Mapping IIT Kharagpur
Autonomous Ground Vehicle lab is one of the very few labs in India with state of the art sensors like Velodyne LiDAR, Carnegie MultiSense S7 stereocamera, SICK LiDAR, Microsoft Kinect, etc. Use of this sensors becomes very important for self-driving problem statements. One of the PS is localization of the vehicle which certainly cannot be achieved using only with GPS data. Hence, we instead thought of generating a Map of surrounding (college campus) first and then localizing the vehicle according to instances/features seen by the sensors. This Alum Meet (January’17), the dune buggy vehicle was installed with all the sensors present in lab and a dataset was collected around the campus. My team is planning to collect many such datasets in future and we are soon going to make them public.
In my previous blog, I had also mentioned arrival of a powerful server at the lab which we are now using not only for Deep learning purpose but also for running simulation platforms like Autoware. Recently, the Autoware package was installed some available algorithms were tested on our collected dataset. So, with such powerful resources in hand, I decided to generate a complete 3D map of IIT Kharagpur both for group benefit and personal satisfaction.
- Testing ndt_mapping on LiDAR dataset [15 Jan’17]
The rosbag collected was played for generating a 3d map using normal distribution transform algorithm and a *.pcd file was obtained. Here are some of the snapshots without any filter or outlier algorithm applied:
Previous Post