A Novel Study on Data Rate by the Video Transmission for Teleoperated Road Vehicles Tito Tang, Frederic Chucholowski, Min Yan and Prof. Dr. Markus Lienkamp 9th International Conference on Intelligent Unmanned Systems
Driverless vehicle delivery Car sharing Delivery to front door Distribution of vehicles Driving vehicles to charging stations Private electric vehicles No private parking lot No private charging station Driving vehicles to charging stations Delivery of charged vehicle to front door ICIUS 2013 2
Autonomous vs. teleoperated vehicles Autonomous vehicles Scenarios with low complexity Highways Teleoperated vehicles Urban traffic Red lights Pedestrians ICIUS 2013 3
System setup ICIUS 2013 4
Transmission Cellular networks UMTS-HSPA: nominal upload rate: 11 Mbit/s LTE: max. nominal upload rate: 75 Mbit/s Raw video Gray-scale video 640 x 480 pixels 25 frames per second 58.6 Mbit/s ICIUS 2013 5
Video CODEC (encoding, DECoding) MPEG-2 MPEG-4 H.264/AVC (MPEG-4 Part10) ICIUS 2013 6
Problem Statement A high bandwidth is required for a smooth display of video streams used in teleoperated system Current available bandwidth through cellular network is limited A novel study on methods to reduce the data rate for the video data transmission ICIUS 2013 7
Canny edge detection algorithm ICIUS 2013 8
LaPlace edge detection algorithm ICIUS 2013 9
Introduction Foveated imaging technique ICIUS 2013 10
H.264/AVC Video Compression Goals: Equally distributed and low datarate Minimum latency Implementation: x264 open source video encoding library Only I- and P-frames No B-frames (no reference to following frames) => No additional encoding delay Baseline profile veryfast and zerolatency encoder presets intra refresh feature distributes I-frames ICIUS 2013 11
Intra refresh feature avoids data rate bursts ICIUS 2013 12
Cameras 8 Cameras, each with a 640x480 resolution 25 Frames per second in grey-scale Front field of view of approx. 240 degrees ICIUS 2013 13
Scenarios Scenario 1: homogeneous without other vehicles Scenario 2: more complex, with vehicles and pedestrians ICIUS 2013 14
CPU Intel Core (TM) i5, 2.4 GHz M520 Windows XP operating system 2.0 GB RAM Table 1 Average Time for Processing Each Frame Scenario 1 Scenario 2 H.264/AVC 10.17 [ms] 11.46 [ms] Canny-Algorithm 27.27 [ms] 27.06 [ms] LaPlace-Algorithm 14.08 [ms] 14.20 [ms] Foveated-Imaging 09.00 [ms] 10.37 [ms] ICIUS 2013 15
Comparison in data size ICIUS 2013 16
Edge-detection algorithms Reduce amount of information Increase in data rate because of sharp corners in frame Foveated-Image-Technique Slight decrease in data rate Scenario 1: reduction of ~19% Scenario 2: reduction of ~17% ICIUS 2013 17
es to reduce data rate for data transmission Canny-Algorithm Laplace-Algorithm Foveated-Imaging-Technique Setup of x264 H.264/AVC video encoder Efficient and robust Minimum encoding latency show that edge-detection techniques do not provide a reduction, but the Foveated-Imaging-Technique does ICIUS 2013 18
Further research Determination of area of interest for Foveated-Imaging-Technique Head-tracking Eye-tracking Possible further reduction using upcoming High Efficiency Video Codec (HEVC or H.265) ICIUS 2013 19
Thank you for your attention ICIUS 2013 20
A Novel Study on Data Rate by the Video Transmission for Teleoperated Road Vehicles Tito Tang, Frederic Chucholowski, Min Yan and Prof. Dr. Markus Lienkamp 9th International Conference on Intelligent Unmanned Systems
ICIUS 2013 22
ICIUS 2013 23