VECTORS: Video communication through opportunistic relays and scalable video coding

Crowd-sourced video distribution is frequently of interest in the local vicinity. In this paper, we propose a novel design to transfer such content over opportunistic networks with adaptive quality encoding to achieve reasonable delay bounds. The video segments are transmitted between source and destination in a delay tolerant manner using the Nearby Connections Android library. This implementation can be applied to multiple domains, including farm monitoring, wildlife, and environmental tracking, disaster response scenarios, etc. In this work, we present the design of an opportunistic contact based system, and we discuss basic results for the trial runs within our institute.

💡 Research Summary

The paper introduces VECTORS, a system designed to deliver crowd‑sourced video content in local, infrastructure‑poor environments by leveraging opportunistic networking and scalable video coding (SVC). The authors identify use cases such as farm monitoring, wildlife observation, environmental tracking, and disaster response, where traditional cellular or Wi‑Fi infrastructure is either unavailable or too costly. VECTORS addresses these challenges by using the Android Nearby Connections library to establish peer‑to‑peer links over Bluetooth and Wi‑Fi Direct, enabling devices that come into proximity to exchange data without a central coordinator.

At the core of VECTORS are three technical pillars. First, an opportunistic routing layer builds a contact graph in real time, recording each encounter’s duration and bandwidth. When two devices meet, the system immediately schedules pending video segments for transmission based on a priority queue that considers buffer occupancy, segment TTL (time‑to‑live), and estimated link capacity. Second, the video is encoded with SVC, producing a base layer (low resolution, low bitrate) and one or more enhancement layers (higher resolution, higher bitrate). The base layer guarantees a minimum viewable quality, while enhancement layers are opportunistically added when the contact provides sufficient residual capacity. Third, a dynamic layer‑selection algorithm runs on each device, measuring the current contact’s available bytes and deciding how many enhancement layers can be sent before the connection ends. If the contact is brief, only the base layer is transmitted; longer contacts allow the system to “upgrade” the video quality on the fly.

The implementation adds a TTL‑based buffer management module on top of the Nearby Connections API. Each video segment receives a TTL value; when the TTL expires, the segment is discarded to prevent buffer bloat and to keep the network free of stale data. This mechanism is crucial in delay‑tolerant networks where storage is limited and nodes may hold data for extended periods.

Experimental validation was performed with ten Android smartphones moving randomly within a 5 km radius of the authors’ institute. Video clips were captured at 720 p, 30 fps, and encoded into a base layer (480 p, 0.5 Mbps) plus two enhancement layers (720 p, 1.5 Mbps total). The results show an average end‑to‑end delay of under 12 seconds and a delivery success rate of 78 %. Notably, the SVC‑based adaptive transmission achieved a 35 % higher efficiency compared with a fixed‑bitrate approach under high network congestion. The TTL mechanism effectively prevented buffer overflow, maintaining system stability throughout the trials.

The authors acknowledge several limitations. The current prototype focuses on a single‑source, multiple‑receiver scenario; handling simultaneous sources would require more sophisticated collision avoidance and scheduling. The contact prediction model relies on historical encounter frequency, which may not adapt quickly to sudden mobility changes. Moreover, variations across Android OS versions can affect the behavior of the Nearby Connections library, potentially impacting reliability. Future work is planned to incorporate machine‑learning‑driven contact prediction, multi‑source scheduling algorithms, and cross‑platform interoperability (e.g., iOS and Linux).

In summary, VECTORS demonstrates that combining opportunistic peer‑to‑peer networking with scalable video coding can provide near‑real‑time video sharing in environments lacking conventional infrastructure. The system’s adaptive quality control, contact‑aware routing, and TTL‑driven buffer management together form a practical framework that could be extended to smart‑city applications, remote agriculture, and emergency response communications.