Welcome to the ICT-OPTIMIX project website
OPTIMIX is a Specific Targetted Research Project in the key line "Networked Media" of the EU Seventh Framework Programme for Research and Development (Contract Number: INFSO-ICT-214625). The project has started on March 1st, 2008 and has a duration of 36 months.
The consortium offers a good balance between industrials and academic research centres, which offers to the project a good theoretical insight and knowledge of the most advanced state-of-the-art solutions as well as a good vision of current market trends and customer wishes.
OPTIMIX project studies innovative solutions enabling enhanced video streaming for point to multi-point in an IP based wireless heterogeneous system, based on cross layer adaptation of the whole transmission chain. The aim of the project is to increase the perceived quality of service for the user thanks to efficient cross-layer mechanisms enabling efficient joint approach between application world and transmission world.
Innovation in the area of sophisticated multimedia source coding schemes aiming to satisfy design criteria and trade-offs in terms of source representation quality, bitrate, delay, encoding/decoding complexity, etc. is a key issue in the modern world where users demand for content “anywhere and anytime”. Today’s approach, relying on traditional separation approaches and focussing on services delivered over homogeneous networks, does not allow to meet the on-going demands to maintain the required Quality of Service (QoS) for each of the users, who have different needs and requirements.
OPTIMIX addressed these issues by proposing a solution where the different entities involved in the end-to-end transmission are enhanced for multicast video transmissions and controlled by two new units: a Master Application (MA) Controller and the Base Station (BS) Controller. The proposed architecture comprises both data and control plane and has been finalized. The control plane is of fundamental importance in the OPTIMIX solution: indeed, it allows the transfer of feedbacks and commands between different network entities, thus enabling the end-to-end optimization.
IEEE 802.21 is used in OPTIMIX clients and BSs to provide physical and data link layer feedbacks to the other entities in the network: 802.21 information is converted into triggers which are handled by the novel triggering engine (TRG). The TRG is present in server, BSs and clients and allows to aggregate and exchange information between trigger sources and trigger consumers at any layer of the networking stack. Besides aggregation in each single entity done by TRG, OPTIMIX proposes a network level aggregation based on IPv6 anycasting. Feedback Aggregation Servers (FAS) are introduced into the IP network to generate aggregated IP packets containing feedbacks of the same type sent from multiple sources, thus decreasing the overhead in the IP core. Beside the design of the signalling architecture, the project detailed the end-to-end adaptation of the stream and the interactions between controllers. Different multicast groups (two for H264/AVC and more for H264/SVC, where the video is accompanied by non-adaptive AMR-WB encoded audio) are used for transmission and each group contains a different scalability layer. Each client thus subscribes to a selection of multicast groups which allow together to obtain the desired quality. The MA Controller collects information on available throughput, loss, bit error and packet losses from the different clients and Base Stations, used to select, using video quality measures, different input parameters (e.g., Quantization Parameters for the video encoder, the rate for Forward Error Correction, etc.). Different criteria for the selection policy are considered: average values, premium users, etc. In each BS the generated stream is further adapted by filtering packets addressed to multicast groups containing enhancement layers exceeding the local and current available throughout and thus avoiding any kind of stream modification at the BS, which could be compatible with end-to-end solutions. Finally, remarkable results have been obtained by enhancing the behaviour of the single modules used in the data plane, such innovative mechanisms for QoS support and mobility management .
A system simulator integrating all the modules composing the OPTIMIX system and allowing both unicast and multicast transmissions has been developed. Its description is available on the OPTIMIX web site. This simulator has been used by the consortium to test and refine (when needed) the implemented algorithms. Moreover, this simulator is a complete and easy instrument for the assessment of the results achieved by the project.
Selected features of the whole communication system have been integrated in a real-time test-bed. The test-bed is composed by a server that can encode on the fly audio and video streams or provide encoded offline streams and assures the transport of the multimedia stream toward the clients. In the server, the optimisation features can be enabled or disabled, allowing to test and exhibit different configurations. The server is then connected to a simulated IP-core network (introducing customizable loss and delay, as well as modelling a Traffic Engineering mechanism with fault resilience capability) and a real IPv6 network obtained through the FP6 ANEMONE test-bed and supporting all IPv6 core networking features and also the mobility specific protocols and extensions implemented to permit users to roam across different networks and access technologies while keeping their actual communicating sessions open anywhere and anytime. Two Access Point (AP) are also connected to the ANEMONE test-bed: an OPTIMIX AP, equipped with a Comsis IEEE 802.11g/n board developed for the project and presenting enhanced features compared to a COTS card like support for partial checksum and management of priorities, and a legacy IEEE 802.11g/n AP. Clients presenting different degrees of optimization communicate with the APs. As an example, an OPTIMIX client, providing feedbacks to the server, integrating enhanced transport/network features (like RTP FEC, header compressions, etc.) and equipped with an enhanced Comsis IEEE 802.11g/n board supporting MAC partial checksum is connected to the OPTIMIX AP, while a legacy client, equipped with a legacy IEEE 802.11g/n card, is connected to the legacy AP.
The end-to-end optimization strategy proposed by the project has been evaluated both by simulation, using the OMNeT++-based simulator developed by the partners, and tests in a real-environment. To this purpose, different degrees of optimization have been considered both for H.264/AVC and H.264/SVC streaming and in case of unicast and multicast transmissions. We considered in particular: - A "reference case", - An "enhanced case", where Forward Error Correction (FEC) with a fixed amount of repair packets introduced in RTP to protect the streaming, - A "partially optimised case", with the Application Controller at the server side adapting encoding and protection, - A "fully optimised case", where both server and Base Station have been modified.
Results include: i) objective metrics like throughput, delay, jitter and overhead, measured both on the wireless network and end-to-end, ii) video quality expressed in terms of PSNR and SSIM, iii) subjective quality results. Subjective validation has been performed in Kingston University and at the University of Budapest, taking into account ITU recommendations ITU-T P910 (Subjective video quality assessment methods for multimedia applications) and ITU-R BT.500-11 (Methodology for the subjective assessment of the quality of television pictures).
Results showed that, with all the considered video sequences, the advantages of the OPTIMIX solutions are evident.
Results obtained in the H.264/AVC case showed that, even considering unfavourable radio channel conditions, the introduction of the sole Master Application Controller determines “good” results (90% of the frames with PSNR>25dB) with all the considered test sequences, while a Fully Optimised scheme produces “excellent” videos (90% of frames above 30dB) with all four sequences. On the contrary, traditional schemes show poor performance. Results obtained with SSIM confirm results obtained with PSNR. Moreover, in case of “Full optimization” the observers participating to the subjective tests were not able to distinguish the original and transmitted videos most of the times. Another important observation concerns the quality fairness achieved by the different schemes. Also from this perspective, we noticed that the introduction of the Application Controller permits to achieve a considerable improvement with respect to the more traditional Enhanced Case. The Fully Optimised solution, then, is characterised by an excellent level of fairness, managing to equally satisfy all the OPTIMIX users. Results obtained with H.264/SVC were quite similar and showed in particular the advantage of protecting with different rates base and enhancement layers.
Tests on Comsis 802.11 boards, confirmed these results and additionally highlighted the gain introduced by a partial CRC done on the packet headers only when a video decoder robust to bit errors, like the AVC one used here, is available.
OPTIMIX achieved also important results in terms of dissemination and contributions to standards. The dissemination part mainly covers publications and participation at conferences, workshops, symposiums, concertation meetings, and other ICT project meetings. Publications were submitted to different conferences (e.g., Globecom, VTC, ICC, ICASSP, ICIP) and journals (e.g., EURASIP, Hindawi, IEEE Signal Processing Letters, IEEE Transactions on Vehicular Technology, IEEE Transactions on Wireless Communications, IEEE Communications Letters) and resulted in more than 150 publications.
Standardisation was done by three partners. Participation at the standardisation bodies ISO/IEC MPEG and ITU-T VCEG, and the JCT-VC was in the focus here. The relevant meetings were attended. Two related proposals on one of the topics, i.e., noise filtering of the reference frame for temporal prediction, were presented and positively acknowledged. A continuation of this effort is planned. Additionally, three contributions have been done to IETF on Secure-RTP, RTP Forward Error Correction and IPv6 Anycast.
Finally the OPTIMIX test-bed allowed the exhibition and promotion of the project achievements. In particular, the test-bed has been used to show project results at presented at THALES Techno Days, an exhibition on innovative products/prototypes open to THALES customers and to THALES employees worldwide.
The end-to-end approach proposed by OPTIMIX project aims at providing to the end user a versatile and adaptable secure infrastructure which can satisfy the needs of rich multimedia based applications. This structure is meant to offer transmissions that will be more bandwidth efficient and more robust, but also to take into account the heterogeneity of multiple users’ requirements and capabilities thus resulting of particular interest for the main business actors.
Directly related to partners activities, video surveillance systems (developed, e.g., by Thales Security Systems and Siemens), low data rate mobile communications (Thales ISR) and telemedicine (Siemens) could benefit from the project results. New service offers and corresponding business opportunities are also expected to emerge from this end-to-end optimisation with cross-layer design, and the demonstrator realised in WP4 is meant to help the exploitation of project results.