V3: A vehicle-to-vehicle live video streaming architecture

Abstract

We propose V3, an architecture to provide a live video streaming service to driving vehicles through vehicle-to-vehicle (V2V) networks. V2V video streaming is challenging because: (1) the V2V network may be persistently partitioned; (2) the video sources are mobile and transient. V3 addresses these challenges by incorporating a novel signaling mechanism to continuously trigger vehicles into video sources. It also adopts a store-carry-and-forward approach to transmit video data in partitioned network environments. We first propose an initial design which supports video streaming to a single receiver. We then propose a multicasting framework that enables video streaming from multiple sources to multiple receivers. Simulation experiments demonstrate the feasibility of supporting V2V video streaming with acceptable performance.

Introduction

Streaming live video content to driving vehicles can provide very attractive services. Many applications can benefit from such services. (i) Driver assistance and safety applications: in case of a car accident or road work, if vehicles that are driving towards this area can receive live video about this area, the drivers can then drive cautiously or simply choose another route. (ii) Business or entertainment applications: road-side businesses, such as hotels and restaurants, can use content-rich video streams to broadcast advertisements to drivers on the road. (iii) Military or scientific applications: vehicles on duty can broaden their view by receiving video from other vehicles’ video cameras.

A widely used approach to broadcast traffic conditions to driving vehicles is to set up bulletin boards on certain spots of a major highway, and display traffic conditions in several lines of text. This approach can provide traffic information to some extent, but it is far from satisfactory: (i) it may suffer extensive delay; (ii) it is not customizable for different drivers on the road; (iii) it can only provide very brief information; (iv) very limited coverage area, this service can only reach vehicles that drive past these bulletins. Imagine a scenario where a vehicle is equipped with an LCD screen at the front panel which displays video information from a remote region. This screen serves similar functions as the rear view mirror, but it can display traffic conditions from any spot on the road. With a glimpse of this screen, the driver can obtain desired information without interrupting normal driving. We use video instead of text for the following reasons: (i) video provides content rich information; (ii) text needs human intervention as it often needs people to capture, organize and summarize to generate descriptive text information; (iii) video needs shorter time to understand, a glimpse of the video screen is enough to grasp the road conditions, while reading information like “Car accident on North Ave., 1 mile south of exit 250, 3 left lanes blocked …” takes significantly longer time.

Two basic approaches can be used to support video streaming to vehicles on the road: an infrastructure-based approach, and a vehicle-to-vehicle (V2V) approach. In this work, we are interested in investigating the problem of supporting a video streaming service using a vehicle-to-vehicle (V2V) approach. A V2V network is a special mobile ad hoc network in which each mobile node is a vehicle with wireless capability. Vehicles in a V2V network are equipped with on-board computing and wireless communication devices. It is also assumed that the vehicles are equipped with GPS devices which enable the vehicles to track their location and trajectory. Some vehicles are equipped with video cameras, as well as enough storage to store recorded video for a certain time, and are therefore able to capture and transmit live video from their surroundings. Other vehicles driving on the road can request to receive video from the camera-equipped vehicle.

Video streaming through a V2V network is practical with the advance of wireless communication and video compression techniques. The IEEE 802.11 standard supports data transfer rates at up to 54 Mbps. Even between high speed driving vehicles, it is reasonable to expect a 1 Mbps data rate [12]; if we assume a 320∗240 screen with 65536 color, a frame rate of about 15 fps, the widely-used MPEG compressed video stream rate is estimated to be about 100 to 150 kbps [6]; furthermore, each vehicle can support relatively large computational and storage resources. Ample power can be supplied from the engine of the vehicle. With enough power and large computational and storage capability, it is possible to explore data intensive streaming video service.

Video streaming in V2V networks is challenging for several reasons. (1) Network partitions: since each vehicle’s radio range is only about two hundred meters, a V2V network may not be connected at all times. Furthermore, due to the gradual market penetration, only a fraction of the vehicles on the road will be equipped with communication capability. We call the fraction of vehicles that are so equipped the penetration ratio. Only such vehicles can participate in the V2V system. A low penetration ratio can cause a V2V network to be constantly partitioned [13], [14], and consequently, traditional mobile ad hoc routing schemes can not be used in such a network. (2) Dynamic conditions: vehicles on a highway or a main road can drive at speeds of 40–60 mph. Wireless connection lifetime between two vehicles could be very short. Thus, a receiver vehicle cannot rely on a fixed set of neighbor vehicles for data delivery. (3) Multiple mobile video sources: in many cases, video data is generated from multiple vehicles. Each vehicle is only responsible for a small portion of the data. These vehicles tend to collect data in an uncoordinated way; data redundancy and data discontinuity is inevitable.

We propose V3, an architecture to support video streaming applications in V2V networks. This architecture is composed of two parts: a video source trigger sub-system and a video data transfer sub-system. The video source trigger sub-system uses a novel signaling mechanism to continuously trigger video sources to send video data back to receivers. The video data transfer sub-system adopts a store-carry-and-forward approach to transmit video data in a partitioned network environment [1], [4], [5]. Several algorithms are proposed to balance the video transmission delay and bandwidth overhead. In this paper, we first describe our design of V3 by focusing on a special case where a single receiver is interested in a single destination region [7]. A more general case involves both multiple receivers and multiple destination regions, as shown in Table 1. This paper then studies the impact of multiple receivers on the system design, and proposes a multicasting framework to support a streaming video service to multiple receivers.

This paper is organized as follows. We give the problem description in Section 2. We then propose the design of V3. In Section 3, we focus on the scenario where video data is sent from a single area to a single receiver. In Section 4, we describe the multicast framework that supports video streaming to multiple receivers. In Section 5, we show some performance evaluation results based on simulation experiments that use detailed vehicle traffic models widely used in the transportation research community. Finally, we conclude this paper in Section 6.