Hacker News Reader

Comment by aharm | Hacker News Reader

Alex Harmsen here, co-founder at Iris Automation. Happy to answer any questions and share more about how the drone market is changing!

View on HN

What is the business case for drone delivery with regard to energy efficiency and total cost of ownership and fundamentally to deliver? Except for high value, low-weight items, how do you see competition in the coming years in the face of fleets of autonomous terrestrial vehicles that don't have to lift their payloads into the air?

I know Amazon has said previously most of their items "can" be delivered by drone, but the question remains should and will they, considering the high cost of current battery technology. Will it ever really be financially viable to lift items into the air in an individual manner when you can get them to their destination far more efficiently via the ground? It seems to me the case is for elimination of the human component in delivery.

If drone delivery is really so important, why aren't the major shipping companies working on it and scraping their existing air freight and hub and spoke distribution systems?

So are you planning to copy everything from Skydio or just the job posting?

https://www.google.com/search?q=The+core+of+our+system+is+th...

The article makes it sound like this is primarily for the drone delivery segment. That makes some sense, because it's at least easy to define the end goal. Do you also have plans for applications where the goal isn't as easy to define? That is, what can you automate beyond obstacle detection/avoidance?

Sounds like the package delivery part was more for journalistic flair than the core market. Collision avoidance beyond-line-of-sight is useful for pipeline inspection, agricultural surveys, search & rescue, forestry management, and many more use cases!

Alex, what is the maximum detection distance you expect is required for Iris to successfully negotiate an avoidance maneuver in its concept of operation. Further, what are the typical targets/obstacles that Iris would intend to detect and avoid?

Seems this didn't get any love, so I'll clarify. There is a tradeoff in exposure area and distance, for a given data rate out of the sensor. Because there are already many projects related to obstacle avoidance in the near field, I am curious if Iris chooses to address traditional air traffic targets (as depicted on the website image). If this is the case, what types of aircraft and at what distances. Put yet another way, does Iris intend to compliment traditional TCAS-like systems (albeit for non-cooperative targets)?

One of the major advantages of the Iris system is being able to deal with other aircraft in the national airspace (especially non-cooperative targets like you mention). It means that any system that will actually be useful will have to be able to see and track objects correctly further than 500m away, like Iris can.

So how does it work?

Is it a sort of SLAM algorithm?

SLAM + deep learning algorithms, all visual

Interesting! Are you guys doing SLAM using deep learning? IIRC SLAM is still mostly done with traditional computer vision.

Can't give away all of our secrets ;) Email me at alex@irisautomation.ca if you want to chat more!

> Today’s SLAM systems help machines geometrically understand the immediate world (i.e., build associations in a local coordinate system) while today’s Deep Learning systems help machines reason categorically (i.e., build associations across distinct object instances). In conclusion, I share Newcombe and Davison excitement in Visual SLAM, as vision-based algorithms are going to turn Augmented and Virtual Reality into billion dollar markets. However, we should not forget to keep our eyes on the "trillion-dollar" market, the one that's going to redefine what it means to "work" -- namely Robotics. The day of Robot SLAM will come soon.

from http://www.computervisionblog.com/2016/01/why-slam-matters-f...

I'm more familiar with "classical" computer vision and pattern recognition, but this makes sense to me.

Definitely on the right path

Nice. Are you guys using Jetsons for cpu+gpu?

Are you using something like OpenCV behind the scenes?

I know it's very early for you guys, but I was curious if you have a ballpark for your product price? Even cheap LIDARs still cost a decent penny, and camera only systems have issues with drift.

One of the main reasons we choose to go with simple cameras is because it can be so much cheaper than LIDAR and RADAR sensors. Won't disclose the actual price, but it is definitely a lot cheaper.

Have you thought to do everything by software only? If yes that may help you acquire much more customers.

Definitely! The core of our product is software so ideally we just sell software licenses (much higher margins). Turns out there are a lot of drone companies that want a plug and play product though, and don't have spare processing power around to run our software.

Reply on news.ycombinator.com