I'm going to hijack this thread to mention that I'm thinking of open sourcing a SLAM system I've been working on under BSD or MIT.<p>Ive been working on this because I'm interested in underwater drones (uuvs), and hope to commercialise this in the form of services.<p>My current understanding is as follows:<p>- most open source systems come from universities and are GPL, with the option for licensing. (PTAM, ORB SLAM, DSO, SVO, LSD)<p>- Optimistically, it would require $1-3 million to build a compedative system privately. The sky is the limit here.<p>- No one, neither the universities nor private entities, is making serious money by licensing just a SLAM system. The well performing startups are somehow coupled to a product or a service.<p>- All systems basically solve the same problems, but mix and match different solutions. For example for initial motion estimation, you could rely on constant motion (DSO, ORB SLAM), image alignment (SVO), or solving directly for rotation between cameras (GSLAM).<p>- As a researcher, to implement a new idea, you either need to implement an entire system from scratch, or modify a GPL'd system. This is great in theory, but since everyone is playing the GPL or pay game they are reluctant to accept contributions.<p>I'm imagining something like opencv, with implementations of common algorithms accompanied by wiki pages describing high level details. Components could be mixed and matched into different reference SLAM systems targeting different use cases, do you want something that will run on a MAV or a self driving car?<p>From a commercial point of view, companies could worry more about tweaking things to work for a particular use case without needing to start from scratch. I believe that universities, with their many departments, have more opportunity than anyone else.<p>My belief is that it's time for a liberally licensed SLAM system. Has anyone come to a similar conclusion or care to comment to the contrary?<p>Thanks HN