🏒 πŸš™ πŸ€– The Physical World Tech Newsletter

By Sam Cash // Physical World Technologies Newsletter

🏒 πŸš™ πŸ€– Issue #40: DeepMind's StreetLearn, AVs and Uber for X

#40・
271

subscribers

45

issues

Subscribe to our newsletter

By subscribing, you agree with Revue’s Terms of Service and Privacy Policy and understand that 🏒 πŸš™ πŸ€– The Physical World Tech Newsletter will receive your email address.

🏒 πŸš™ πŸ€– Issue #40: DeepMind's StreetLearn, AVs and Uber for X
Welcome to my newsletter, where I discuss thoughts and news on the intersection of the built world and technology #retail #mobility #realestate #tech
Please contribute to the community by forwarding this to someone who might enjoy it also.
You can follow me on Twitter here

One Big Thought
DeepMind’s StreetLearn - last week DeepMind released StreetLearn (NeurIPS paper). Navigating through an unstructured environment is a basic capability of intelligent, sentient being, humans for example. Humans are able to do this as we not only have a sense for where we are in the present but also where we are looking to get to, using visual cues to refine our understanding of space and direction. Machines, to date, have only been able to navigate using coordinates, pre-defined routes and/or using positioning systems and localisation to wayfind.
What DeepMind is asking the agent to do is to wayfind to a specific location using only the dataset of StreetView images (the street-level photos you see when use Google Maps). This is the human equivalent of travelling through a city and memorising the various streets, locations, landmarks using only non-written visual cues.
Google’s DeepMind have done this by building a neural network agent that inputs images observed from the environment and predicts the next action it should take in that environment. They then train it end-to-end using deep reinforcement learning. The result is that the agent can navigate through a city having never built a map, merely knowing where visual representations or artifacts lie.
Learning to navigate in cities without a map
Learning to navigate in cities without a map
To date much of localisation and positioning is done using fairly imprecise and outdated infrastructure such as GNSS (global navigation satellite system) systems like GPS, Galileo and BeiDou and to a lesser extent CORS (continuously operating referrence stations) base stations, both of which triangulate location based on pinging signals to orbital and ground-level receivers, respectively.
As more intelligent robots, autonomous vehicles (AVs) for example, permeate our streets there is an increased need for precision in localisation, for these robots to accurately wayfind. Currently, AVs use a blend of a few techniques in order to manouever through streets and around objects. This is often using GPS in-part for localisation, static mapping layers, whilst overlaying dynamic data layers from sensors such as LIDARs and cameras.
Why is this important?
  • Whilst the use of the StreetLearn model isn’t going to have an immediate effect on localisation - more interestingly, it shows there are other more novels solutions which intelligent agents can produce when solving this trying to solve this issue. We could start seeing visual localisation based on existing visual data sets
AV Readiness report - KPMG just released their annual report on which countries are best prepared for the adoption of autonomous vehicles:
*spoiler alert*
Vague Scientist
β€˜Uber for X’ - is famed tech parlance and a proxy for the booming on demand economy. Someone, I’m not entirely sure who, has put together a full list of β€˜Uber for X’ startups, which lists all startups and their current status, who claim to be the Uber equivalent in a specific vertical
Dutch e-bikes - electrically powered bicycles are now biggest seeling single class of bikes in Holland. In 2018, 40% of the 1 million bikes sold in Holland were e-bikes. This is largely due to the cost of these bikes decreasing dramatically with the ASP being €1200, and having a populace enthused about adopting new biking technologies
Waymo selling sensors - Google’s Waymo is repordtedly going to start selling its own LIDARs in a bid to lower costs for the units. Until very recently, much of AV development has been held back by the high cost of LIDAR (>$80k) units
Lyft v Uber - with Lyft revealing its FY2018 numbers in S-1 filing, DealRoom decided to do a like-for-like comparison of its financials
Thanks for reading!
Sam
Did you enjoy this issue?
Sam Cash // Physical World Technologies Newsletter

The intersection of the physical world and technology; with a focus on future mobility,real estate, retail and cities.

Tweet Β Β Β  Share
In order to unsubscribe, click here.
If you were forwarded this newsletter and you like it, you can subscribe here.
Powered by Revue