Video Friday: Deep Learning for Cars, Space Invaders With Drones, and Disagreeable Robot

Your weekly selection of awesome robot videos

5 min read

Evan Ackerman is IEEE Spectrum’s robotics editor.

Erico Guizzo is IEEE Spectrum's Digital Innovation Director.

UT Austin Dreamer humanoid robot
Image: UT Austin via YouTube

Video Friday is your weekly selection of awesome robotics videos, collected by your Automaton bloggers. We’ll also be posting a weekly calendar of upcoming robotics events for the next two months; here's what we have so far (send us your events!):

ISER 2016 – October 3-6, 2016 – Tokyo, Japan
Cybathlon Symposium – October 07, 2016 – Zurich, Switzerland
Cybathalon 2016 – October 08, 2016 – Zurich, Switzerland
Robotica 2016 Brazil – October 8-12, 2016 – Recife, Brazil
ROSCon 2016 – October 8-9, 2016 – Seoul, Korea
IROS 2016 – October 9-14, 2016 – Daejon, Korea
NASA SRC Qualifier – October 10-10, 2016 – Online
ICSR 2016 – November 1-3, 2016 – Kansas City, Kan., USA
Social Robots in Therapy and Education – November 2-4, 2016 – Barcelona, Spain
Distributed Autonomous Robotic Systems 2016 – November 7-9, 2016 – London, England

Let us know if you have suggestions for next week, and enjoy today's videos.

IROS is approaching fast. Here’s a taste of what’s to come:

We’ll be there, of course.

[ IROS 2016 ]

Well-known auto manufacturer NVIDIA is demoing the capabilities of its new deep neural network-powered self-driving car, which has a robot in Star Wars named after it:

In contrast to the usual approach to operating self-driving cars, we did not program any explicit object detection, mapping, path planning or control components into this car. Instead, the car learns on its own to create all necessary internal representations necessary to steer, simply by observing human drivers.

The car successfully navigates the construction site while freeing us from creating specialized detectors for cones or other objects present at the site. Similarly, the car can drive on the road that is overgrown with grass and bushes without the need to create a vegetation detection system. All it takes is about twenty example runs driven by humans at different times of the day. Learning to drive in these complex environments demonstrates new capabilities of deep neural networks.

The car also learns to generalize its driving behavior. This video includes a clip that shows a car that was trained only on California roads successfully driving itself in New Jersey.


Thanks Michael!

Is there such a thing as a robot hand that's too capable? Yes, if it's in one of those claw machines:

The process by which RightHand Robotics designs and prototypes their hands is pretty cool, too:

We'll be seeing plenty more of these hands, as RightHand is sponsoring this year's IROS Robotic Grasping and Manipulation Competition, which will take place at IROS in Korea in a few weeks.

[ RightHand Robotics ]

Thanks Leif!

Unlike the arcade version of Space Invaders, this version has real consequences if the aliens reach the ground. Namely, you're out a really expensive drone.

This would be more fun with super high-powered battle lasers that could vaporize the drones, but still, not a bad start.

[ For Real ] via [ DIY Drones ]

I'm not looking at you. I'm not looking at you. I'm not looking at you.

[ UT Austin ]

Watching a robot not get wrecked by a wrecking ball is more interesting than it sounds:

[ Megabots ]

Do I want one of these? Yes, I want one of these.

And wow why don't I have a cookie warming lamp in my house? That's genius.


Can I just buy one of these to take me wherever I want, please?

[ Nissan ] via [ Gizmodo ]

While not strictly a robot, there's enough robotics history and application here that it's worth watching this video about using gecko-inspired adhesives to grip things in space:

[ Stanford ]

The Cybathlon competition is next week, and IHCM is ready to totally rock it with their exoskeleton:

This is a practice run of the full Cybathlon Powered Exoskeleton Course. IHMC will be competing in the Powered Exoskeleton Race on October 8, 2016 in Zurich, Switzerland. The race consists of 6 tasks with a 10-minute time limit. We are skipping one of the tasks, so this video shows only 5. Mark, our, pilot, who is paralyzed with a spinal cord injury, is fully controlling the exoskeleton. The tether is only for fall protection and to log controller data. The total time to complete the 5 is 8:54.

[ IHMC ]

Thanks Peter!

Looks like Valkyrie has been keeping busy over in Edinburgh:

[ University of Edinburgh ]

Drone-assisted water rescue requires specially developed equipment that can rise to the challenges often present in aquatic environments. The drone used in this rescue simulation was the microdrones md4-1000. This quadcopter is highly effective for water rescue because it has uniquely designed motors, rugged carbon fiber housing, highly efficient batteries, and an integrated GPS system. These features allow the UAV to fly and stay in position in strong winds over the water.

[ Microdrones ]

In this work, we show a fully integrated system that is energy efficient and enables MAVs to pick up and deliver objects with partly ferrous surface of varying shapes and weights. This is achieved by using a combination of an electro-permanent magnetic gripper with a passively compliant structure and integration with detection, control and servo positioning algorithms.

[ ASL ]

The German Research Center for Artificial Intelligence (DFKI), with locations in Kaiserslautern, Saarbrücken, Bremen (with a branch office in Osnabrück) and a project office in Berlin, is the leading research center in Germany in the field of innovative commercial software technology using Artificial Intelligence.

The DFKI research department Robotics Innovation Center (RIC), headed by Prof. Dr. Frank Kirchner, develops mobile robot systems that solve complex tasks under water, in space and in our everyday life. The goal is to design robots that operate autonomously and interact safely with humans, their environment and other systems. The RIC closely cooperates with the Robotics Group at the University of Bremen.

[ DFKI ]

Making autonomous robots is hard. Making autonomous robots that are waterproof is harder.

[ RoboSub ]

Here's another amazing retro robot video courtesy of Georgia Tech: Meet Centipede, which was doing robot-y stuff 45 (!) years ago.

The company MB Associates (San Ramon, CA) was a forerunner in the design of a number of different robot systems. They designed a teleoperated arm that was the foundation for an arm used by NASA and they built a variety of robots for defense applications. The Centipede robot used a set of “standard” modules to build a high agile mobile platform that can be used for transportation and dismounted operation. The modular design and the integrated control from a simple joystick is interesting for a robot from ~1970.

[ Georgia Tech ]

Fetch Robotics CTO Michael Ferguson tells you everything you ever wanted to know about running ROS-based robots in warehouses. Get comfortable:

[ Fetch Robotics ]

This week's CMU RI Seminar for your hardcore roboticists out there: Ashish Kapoor, Microsoft Research: Safe and Optimal Path Planning in Uncertain Skies

Achieving optimality while staying safe is one of the key problems that arise when planning under uncertainty. We specifically focus on path planning for aerial vehicles, where the uncertainties arise due to unobserved winds and other air traffic. A flight plan or a policy that doesn’t take into account such uncertainties can not only result in highly inefficient flight paths but can also jeopardize safety. In this talk, we will first focus on how to reduce uncertainty in wind predictions by using airplanes in flight as a large-scale sensor network. In particular, we explore how information from existing commercial aircraft on their normal business can be harnessed to observe and predict weather phenomena at a continental scale in greater detail that currently available. In the second part of the talk, we consider the problem of path planning under uncertain winds and traffic conditions. Specifically we propose planning algorithms that trade off exploration and exploitation in near-optimal manner and have appealing no-regret properties. Further, we will also discuss how Probabilistic Signal Temporal Logic (PrSTL) can be adapted to the robotic path planning problems in order to guarantee safety. We will present results from longitudinal real-world studies that demonstrate effectiveness of the framework.

[ CMU ]

The Conversation (0)