Hey guys,
I've finally managed to make enough progress to get to a point where it's worth sharing my project. Allow me to introduce Mog-Y!
So, the purpose of Mog-Y is to provide an experimental platform to develop an architecture for autonomous vehicles that is scaleable and can be applied to as many different configurations as possible. I've got a plan to develop more vehicles of various sizes, and one or two cools things in mind.
So the vehicle uses a Pixhawk for vehicle control and sensor data collection, and this will be linked to an Nvidia Jetson board. The Jetson board will be used for high-level tasks including comms, re-routing and controlling other things. I need to start putting money aside for a Velodyne Puck to enable the more interesting route planning on the fly.
The voltage regulator for the LiPo is in the post so I'll have the jetson up and running in the next few days. And then the scary programming begins!
Replies
Gary - yup there are lots of options. I've been looking at the options for some time, and trying ot decide the best way forward. I guess I jumped ahead of myself with the Velodyne comment. I definitely plan to use the pucks, they look great for the price in comparison to the cost of their previous products and I don't want to develop a solution that won't work when it comes to the kinds of data rates that 300000 points per second could create. That's why I went for the jetson board - I'm hoping it's cuda cores can take care of the vast amount of processing of that point data in to something useful to interpret. But, as you rightly point oiut, there are plenty of options to begin with. Sick is another supplier and there are some used examples on Ebay here in the UK, but they look pretty big for this rover.
Which Hokuyo unit do you have? I need to develop a system that works outdoors because I need to accurately cover quite a large area. I think I am going to go along the Inertial Nav route rather than relying on GPS because, again RTK GPS is out of the question price wise, and of course, there are weaknesses with GPS in terms of denial of GPS signal and so on. This leads on to sensor calibration and drift issues, but I got talking to a sensor producer at the recent commercial UAV show in London and they seem to think they can help with calibrated sensors at a realistic price, so we'll see.
Cheers,
Phill
Hi Phil,
Project Truck looks Great.
I am slowly working in similar direction with my own large 2 wheel hanging pendulum robot (26" brushless bicycle wheels).
I am also looking at the Nvidia Jetson board.
But, if I could make a suggestion to at least consider a laser ranger or simple 2D scanner instead they are a lot cheaper than the Velodyne solutions.
I have a Hokuyo scanner which is easy to servo scan vertically and costs less than a quarter of the Velodyne.
Or for really cheap, you can servo (or brushless gimbal) scan a laser ranging module such as the $300.00 Lightware SFO2 or soon (hopefully) to be released $80.00 Lidar Lite in 2 dimensions for full 3D scanning.
http://www.lightware.co.za/shop/en/lrf-modules/7-sf02f.html
http://www.dragoninnovation.com/projects/32-lidar-lite-by-pulsedlight
Or of course, the new Kinect 1 TOF camera for Windows from Microsoft can give you a really nice point cloud as long as you aren't trying to deal with direct sun light.
Best,
Gary
Hi Phill,
Nice work! Thanks for sharing your rover project and thoughts and ideas concerning future developments.
Regards,
TCIII AVD