is a regular speaker at the Embedded Linux Conference, Sensors Expo, Arm TechCon and the Embedded Systems Conference as well as other Linux and IoT-oriented conferences. Ongoing projects include work with mesh wireless topologies including Wi-Fi and Bluetooth mesh, IEEE 802.15.4/6LoWPAN and commercial reverse engineering. Additionally, he continues to play the role of mentor in working with the FIRST Robotics and CyberPatriot programs in the high schools.

View full profile

Topics Covered

Robotics

Talk

Implementing a Simultaneous Localization and Mapping (SLAM) system for Robotics/Unmanned Vehicles

Mike Anderson

45:07 EOC 2020

Mike Anderson

45:07

As robotics and unmanned vehicles like drones start to proliferate, many developers find that they need a means of fixing the device to a specific location and then have the device navigate to a destination while avoiding obstacles. In addition, we frequently want the device to create a map of the surrounding area for future reference. In order to most efficiently accomplish these goals, we need to create a SLAM for the device. In this session, we will discuss the various approaches and requirements for 3D SLAMs and how they are constructed. We will address implementations ranging from autonomous cars down to robots based on platforms such as the Raspberry Pi.

1 / 4

Please log in or create an account to test your knowledge and see the answers.

According to the talk, which set of sensors is the speaker describing as the basic, practical minimum to build a SLAM system that can both localize and map a new environment?

A Wheel odometry for displacement, an IMU for orientation, and a range sensor (e.g., LIDAR/radar/sonar) for distances

B A single camera only, since visual SLAM can always produce full maps without other sensors

C Only GPS, because latitude/longitude and elevation are sufficient for mapping and navigation

D Just an IMU (accelerometer + gyroscope) because orientation and dead-reckoning are enough

E Send all raw sensor data to the cloud and rely on remote servers to perform localization and mapping

Why does the speaker recommend FastSLAM over a naïve Extended Kalman Filter (EKF) implementation for larger mapping problems?

A Because EKF is quadratic in computational complexity (O(n^2)), while FastSLAM achieves similar results with much better (logarithmic) scaling

B Because EKF assumes no noise in sensors, but FastSLAM assumes perfect sensors

C Because EKF cannot run on any embedded hardware, while FastSLAM requires no computation at all

D Because EKF always diverges and never converges even with accurate sensors, whereas FastSLAM always produces a perfect map

E Because FastSLAM uses deep learning and neural networks, which EKF cannot use

What practical advantage of visual SLAM (V‑SLAM) with fiducial markers does the speaker highlight for applications like warehouse route following?

A Fiducials let a camera-based system follow a pre-planned path without having to build a full map of the entire environment, reducing computation and sensor needs

B Fiducials allow a single camera to measure absolute distance to every object in the environment with millimeter accuracy

C Using fiducials eliminates the need for any orientation sensing (no IMU required) because the camera provides absolute heading

D Visual SLAM with fiducials makes LIDAR unnecessary because cameras always work better in fog and rain

E Fiducials enable a robot to autonomously generate high-resolution 3D point clouds without any odometry or range sensors

What limitation of automotive LIDARs related to laser wavelength and eye safety does the speaker point out, and how do manufacturers address it?

A Near‑infrared LIDARs (around 600–1000 nm) are limited by eye-safety (Class 1) so manufacturers use longer wavelengths (≈1,300–1,550 nm) that permit higher power and longer range

B All LIDARs use visible green lasers so they are never subject to eye-safety limits and therefore always have maximum range

C Eye-safety is not a concern for LIDAR; the real limit is battery capacity, so manufacturers only improve batteries

D To avoid eye-safety rules, manufacturers stop using lasers entirely and only use sonar for automotive distance sensing

E Manufacturers avoid longer wavelengths because those penetrate human tissue more deeply and are thus banned for automotive use

Formatting help

italics	surround text with asterisks
bold	surround text with two asterisks
hyperlink	[hyperlink](https://example.com) or just a bare URL
inline code	surround text with single `backticks`
code block (multi-line)	wrap on its own lines with three backticks: ``` your code here ```
~~strikethrough~~	surround text with ~~two tilde characters~~
quote	prefix with >

Upvotes Newest Oldest

ThomDenholm

Score: 0 | 6 years ago | 1 reply

Really interesting, I especially appreciate the new terminology.

RealMikeAndersonSpeaker

Score: 0 | 6 years ago | no reply

Thank you!

Score: 0 | 6 years ago | 1 reply

I enjoyed your presentation very much. Will you be presenting at the online version of the Embedded Linux Conference this year?

RealMikeAndersonSpeaker

Score: 0 | 6 years ago | no reply

I'm hoping to. I'm thinking that this year's presentation will be on GDB debugging ins and outs. there have been a lot of enhancements to GDB recently. SO, they've asked if I can present that one.

RealMikeAndersonSpeaker

Score: 0 | 6 years ago | 3 replies

Good afternoon, everyone!

Jacob_Beningo

Score: 0 | 6 years ago | 2 replies

Hello Mike! Looking forward to the talk!

RealMikeAndersonSpeaker

Score: 0 | 6 years ago | no reply

Thanks for the chance to get this out there Jacob.

RealMikeAndersonSpeaker

Score: 0 | 6 years ago | no reply

OK, it just took a bit for the reply to show up. We're all good.

Stephane.Boucher

Score: 0 | 6 years ago | 1 reply

Hi Mike! Thanks for being part of the Embedded Online Conference. Do you remember being the star of this video? You were so good - just one take! At about 1:12.. https://www.youtube.com/watch?v=AKdvKplxoSI

RealMikeAndersonSpeaker

Score: 0 | 6 years ago | no reply

This presentation was done in one take as well. A few stumbles, but overall not too bad.

DrewFustini

Score: 1 | 6 years ago | 1 reply

Hi Mike! Excited to watch your talk. It's not an embedded conference without you :)

RealMikeAndersonSpeaker

Score: 0 | 6 years ago | no reply

Thanks, Drew!

Partners & Sponsors

STMicroelectronics

Arm

Microchip

The Qt Company

Cypress

NXP

IAR

PE Micro

LDRA

TUXERA

Become a Sponsor