Miscellaneous


Steve Viscelli: Trucking and the Decline of the American Dream | Lex Fridman Podcast #237

Nov 4, 2021

Steve Viscelli is a former truck driver and now an economic sociologist at University of Pennsylvania studying freight transportation, including autonomous trucks.

Outline:

0:00 - Introduction
0:44 - Ethnography
12:57 - Challenges of driving a truck
31:36 - Trucking industry: State of affairs
1:04:41 - Future of autonomous trucks
1:30:57 - Solving the automated truck dilemma
2:02:52 - Role of society in automated trucking
2:30:01 - Tesla and revolutionizing the trucking industry
2:49:41 - Hope and final thoughts
 

These are the self-driving cars to watch in 2022

Feb 22, 2022

You can’t lean back and strap on your Oculus behind the wheel just yet, but we’re getting tantalizingly closer to that day. Cooley shows you the autonomous cars he’s most excited about seeing this year.
 

UW Off-road autonomous driving - Off Trail Run

Apr 3, 2022

This video shows a test of autonomous off-road driving conducted by University of Washington researchers from the Paul G. Allen School of Computer Science and Engineering and the Applied Physics Laboratory. This test is on the "Offtrail" course, a short run that features unstructured terrain with scattered tree obstacles, very muddy forest trails, a number of drainage ditches and mounds, and a hairpin turn. The course in this video is defined by five waypoints, and the goal of this test is to run the course aggressively with no human interventions. The autonomous vehicle uses only onboard sensors and compute and does not reference GPS or predefined maps for localization.

Compared to on-road driving that is engineered to be simple, repeatable, and predictable, off-road terrain lacks man-made structure and contains challenging features including vegetation, uneven and low-friction surfaces, reduced visibility, obstacles, and rapidly changing terrain surface properties. When navigating at high-speed in these conditions, existing approaches to perception, planning, and control fail.
 

TartanDrive: Roboticists go off road

May 23, 2022

Researchers from Carnegie Mellon University took an all-terrain vehicle on wild rides through tall grass, loose gravel and mud to gather data about how the ATV interacted with a challenging, off-road environment.

They drove the heavily instrumented ATV aggressively at speeds up to 30 miles an hour. They slid through turns, took it up and down hills, and even got it stuck in the mud — all while gathering data such as video, the speed of each wheel and the amount of suspension shock travel from seven types of sensors.

The resulting dataset, called TartanDrive, includes about 200,000 of these real-world interactions. The researchers believe the data is the largest real-world, multimodal, off-road driving dataset, both in terms of the number of interactions and types of sensors. The five hours of data could be useful for training a self-driving vehicle to navigate off road.

"Roboticists Go Off Road To Compile Data That Could Train Self-Driving ATVs"
TartanDrive Dataset Likely Largest for Off-Road Environments

by Byron Spice
May 25, 2022
 

Developing safe autonomous vehicles

Mar 22, 2023

Uncertainty about self-driving cars / autonomous vehicles (AVs) is at an all-time high. Michigan Engineering researchers aim to change that. Training AVs to recognize safety hazards is a complicated task. Autonomous vehicles can typically handle 99.99% of safety use cases. Once you get to the 0.001%, AVs may not be able to handle these case uses because they haven’t seen the scenarios yet. This 0.001% is the curse of rarity. Training autonomous vehicle software is especially time-consuming and expensive, because individual safety use cases come up so rarely in normal driving conditions.
news.engin.umich.edu/2023/03/simulated-terrible-drivers-cut-the-time-and-cost-of-av-testing-by-a-factor-of-one-thousand

To fix this problem, a team of researchers used artificial intelligence to train virtual vehicles that can challenge autonomous vehicles in a virtual or augmented reality testing environment. The virtual cars were only fed safety-critical training data, making them better equipped to challenge AVs with more of those rare events in a shorter amount of time. In an era of uncertainty towards AVs, this solution can save auto manufacturers a prohibitive amount of time and money to ensure their systems are safe.

This research was led by Professor Henry Liu, Director of Center for Connected and Automated Transportation (CCAT), Director of Mcity
cee.engin.umich.edu/people/liu-henry

aper:
"Dense reinforcement learning for safety validation of autonomous vehicles"
Journal: Nature
Date: March 22, 2023
nature.com/articles/s41586-023-05732-2
DOI: 10.1038/s41586-023-05732-2
 
Back
Top