[Paper Review] Real-Time Rate-Distortion Optimized Streaming of Wireless Video

Ahmed Abdelhadi, Andreas Gerstlauer+1 more|arXiv (Cornell University)|Jun 7, 2014

Video Coding and Compression Technologies9 references3 citations

TL;DR

This paper proposes a low-complexity, real-time rate-distortion optimized (RDO) streaming algorithm for wireless video transmission over mobile autonomous aerial vehicle (AAV) networks. By jointly optimizing video encoding and transmission based on rate-distortion trade-offs, the method achieves significantly improved temporal video quality and reliability, outperforming non-optimized approaches in both simulation and real-world AAV testbed experiments using Horus.

ABSTRACT

Mobile cyberphysical systems have received considerable attention over the last decade, as communication, computing and control come together on a common platform. Understanding the complex interactions that govern the behavior of large complex cyberphysical systems is not an easy task. The goal of this paper is to address this challenge in the particular context of multimedia delivery over an autonomous aerial vehicle (AAV) network. Bandwidth requirements and stringent delay constraints of real-time video streaming, paired with limitations on computational complexity and power consumptions imposed by the underlying implementation platform, make cross-layer and cross-domain co-design approaches a necessity. In this paper, we propose a novel, low-complexity rate-distortion optimized (RDO) algorithms specifically targeted at video streaming over mobile embedded networks. We test the performance of our RDO algorithms using a network of AAVs both in simulation and implementation.

Motivation & Objective

• Address the challenge of reliable, low-latency video streaming in mobile cyberphysical systems with constrained bandwidth and computational resources.
• Design a cross-layer, cross-domain co-design approach for real-time video delivery over dynamic AAV networks.
• Develop a low-complexity rate-distortion optimized (RDO) algorithm tailored for embedded wireless platforms with strict power and processing constraints.
• Evaluate the proposed RDO algorithm in both simulation and real-world implementation using the Horus AAV testbed.
• Compare performance across different video codecs and transmission strategies in terms of temporal and spatial distortion metrics.

Proposed method

• Design a low-complexity rate-distortion optimization (LCRDO) framework that balances video quality and transmission efficiency in real time.
• Integrate RDO with video codecs such as MJPEG/SMOKE, MPEG2, and H.264 to adapt encoding based on channel conditions and buffer states.
• Implement an adaptive transmission strategy that prioritizes frames based on distortion metrics and temporal dependencies.
• Use optical flow and SSIM (Structural Similarity Index) to quantify temporal and spatial distortions in received video streams.
• Conduct simulations using dynamic AAV mobility and channel models before validating in a real-world Horus testbed.
• Apply feedback mechanisms and evaluate performance across unicast and multiple unicast topologies in real flight experiments.

Experimental results

Research questions

• RQ1 How can rate-distortion optimization be made practical for real-time video streaming in mobile embedded networks with limited computational and bandwidth resources?
• RQ2 What is the impact of RDO-based transmission on temporal video quality and frame recovery in dynamic AAV networks?
• RQ3 How do different video codecs (MJPEG/SMOKE, MPEG2, H.264) compare in terms of distortion and reliability under RDO-optimized transmission?
• RQ4 Can RDO-based streaming maintain high video quality and low latency in real-world AAV flight experiments with variable channel conditions?
• RQ5 What are the key trade-offs between computational complexity, bandwidth usage, and video quality in AAV-based video streaming?

Key findings

• The LCRDO algorithm achieved the highest fraction of received video duration (case c) across all flight experiments, indicating superior temporal video quality and reduced frame loss.
• The average SSIM index was approximately equal across all transmission cases, indicating that spatial distortion is primarily influenced by the codec rather than the transmission algorithm.
• MJPEG/SMOKE outperformed MPEG2 in average SSIM due to its frame-dropping behavior on partial frames, which avoids propagating distortion from incomplete decoding.
• The LCRDO-Adaptive algorithm successfully supported simultaneous transmission of two video streams to a single ground station, demonstrating feasibility for multi-source/multi-destination AAV networks.
• For low-computation platforms, MJPEG/SMOKE provided the best performance; for moderate computation, MPEG2 was more suitable; for high-computation, H.264 offered better bandwidth efficiency.
• Results showed a 5% variation across repeated experiments, indicating measurement accuracy within 95% confidence.