Today it is pretty easy to build a robot with an onboard camera and have fun manually driving through that first-person view. But builders with dreams of autonomy quickly learn there is a lot of work between camera installation and autonomously executing a “go to chair” command. Fortunately we can draw upon work such as View Parsing Network by [Bowen Pan, Jiankai Sun, et al]
When a camera image comes into a computer, it is merely a large array of numbers representing red, green, and blue color values and our robot has no idea what that image represents. Over the past years, computer vision researchers have found pretty good solutions for problems of image classification (“is there a chair?”) and segmentation (“which pixels correspond to the chair?”) While useful for building an online image search engine, this is not quite enough for robot navigation.
A robot needs to translate those pixel coordinates into real-world layout, and this is the problem View Parsing Network offers to solve. Detailed in Cross-view Semantic Segmentation for Sensing Surroundings (DOI 10.1109/LRA.2020.3004325) the system takes in multiple camera views looking all around the robot. Results of image segmentation are then synthesized into a 2D top-down segmented map of the robot’s surroundings. (“Where is the chair located?”)
The authors documented how to train a view parsing network in a virtual environment, and described the procedure to transfer a trained network to run on a physical robot. Today this process demands a significantly higher skill level than “download Arduino sketch” but we hope such modules will become more plug-and-play in the future for better and smarter robots.
In the early days of spaceflight, when only the governments of the United States and the Soviet Union had the ability to put an object into orbit, even the most fanciful of futurists would have had a hard time believing that commercial entities would one day be launching sixty satellites at a time. What once seemed like an infinite expanse above our heads is now starting to look quite a bit smaller, and it’s only going to get more crowded as time goes on. SpaceX is gearing up to launch nearly 12,000 individual satellites for their Starlink network by the mid-2020s, and that’s just one of the “mega constellations” currently in the works.
It might seem like overcrowding of Earth orbit is a concern for the distant future, but one needs only look at recent events to see the first hints of trouble. On September 2nd, the European Space Agency announced that one of its research spacecraft had to perform an evasive maneuver due to a higher than acceptable risk of colliding with one of the first-generation Starlink satellites. Just two weeks later, Bigelow Aerospace were informed by the United States Air Force that there was a 1 in 20 chance that a defunct Russian Cosmos 1300 satellite would strike their Genesis II space station prototype.
A collision between two satellites in orbit is almost certain to be catastrophic, ending with both spacecraft either completely destroyed or severely damaged. But in the worst case, the relative velocity between the vehicles can be so great that the impact generates thousands of individual fragments. The resulting cloud of shrapnel can circle the Earth for years or even decades, threatening to tear apart any spacecraft unlucky enough to pass by.
Fortunately avoiding these collisions shouldn’t be difficult, assuming everyone can get on the same page before it’s too late. The recently formed Space Safety Coalition (SSC) is made up of more than twenty aerospace companies that realize the importance of taking proactive steps to ensure humanity retains the unfettered access to outer space by establishing some common “Rules of the Road” for future spacecraft.
At the turn of the 21st century, it became pretty clear that even our cars wouldn’t escape the Digital Revolution. Years before anyone even uttered the term “smartphone”, it seemed obvious that automobiles would not only become increasingly computer-laden, but they’d need a way to communicate with each other and the world around them. After all, the potential gains would be enormous. Imagine if all the cars on the road could tell what their peers were doing?
Forget about rear-end collisions; a car slamming on the brakes would broadcast its intention to stop and trigger a response in the vehicle behind it before the human occupants even realized what was happening. On the highway, vehicles could synchronize their cruise control systems, creating “flocks” of cars that moved in unison and maintained a safe distance from each other. You’d never need to stop to pay a toll, as your vehicle’s computer would communicate with the toll booth and deduct the money directly from your bank account. All of this, and more, would one day be possible. But only if a special low-latency vehicle to vehicle communication protocol could be developed, and only if it was mandated that all new cars integrate the technology.
Except of course, that never happened. While modern cars are brimming with sensors and computing power just as predicted, they operate in isolation from the other vehicles on the road. Despite this, a well-equipped car rolling off the lot today is capable of all the tricks promised to us by car magazines circa 1998, and some that even the most breathless of publications would have considered too fantastic to publish. Faced with the challenge of building increasingly “smart” vehicles, manufacturers developed their own individual approaches that don’t rely on an omnipresent vehicle to vehicle communication network. The automotive industry has embraced technology like radar, LiDAR, and computer vision, things which back in the 1990s would have been tantamount to saying cars in the future would avoid traffic jams by simply flying over them.
In light of all these advancements, you might be surprised to find that the seemingly antiquated concept of vehicle to vehicle communication originally proposed decades ago hasn’t gone the way of the cassette tape. There’s still a push to implement Dedicated Short-Range Communications (DSRC), a WiFi-derived protocol designed specifically for automotive applications which at this point has been a work in progress for over 20 years. Supporters believe DSRC still holds promise for reducing accidents, but opponents believe it’s a technology which has been superseded by more capable systems. To complicate matters, a valuable section of the radio spectrum reserved for DSRC by the Federal Communications Commission all the way back in 1999 still remains all but unused. So what exactly does DSRC offer, and do we really still need it as we approach the era of “self-driving” cars?
The GP2Y0D02 is an infrared proximity sensor with a detection field that extends 80cm. This type of sensor can be used to build collision avoidance systems for robots. We’ll demonstrate this sensor using a single resistor and a multimeter.